Tag: AI Visibility

How to Build an AI Brand Claim Correction Workflow

An AI answer says your product lacks a feature it has, assigns your company to the wrong owner, or repeats a policy you retired. The tempting response is to regenerate the answer until it looks right. That may produce a better output, but it does not tell you whether the underlying claim has been corrected.

You need a workflow that turns a bad answer into a documented case: capture the claim, decide whether it is truly inaccurate, identify the evidence influencing it, correct that evidence where possible, and verify the result without treating one favorable retest as proof.

Capture the claim before anyone starts correcting it

An AI error is not actionable when the entire report is, AI got our brand wrong. Your unit of work should be one exact claim in one observable response. If an answer contains three inaccuracies, open three claim records. They may have different evidence, owners, risks, and correction paths.

Create the record before editing a page, contacting a publisher, or changing structured data. Otherwise, you lose the baseline needed to determine what changed.

Save the inaccurate sentence verbatim and preserve the surrounding answer. A cropped sentence can hide a qualification that changes its meaning.
Record the exact prompt, AI product or search surface, visible model name if one is provided, response mode, language, location, and any account or personalization setting that could affect the result.
Add the capture date, a screenshot, and the full response in a durable format. Redact personal or confidential information before sharing the case outside authorized systems.
Save every citation, linked page, domain, and quoted passage returned with the answer. Note explicitly when no citation is shown.
Write the correct replacement claim in one sentence. Avoid promotional wording; state the narrow fact you can prove.
Attach the evidence supporting that replacement, including the authoritative URL, page section, document owner, and effective date where one exists.

Then run a small, fixed baseline set. Include the original prompt, a natural paraphrase, and the adjacent question a prospective customer is likely to ask. If the problem appeared in a comparison query, include both the comparative and standalone brand forms. Log each response separately.

Do not combine different AI products, model modes, languages, or countries into one result. A claim that appears on one surface and not another is still worth recording, but it is not evidence that every system holds the same representation. Likewise, a single occurrence establishes that the error happened; it does not establish how prevalent it is.

Classify the failure while the evidence is fresh. Useful labels include fabricated, outdated, misattributed, context omitted, source contradicted, and technically true but materially misleading. These labels make the next decision easier because an outdated policy needs a different remedy from a claim invented without a visible citation.

Triage inaccurate claims by harm, evidence, and correctability

Not every unfavorable statement is inaccurate, and not every inaccuracy deserves an urgent campaign. Validate the claim before you send a correction request. If your own product pages disagree, the immediate problem is not the AI system; it is the absence of a stable, supportable brand fact.

Ask four questions in order:

Can you prove the claim is wrong? Identify the specific factual conflict and the dated evidence that resolves it.
What decision could it affect? Consider purchasing, renewal, hiring, partnership, compliance, safety, and reputation rather than relying on how embarrassing the answer feels.
How broadly does it recur? Use the fixed prompt set instead of repeatedly improvising prompts until you find either the answer you want or the answer you fear.
Is there a correctable evidence path? A cited publisher page, outdated first-party page, incorrect profile, or contradictory product document gives you a concrete target. An uncited answer requires investigation before outreach.

Use three practical queues. Put objectively false claims with serious commercial, safety, regulatory, or reputational consequences in the urgent queue. Put material but lower-consequence errors with identifiable evidence in the planned queue. Monitor isolated, low-impact, ambiguous, or genuinely subjective statements until you have enough evidence to act.

Do not submit a factual correction simply because an answer is negative. A documented limitation, a supported criticism, or an opinion cannot be repaired by replacing it with brand copy. Correct the underlying fact, supply missing context, or respond through the appropriate communications process.

Claims alleging fraud, criminal conduct, regulatory violations, dangerous behavior, or other matters with legal consequences need special handling. Preserve the complete evidence, restrict internal circulation where appropriate, and have qualified counsel approve any external demand. A hurried accusation or an attempt to remove relevant records can create a larger problem than the AI answer itself.

Choose the evidence layer that can actually be corrected

An AI response is an output, not a single brand profile you can open and edit. Your correction target is usually an evidence layer that the system found, cited, retrieved, or learned from. Begin with the citations in the response, then work outward to exact wording searches, first-party content, structured data, public profiles, and other pages that repeat the same claim.

Observed pattern	Likely correction target	First action
The answer cites an inaccurate third-party page	The cited publisher or data owner	Prepare a narrowly scoped correction request with the exact passage, replacement wording, and proof
The answer cites an outdated page you control	Your canonical product, policy, company, or documentation page	Correct the visible content and reconcile every owned page that contradicts it
Several sources publish conflicting versions	The broader evidence set	Establish one canonical fact, update owned properties, and approach the most consequential external sources separately
No citation is visible	Still unknown	Search for the exact phrasing and distinctive fragments, inspect owned content, and collect more logged responses before assigning a target
The statement is technically true but missing a decisive qualification	Content clarity and context	Publish the qualification beside the claim rather than relying on a distant disclaimer

First-party consistency matters because machines and people should not have to decide which of your pages is current. Pick one canonical location for each important brand fact. State the fact plainly, name its scope, add an effective or updated date when timing matters, and link supporting documents from that location. Remove or revise contradictory wording across product pages, help content, press materials, policy pages, downloadable files, and public profiles you control.

Use JSON-LD to express facts that are already visible and supportable, not to create an alternate machine-only version of the brand. Organization, Product, and Offer markup can clarify entities and properties, but markup is not proof by itself and cannot repair an inaccurate publisher page. Keep structured data aligned with the visible page and your canonical record. If the prose says one thing and the schema says another, you have introduced another conflict.

Third-party errors require a source-level correction. Identify who can change the exact record: an editor, database operator, directory owner, review platform, syndication partner, or other publisher. Do not send a general reputation complaint when you can point to a sentence, explain the factual defect, and provide a supported replacement.

A vendor-announced integration connects inaccurate-claim flags from FactCheck with Noble’s Mention Refresh for source-correction work. The useful pattern is the handoff: detection should create an evidence-backed correction task, not end at a dashboard alert. That integration is not evidence that every publisher will accept a request or that every AI output will change afterward.

Run the correction as a controlled handoff

The handoff is where most correction programs become vague. Monitoring finds an error, communications assumes SEO owns it, SEO assumes legal or product has approved the replacement, and nobody has authority to contact the source. Assign four responsibilities for every validated case, even if one person fills more than one role:

The claim owner decides what the correct, supportable brand fact is.
The evidence owner supplies the records that prove it.
The correction owner updates an owned property or contacts the external source.
The verification owner reruns the fixed test set and decides whether the closure rule has been met.

Package the case so the correction owner does not have to reconstruct it. A complete correction packet should contain:

A short case title naming the entity, incorrect claim, and affected surface.
The verbatim AI claim, original prompt, capture details, and full response.
The URL and exact passage believed to support or repeat the error.
A neutral explanation of why the passage is inaccurate or incomplete.
The smallest replacement wording that resolves the defect.
Links or attachments proving the replacement, with an internal approver named.
The requested action, responsible owner, priority, and next review point.

For a page you control, make the correction visible in the main content. Reconcile page titles, summaries, downloadable files, structured data, and related documentation where they repeat the old claim. Preserve any record your legal, compliance, or archival obligations require. When an old URL must remain available, add clear current context instead of silently leaving obsolete wording to circulate.

For an external page, keep the request factual and easy to process. Name the URL and passage. Explain the error in one short paragraph. Supply the replacement and direct evidence. Ask for confirmation when the page changes. Do not mix a correction request with a demand for a promotional backlink, preferred positioning, or removal of an accurate criticism; that obscures the factual issue.

Automation can create the case, attach captures, route approvals, assign owners, and schedule follow-up. It should not invent the replacement fact or send consequential external messages without review. The risky step is not copying fields between systems. It is deciding what the public record should say.

Use explicit workflow states: detected, validating, validated, target identified, correction approved, submitted, source changed, retesting, closed, and monitor only. Require an artifact for each important transition. Validation needs proof. Submission needs a copy of the request. Source changed needs a before-and-after record. Closure needs the retest log.

Separate the source task from the AI-output task. The source task can close when the target page or record is corrected. The output task stays open until your verification rule is satisfied. This distinction prevents a successful outreach email from being mistaken for a corrected brand representation.

Verify the result without overreading one clean answer

A corrected page does not guarantee an immediate or universal change in generated answers. The system may retrieve another page, use a different response path, preserve older information, or vary its wording from one run to the next. Do not promise a universal refresh time when the product, model mode, retrieval behavior, and evidence path can differ.

Retest against the baseline you saved. Use the same prompts, settings, language, and surface first. Then run the approved paraphrases and adjacent questions. If several AI products matter to your business, treat each one as a separate test panel rather than averaging them into a reassuring overall result.

At each checkpoint, record the answer, whether the inaccurate claim appeared, which qualification was present, and what the response cited. This produces four meaningful outcomes:

The source is corrected and the claim disappears across repeated checks. Keep the evidence and move the case toward closure.
The source is corrected but the claim persists. Investigate other cited pages, repeated phrasing, cached copies, and conflicting owned content before reopening outreach to the same publisher.
The claim varies between runs. Keep the case in retesting; a favorable generation has not established a stable correction.
The claim disappears but the underlying source remains wrong. Do not close the source task. The error can return or affect another answer.

Measure the workflow rather than claiming credit for every output change. Useful operational measures include the number of validated claims still open, time from validation to source change, share of cases with an identifiable evidence target, recurrence within a fixed prompt panel, and the number of cases reopened after apparent resolution. Define each measure before reporting it, and keep raw counts beside rates when the test panel is small.

Recurrence is especially useful when it has a fixed denominator: erroneous answers divided by completed runs in the same prompt panel at the same checkpoint. Changing the prompts, surfaces, or number of runs midstream makes the before-and-after rate hard to interpret. Add new discovery prompts to the next test version rather than quietly inserting them into the current baseline.

Key takeaways

Preserve the exact claim, response context, prompt, surface, and citations before changing anything.
Validate that the statement is objectively inaccurate; negative, incomplete, and false are different correction cases.
Correct the evidence layer that can be changed, including contradictory first-party content and inaccurate third-party pages.
Give every case a claim owner, evidence owner, correction owner, verification owner, and explicit workflow state.
Close source correction and AI-output verification separately, using repeated checks against a fixed baseline.

Start with the highest-consequence claim for which you already have decisive evidence. Build one complete case, assign its owners, and follow it from capture through repeated verification. That case will expose the missing approvals, evidence gaps, and handoff failures you need to solve before scaling the workflow.

References

Try Profound Blog — Act on inaccurate claims flagged by FactCheck via Noble’s Mention Refresh

July 21, 2026

How Content, Entities and Category Framing Shape AI Visibility

You have useful content, a clean About page and valid organization markup. Yet your brand still disappears when someone asks an AI assistant for options in your market. The missing piece may not be authority. The system may know who you are without considering you eligible for the category named in the prompt.

You can diagnose that problem by separating three jobs: establish the category in which you belong, make the relevant entities and relationships unambiguous, and publish evidence that supports recommending you for the user’s task. That distinction turns AI visibility from a vague branding exercise into work you can assign, test and improve.

Key takeaways

Brand recognition and recommendation eligibility are different. An AI system can identify your company accurately and still exclude it from an unbranded category answer.
Choose category language before planning content or schema. Your primary category should describe what you sell now; adjacent categories should reflect real customer language and a defensible part of your offer.
Build an entity map before building more pages. It should connect your organization, offers, audiences, problems, methods, people, proof and category claims.
Use JSON-LD to declare facts that visible content already supports. Schema can reduce ambiguity, but it cannot manufacture relevance or compensate for missing evidence.
Category association is also built away from your website. Relevant reviews, editorial coverage, comparisons and co-mentions help establish the contexts in which your brand is considered.
Measure recognition, category eligibility, recommendation and supporting evidence separately. A single visibility score hides the reason you are being omitted.

First, determine whether you have a recognition or category problem

Start with two prompts that look similar but test different things:

Recognition prompt: What is [Brand], and what does it offer?
Category prompt: Which [category] providers should [audience] consider for [task]?

If the first answer is accurate and the second omits you, rewriting your About page again is unlikely to address the main constraint. Your entity is recognized, but it is not being retrieved or selected in that category context.

Observed result	Likely problem to investigate	Best first check
Your brand is described incorrectly when named	Entity ambiguity or inconsistent facts	Compare names, descriptions, offers and relationships across core pages, markup and authoritative profiles
Your brand is understood but absent from an unbranded category prompt	Weak category association	Inspect the categories used in your own copy and in third-party coverage
You appear for a primary category but not an adjacent one	Category-specific evidence gap	Look for useful content and independent mentions that connect you to the adjacent category
You are included but the recommendation rationale is vague	Weak differentiation or insufficient proof	Identify which claims lack examples, evidence or a clear audience fit
A relevant page is cited but your brand is not recommended	Informational relevance without brand-level eligibility	Check whether the page clearly connects its subject, your offer and the user’s decision

The effect of category wording can be substantial. A controlled test covering 14,140 API runs across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews evaluated 12 athletic apparel brands in the U.K. over seven days. Changing the category from athleisure to athletic footwear moved New Balance from a 1% appearance rate to 90%, while lululemon moved from 90% to 0%.

That is strong evidence that framing controlled recommendation behavior in that test. It is not a universal performance benchmark: one market, one prompt design and one testing period cannot establish how every model will treat every category. The practical lesson is narrower and more useful. Test the category noun instead of assuming that general brand strength transfers across every way a customer might describe your market.

Define one primary category and a small set of adjacent frames

Your primary category should be the plainest accurate answer to: What kind of provider, product or organization is this? An adjacent frame is a different but truthful way a buyer may classify the same offer. For example, a platform may belong firmly to one software category while also serving a narrower workflow, audience or outcome category.

Do not collect every loosely related label. For each candidate category, record:

Customer language: Do real buyers use this term when expressing the need you solve?
Offer fit: Can you point to a current product, service or capability that makes the label true?
On-site evidence: Is the category explained on a crawlable page, or does it appear only in a slogan?
Independent evidence: Do credible third parties discuss you in that context or alongside established members of the category?
Decision value: Would visibility for this category attract the audience and use case you actually want?

Then write a control sentence: [Brand] is a [primary category] for [audience], helping them complete [task] through [offer or method]. Treat this as an editorial constraint, not a slogan and not a Schema.org type. Every element must be demonstrably true, and the same relationship should be understandable from your core pages.

Build an entity map that gives every page a job

Once the category is chosen, map the things a search system must connect to decide that you belong. Entities are not limited to your company and founder. They include products, services, people, audiences, locations, problems, methods, features and other identifiable concepts. The useful unit is not an isolated noun; it is a relationship that helps explain the brand.

Create an entity ledger with one row for each important relationship:

Subject: the organization, person, offer, category, audience or problem being described.
Relationship: offers, serves, solves, teaches, authored, includes, supports or another accurate connection.
Object: the entity on the other side of that relationship.
Visible evidence: the page and passage where a reader can verify the claim.
Structured declaration: the standards-supported markup, if any, that can express it accurately.
Independent corroboration: a review, profile, comparison, citation or other external evidence.
Gap: missing, vague, contradictory or fully supported.

Use those relationship words as planning labels. They are not automatically valid Schema.org properties. Your conceptual model can and often should be richer than the standardized vocabulary you publish.

This distinction matters in specialized markets. One higher-education framework found that 23 existing Schema.org entities were insufficient and added more than 60 domain-specific concepts to represent a prospective student’s journey. You can use a custom ontology internally to expose content gaps without pretending that proprietary terms are recognized Schema.org vocabulary.

Turn the map into a content system, not one oversized page

Assign each important relationship to a canonical page. Your About page should establish organization identity and positioning. An offer page should explain what the offer does, whom it serves and how it differs. A method page should explain the process. A use-case page should connect a specific audience and task to the offer. An author page should establish the person behind relevant expertise. Supporting resources should answer the questions that arise before and after the main decision.

This division helps with the way AI search may expand a request. A query can trigger related searches across subtopics and data sources so that the system can assemble an answer to the broader task. A buyer asking for a category recommendation may also need selection criteria, implementation details, limitations, alternatives, audience fit and next steps. One page does not have to answer everything, but your site should make the connections explicit.

Use this brief for every page you keep or create:

Page job: State the single decision or question this page resolves.
Primary entities: Name the organization, offer, audience, problem and category involved.
Direct answer: Put the answer near the beginning in visible text. Do not make a reader infer it from a slogan, image or schema block.
Boundary: Explain who or what the answer is for, where it applies and what it does not cover.
Evidence: Support claims with concrete capabilities, examples, authorship or other facts you can substantiate.
Related questions: Link to the next useful pages with anchor text that describes the relationship, rather than generic text such as learn more.
Duplication check: Merge or differentiate pages that make the same claim about the same entities without serving different intents.

The standard is comprehension, not length. A clear page names its subject, answers the intended question and connects to the next part of the task. More copy only helps when it adds a missing entity, relationship, condition or piece of evidence.

Use JSON-LD to declare truth, not manufacture relevance

Schema is valuable because it can state entities and relationships explicitly in a vocabulary machines already recognize. It is best treated as a declaration layer over a coherent site, not a lever that forces a model to recommend you.

The evidence does not support a simple claim that adding markup produces more AI citations. Microsoft Bing’s Fabrice Canel stated in March 2025 that Copilot uses schema to understand content, while other published tests found no effect on LLM visibility or no direct reading of on-page schema. Those findings measure different things, including machine understanding, direct model access, citations and observed visibility. Treating them as one outcome creates a false yes-or-no debate.

A safer operating position is straightforward: accurate markup can reduce ambiguity for systems that consume it, but visibility remains a downstream result influenced by content, retrieval, category fit and external evidence. Do not promise a citation lift from markup alone.

Implement JSON-LD in this order:

Resolve identity first. Decide which organization, people, offers and other entities are canonical. Use stable identifiers so the same entity is not represented as several disconnected things.
Confirm the visible facts. A reader should be able to verify every material claim in the markup from the page or an appropriate linked page. Structured data should match what users can actually see.
Use established vocabulary where it fits. Choose the most accurate standard types and properties available. Do not force a marketing phrase into a technical type merely because the phrase is commercially important.
Connect entities deliberately. Markup should describe a coherent graph rather than produce unrelated blocks for the organization, author, service and page.
Keep custom concepts separate. Use your internal ontology to plan coverage and analyze gaps. Publish custom terms only where a consuming system understands that vocabulary; do not misrepresent them as standard Schema.org definitions.
Remove decorative markup. If a block exists only to qualify for a feature or repeat keywords, but adds no accurate entity relationship, it is not solving your AI visibility problem.

When markup and visible copy disagree, repair the underlying page first. Otherwise you are making two incompatible claims about the same entity and asking machines to decide which one is true.

Create off-site category evidence, then measure the whole system

Build corroboration in the category you want to earn

Your site can declare its category, but it cannot independently establish how the wider market describes you. Category coding appears to combine an entity anchor with the third-party material accumulated around a brand, including reviews, editorial comparisons, roundups and co-mentions. This helps explain why editing a description does not instantly move a brand into a different recommendation set.

Audit the external evidence for each priority category:

Which publications, communities and comparison pages appear in AI answers for the category?
Which brands are repeatedly mentioned together, and what language is used to explain their inclusion?
Which attributes make a provider category-eligible: audience, use case, product form, method, price position or another verifiable characteristic?
Where is your brand already mentioned, and which category does that coverage reinforce?
Does the cited coverage still describe your current offer accurately?

Use the findings to shape public relations and content distribution. Give relevant publishers a truthful reason to place your brand in the target context: a category-specific capability, credible expert contribution, useful case evidence or a clear point of view. A generic mention may improve recognition while doing nothing to connect you to the category that matters.

Do not pursue an adjacent category that your product cannot support. Repetition can amplify an association, but it cannot make a misleading position useful to the customer. Establish the offer and on-site evidence before trying to earn external corroboration.

Measure recognition, eligibility, recommendation and evidence separately

Create a controlled prompt matrix for every primary and adjacent category. Keep the audience, task and wording stable, then change only the category expression you want to test. Run each prompt in a fresh conversation so earlier messages do not supply the brand or category context.

Record these fields for each model and prompt:

Recognition: Can the system describe your brand accurately when it is named?
Eligibility: Does the brand appear in an unbranded list for the category?
Recommendation: Is it merely mentioned, or actively presented as suitable for the audience and task?
Rationale: Which capabilities, use cases or associations explain its inclusion or exclusion?
Evidence: Which URLs, publishers or page types support the answer?
Representation: Are the description, category and sentiment accurate?
Conditions: Which model, prompt, date and conversation state produced the response?

Do not compress these observations into one score until you have inspected them separately. A brand that is recognized everywhere but eligible nowhere has a different problem from one that is regularly recommended with the wrong description.

Use the pattern to choose the next action:

Recognition is weak: reconcile identity, core descriptions, canonical pages, profiles and structured relationships.
Recognition is strong but category eligibility is weak: repair category language and build relevant third-party association.
Eligibility is strong but recommendation is weak: clarify audience fit, differentiation, limitations and supporting proof.
Recommendation is strong but evidence is poor: strengthen pages that make the rationale attributable and easy to cite.
Results differ sharply by category: plan content and outreach for each frame independently instead of treating visibility as a brand-wide property.
Results differ sharply by model or prompt: preserve the raw responses and gather more controlled observations before declaring a trend.

Prioritize gaps using three questions: Does this category matter commercially? Is the missing association visible across controlled prompts? Can you support it truthfully with your present offer and evidence? A high-volume label that fails the third test is not an optimization opportunity. It is a positioning error.

Start with one primary category and one defensible adjacent frame. Run the prompt matrix, map the entities behind both, assign each important relationship to a page, align visible copy with JSON-LD, and then pursue independent coverage in the context that is still missing. That sequence gives you something more useful than a visibility score: a reason for the result and a specific next move.

References

July 21, 2026

How to Use Profound Aim Brainstorm Mode Productively
You can have useful AI Search data and still face a blank next step. The data may expose several promising directions, but it cannot choose which uncertainty your team should resolve first.

Brainstorm Mode within Profound Aim is designed for that handoff: it guides a broad goal toward scoped, ready-to-run Agents. The practical value is not producing more ideas. It is reducing the distance between an ambition and a task that can inform a real decision. To get that value, you need to give Brainstorm Mode strategic direction without prematurely prescribing the analysis.

Use Brainstorm Mode to close a decision gap

Brainstorm Mode is most useful when you know the outcome you want but do not yet know what an Agent should investigate. That is a decision gap: your team has a business objective and relevant data, but the next analytical question remains unclear.

Good reasons to start in Brainstorm Mode include:
- You can describe the business outcome, but several parts of the AI Search data could be relevant.
- You have noticed a visibility pattern and need to decide which part deserves deeper investigation.
- Different teams are proposing different explanations for the same result.
- You need to turn a broad AI visibility priority into work that has a clear boundary.
- You know someone can act on the answer, but you have not yet defined the question that would produce it.
Brainstorming adds less value when the task is already precise. If you know the exact question, scope, evidence and required output, you may already have an Agent brief. Starting another ideation cycle can introduce ambiguity that was not there before.

There is a simple readiness test: complete the sentence, “When this Agent finishes, we will decide whether to ______.” If you cannot fill the blank with a decision your team is prepared to make, the problem is not Agent scope yet. You still need alignment on the purpose of the work.

Give Aim a broad goal without giving it an empty one

Broad and vague are not the same. A broad goal leaves room to discover the right investigation. A vague goal hides the decision, audience and boundary that make an investigation useful.

“Improve our AI visibility” is vague. It does not say which part of the business matters, what kind of visibility problem is in scope or what anyone will do with the result. Brainstorm Mode may still be able to propose work, but you will have no strong basis for judging whether that work matters.

A useful goal normally contains these ingredients:
- Outcome: the change you want to support, such as choosing a content priority or understanding a visibility weakness.
- Business scope: the brand, offering, product area or customer problem that matters.
- Audience scope: the market, language, geography or buyer context that should govern relevance.
- Decision: what the team expects to choose after seeing the evidence.
- Evidence boundary: what the available AI Search data can reasonably help examine.
- Constraint: what should remain outside the first investigation so the Agent does not become an entire strategy project.
You can assemble those ingredients with this reusable structure:

Help us decide [decision] for [brand, offering or audience] by using our AI Search data to investigate [uncertainty]. Keep the first Agent focused on [scope], and produce evidence we can use to [next action].
Goal-framing template

For example, replace “Improve our AI visibility” with: “Help us decide which content area should receive the next optimization effort. Use our AI Search data to investigate where visibility is weakest within the product area we plan to grow, and keep the first Agent focused on identifying and characterizing the gap rather than recommending a complete content strategy.”

The improved version is still broad enough for Brainstorm Mode to shape the work. It also supplies a decision, a business boundary and a stopping point. That stopping point matters. Without it, one Agent can easily become responsible for finding a problem, explaining it, designing a strategy, writing content and evaluating results. Those are different jobs with different evidence requirements.

Review every proposed Agent as a research brief

“Ready to run” describes an operational state, not automatic strategic importance. Before running a proposed Agent, make sure its result could actually change what you do. A technically valid investigation can still be too broad, unanswerable from the available data or disconnected from the decision owner.

Use this pre-run check:
- One primary question: Can you express the Agent’s job as one question without joining several assignments with “and”?
- Defined boundary: Does the brief identify the relevant brand, topic, audience or market while excluding unrelated areas?
- Available evidence: Can the AI Search data support the requested analysis, or is the Agent being asked to infer facts the data does not contain?
- Usable output: Will the result help someone choose, prioritize, approve, reject or investigate something specific?
- Inference discipline: Does the brief distinguish observed patterns from possible explanations?
- Named owner: Is there a person or team prepared to use the result?
Break apart bundled Agents

A bundled Agent might be asked to find every visibility gap, explain every cause, compare all relevant competitors, build a content strategy and produce implementation briefs. It sounds comprehensive, but each stage depends on choices made in the previous one. If the first interpretation is weak, every later deliverable inherits the problem.

Start with the smallest question that can change the next action. An initial Agent might identify and characterize an in-scope visibility gap. A later Agent can investigate evidence-linked explanations for the selected gap. Content planning should begin only after you decide that the gap is important enough to address.

This sequence also makes poor outputs easier to diagnose. You can tell whether the difficulty came from the goal, the data boundary, the interpretation or the proposed action instead of debugging one oversized deliverable.

Separate observations from explanations

AI Search data can reveal a pattern. A pattern does not, by itself, prove why that pattern exists. “The brand appears less often for this topic” is an observation. “The brand appears less often because of a particular content weakness” is an explanation that still needs support.

If a proposed Agent asks why something is happening, require it to distinguish direct evidence from inference. The useful output is not an unsupported diagnosis stated confidently. It is a set of plausible explanations connected to the available evidence, with the remaining uncertainty made visible. That gives your team something it can test instead of a conclusion it can only accept or reject.

Turn the first Agent into a controlled decision loop

The fastest way to create a pile of unused analysis is to run every plausible Agent at once. The outputs arrive without an order of operations, overlap in scope and often answer questions that no longer matter after the first decision.

Use Brainstorm Mode as the beginning of a controlled sequence:
1. Write the decision sentence: “When this Agent finishes, we will decide whether to ______.”
2. Frame the broad goal around that decision and the relevant AI Search data.
3. Use Brainstorm Mode to translate the goal into a proposed Agent or set of Agents.
4. Apply the pre-run check and select the smallest Agent whose result could change the decision.
5. Run that Agent before commissioning downstream analysis.
6. Record the finding, the interpretation and the decision as separate items.
7. Create another Agent only when the decision exposes a new uncertainty that must be resolved.
A working note for each completed Agent can remain short:
- Finding: What is directly supported by the output and underlying data?
- Interpretation: What might the finding mean, and which part remains an inference?
- Decision: What will the team do, defer or reject because of the finding?
- Owner: Who is responsible for the next action?
- Validation: What later AI Search signal would help determine whether the action had the intended effect?
Consider a team deciding which product area deserves its next content investment. The first Agent could identify which in-scope topic area shows the most decision-relevant visibility weakness in the available data. The team then selects a topic based on business importance, not merely the size of the gap. A second Agent, if needed, can examine answer patterns for that topic and organize evidence-linked hypotheses. Only then does the team choose a content intervention and define how it will evaluate the result.

That order preserves human judgment at the points where data cannot make the business choice. Brainstorm Mode helps structure the investigation; it does not remove the need to decide which market, audience, risk and opportunity matter.

Key takeaways
- Use Brainstorm Mode when you have a meaningful AI Search goal but have not yet converted it into an answerable investigation.
- Frame the goal around a decision, business boundary, audience and evidence source instead of asking generally for better visibility.
- Reject proposed Agents that combine discovery, diagnosis, strategy, production and measurement in one assignment.
- Make every Agent distinguish data-backed observations from explanations that remain hypotheses.
- Run the smallest useful Agent first, make a decision and generate follow-up work only when a new uncertainty appears.
Before you open Brainstorm Mode, write one sentence: “When the first Agent finishes, we will decide whether to ______.” Use that decision to frame the goal you bring into Aim. If the blank is still empty, pause the Agent design and settle the business question first.

References
- Try Profound Blog — Introducing Brainstorm Mode within Aim
July 20, 2026

How to Choose a Generative Engine Optimization Agency

If you are choosing a generative engine optimization agency, finding candidates is the easy part. The difficult part is deciding whether a firm can improve your visibility in AI-generated answers or has simply put a GEO label on its existing SEO package.

You need a proposal that connects questions your buyers ask to sources an answer engine can retrieve, understand, trust, and cite. You also need measurement you can audit. The framework below will help you test both before you sign a long engagement.

Key takeaways for choosing a GEO agency

Hire for an operating system, not a label. The agency should connect audience research, content, technical access, entity clarity, external authority, and measurement.
Require a reproducible baseline built from a defined set of questions, answer environments, markets, and evaluation rules.
Ask to see the evidence chain from observed problem to recommendation, implemented change, later answer, and business interpretation.
Treat schema markup as a supporting layer. JSON-LD can clarify what a page describes, but it cannot manufacture authority or guarantee a citation.
Reject guaranteed mentions, citations, rankings, or recommendations. An agency can influence the inputs to an answer system, but it cannot control the answer selected for every user.
Start with a bounded, commercially meaningful scope. Expand only when the agency can show its work and your team can verify the resulting evidence.

What a real GEO agency should actually own

Generative engine optimization is the work of improving how accurately and often a company, product, service, or expert is represented in AI-generated answers. It overlaps with SEO, but the unit of performance changes. A conventional search program often concentrates on pages and rankings. GEO must also examine whether an answer system retrieves the right information, understands the entity behind it, includes the brand in the relevant context, and cites an appropriate source when citations are shown.

The specialist label alone proves little. In 2026, buyers can already compare seven firms presented as GEO agencies. That makes the label a useful way to build a shortlist, but not evidence that a particular agency has a distinct method.

A credible scope should connect the following workstreams:

Audience-question mapping: The agency identifies the questions that matter before, during, and after a buying decision. It groups them by intent instead of treating every prompt containing your category name as equally valuable.
Baseline visibility: It records where your brand appears, where competitors appear, which sources are cited, and whether the resulting description of your business is accurate.
Content and evidence planning: It finds missing definitions, explanations, comparisons, proof points, policies, product details, and expert material. Each recommendation should answer a documented information need rather than merely add more words to the site.
Technical accessibility: It checks whether the intended pages are discoverable, indexable, internally connected, and available to the retrieval systems included in the engagement. A page cannot support an answer if the relevant system cannot reach or interpret it.
Entity and structured-data work: It aligns names, descriptions, relationships, authorship, organization details, and supported schema markup with the visible content. Markup should describe evidence that actually exists on the page.
External corroboration: It considers reputable third-party mentions, reviews, profiles, expert contributions, public relations, and other off-site signals. Publishing a claim on your own domain does not automatically make that claim persuasive.
Measurement and iteration: It repeats a documented evaluation process, connects changes to observations, and tells your team what to keep, revise, investigate, or stop.

These workstreams cross organizational boundaries. Content teams control explanations. Developers control templates and access. Communications teams influence external mentions. Subject-matter experts validate claims. A serious agency identifies those dependencies in the proposal and assigns an owner to each action. A vague promise to “optimize your site for LLMs” is not an implementation plan.

Use the rebranded-SEO test

Ask the agency to show a recommendation it would make specifically because of AI-answer behavior, then ask how it would measure the effect. The response should go beyond adding keywords, publishing generic articles, or installing schema across the site.

A defensible answer might involve a missing question class, an inaccurate entity relationship, a source routinely used in relevant answers, an unsupported claim, weak external corroboration, or a page that is available to search engines but unsuitable for direct answer extraction. The agency should be able to show the observation that led to the recommendation and the evidence it would inspect afterward.

This does not make traditional SEO irrelevant. Useful pages still need clear information architecture, accessible content, descriptive headings, internal links, and credible evidence. The warning sign is an agency that either treats GEO as identical to SEO or presents it as a complete replacement for SEO. The work overlaps, but the questions being measured are not identical.

Demand an AI-visibility measurement system you can audit

AI-generated answers can vary with the wording of a question, the interface used, available retrieval features, market, language, and evaluation date. A collection of favorable screenshots is therefore not a baseline. It is a collection of examples.

Before accepting an agency’s visibility score, ask for the measurement protocol behind it. The protocol should define:

Answer environments: Which models, search experiences, assistants, modes, or features are included? Which are explicitly outside scope?
Question set: What exact questions are monitored? How were they selected, and which audience, buying stage, product line, or market does each represent?
Core and exploratory questions: Which questions stay stable so you can compare observations over time, and which may change as new customer language or opportunities emerge?
Evaluation context: What language, location, account state, date, and other relevant settings are recorded with each observation?
Classification rules: What counts as a mention, recommendation, citation, accurate description, competitive inclusion, or absence?
Evidence archive: Does the agency preserve the exact question, raw answer, cited URLs, evaluation context, and timestamp rather than only a derived score?
Change log: Can you see which pages, claims, markup, links, or external activities changed between measurement periods?

The denominator matters as much as the result. “We increased citations” is not interpretable unless you know how many eligible responses were evaluated, whether the monitored questions stayed comparable, and whether branded questions were mixed with non-branded discovery questions. A brand should naturally appear more often when its name is already in the prompt. That does not prove improved discovery.

Ask the agency to separate several kinds of outcomes:

Brand inclusion: The brand appears in responses to relevant, eligible questions.
Owned-source citation: An eligible answer cites a page controlled by your organization.
Representation accuracy: The answer correctly describes what you offer, who it is for, and any important limitations.
Competitive consideration: The brand appears in a relevant comparison or recommendation context, not merely in a list created by a branded question.
Source quality: Citations point to the most appropriate current page rather than an outdated, weak, or unrelated URL.
Downstream behavior: Referral visits, engaged sessions, qualified inquiries, assisted conversions, or other agreed business signals move in a useful direction.

Do not collapse all of these into a single proprietary visibility number. A composite score may be convenient for reporting, but you should still receive the underlying records and definitions. Otherwise, you cannot tell whether a change came from broader discovery, more branded prompting, a modified scoring formula, or a genuine improvement in how the brand is represented.

Business attribution also needs restraint. An AI answer may influence a buyer without producing a trackable click, while a referral visit may occur without causing a sale. Ask the agency to report visibility indicators and commercial outcomes separately, then explain the plausible connection without presenting correlation as proof of causation.

Score every agency proposal against the same evidence

Marketing language makes proposals difficult to compare. A common scorecard forces each agency to reveal its method, implementation assumptions, and reporting limits. Use the same criteria for every finalist and request supporting examples wherever a claim remains abstract.

Area	What an acceptable proposal contains	Warning sign
Scope	Named answer environments, markets, languages, products, audiences, and question groups	Promises visibility “across AI” without defining where or for whom
Baseline	A reproducible method, recorded context, raw observations, and clear classification rules	A visibility score or screenshots with no query set, denominator, or methodology
Strategy	Prioritized hypotheses linking visibility gaps to specific content, technical, entity, or authority work	A generic publishing calendar produced before the visibility gaps are examined
Content	Question-level briefs, evidence requirements, expert review, update rules, and a defined approval process	High-volume AI-generated pages treated as the main deliverable
Technical work	Checks for access, indexability, rendering, internal discovery, canonical signals, structured data, and implementation ownership	Schema installation presented as a complete GEO strategy
External authority	A plan for relevant third-party corroboration with editorial standards and approval controls	Guaranteed placements, undisclosed paid mentions, or citation schemes
Reporting	Raw evidence, change logs, limitations, business context, and next actions	A dashboard that shows movement but cannot explain what changed
Commercial terms	Deliverables, responsibilities, tool costs, data ownership, exit rights, and change-control terms	A long commitment before the method, baseline, and implementation dependencies are visible

Ask questions that force the method into the open

A polished presentation can hide an undeveloped process. These questions require the agency to move from claims to inspectable work:

Which specific answer experiences are included, and why do they matter to our buyers?
How will you build the monitored question set, and how will you prevent branded prompts from inflating the result?
What raw data will we receive behind every score?
Can you walk us through a sanitized example from observed answer to diagnosis, recommendation, implementation, and later evaluation?
How do you distinguish an owned-page problem from a lack of third-party corroboration?
Which recommendations will require developers, subject-matter experts, legal reviewers, communications teams, or product owners?
How do you verify factual claims before publishing or marking them up?
What work will you refuse to do because it is unreliable, misleading, or likely to create reputational risk?
How will you report an answer that mentions us often but describes us inaccurately?
Which tools, question sets, observations, content briefs, and reports can we export when the engagement ends?
What evidence would make you advise us not to expand the program?

The final question is especially revealing. A consultancy should have a stopping rule. If every possible result leads to a larger retainer, the measurement system is serving the sale rather than the decision.

Treat guarantees as a control problem, not a bonus

No agency controls how an independent answer system generates every response. Guarantees of permanent citations, universal coverage, or fixed recommendation positions should therefore reduce your confidence, not increase it.

Ask for controllable commitments instead: audits completed, questions mapped, pages improved, factual evidence reviewed, markup validated, outreach approved, observations recorded, and reports delivered. Then evaluate whether those actions improve the agreed indicators. This keeps the contract enforceable without pretending the agency controls a third-party model.

Structure the first engagement so you can inspect the work

A bounded first engagement is not merely a cheaper version of a retainer. It is a way to test whether the agency’s diagnosis, execution, and measurement connect. Choose a commercially meaningful topic area with enough existing evidence to examine, then define what the agency must deliver before expansion is considered.

Your kickoff document should contain:

A clear business objective and the audience decisions connected to it
The products, services, markets, and languages in scope
The approved question set and baseline protocol
A record of current brand mentions, citations, inaccuracies, and important absences
A prioritized backlog with an owner, dependency, rationale, and acceptance condition for each action
Rules for factual review, brand approval, technical deployment, and external communications
A change log connecting completed work to the pages or assets affected
Conditions for expanding, revising, pausing, or ending the work

Do not define acceptance as a guaranteed position in an AI response. Define it through deliverables the agency controls and observations your team can verify. For example, an important question gap can lead to an evidence-backed page, expert approval, correct technical implementation, inclusion in the monitoring set, and a documented follow-up evaluation. Visibility movement can then inform the decision to continue, but it is not fabricated into a contractual certainty.

Protect the assets and access your team will need later

The contract should say who owns the question taxonomy, raw response records, scoring definitions, dashboards, content briefs, written content, schema specifications, technical documentation, outreach records, and reporting history. It should also state which formats you can export without the agency’s proprietary platform.

Clarify third-party software fees, data-retention limits, credential handling, approval requirements for automated publishing, and the process for removing access at the end of the engagement. If the agency will contact publishers, customers, partners, or experts in your name, require an approval workflow. Poor outreach can create a reputational cost long after the campaign ends.

Include a handoff requirement as well. Your team should leave with the current measurement protocol, unresolved issues, deployed changes, pending outreach, known limitations, and the next recommended decisions. A dashboard login that disappears on termination is not a usable knowledge transfer.

Send every shortlisted agency the same brief and score each response against the table above. Then ask the finalists to walk a sample question through their complete evidence chain. Choose the firm that makes its assumptions, data, dependencies, and limits easiest to inspect. If that chain is unclear before the contract, a more elaborate report will not make it clearer afterward.

References

HiGoodie Blog – Top Generative Engine Optimization (GEO) Agencies

July 20, 2026

Profound Claude Connector: A Practical AI Visibility Workflow
If you have connected Profound to Claude and are staring at an empty conversation, do not begin with a broad request such as “analyze our AI visibility.” That leaves Claude to choose the scope, comparisons, and standard of proof. The response may sound decisive while answering a different question from the one your team needs resolved.

Profound is now available as an official Anthropic connector. The practical opportunity is a shorter path from authorized Profound data to analysis inside Claude. You still need to define the decision, verify what the connection exposes, and keep measured evidence separate from Claude’s interpretation.

What the Profound connector changes – and what it does not

Treat the connector as an access layer, not a new measurement system. Profound remains the origin of the connected data. Claude can help you inspect, organize, compare, and explain what the connection returns. It cannot recover fields that were not returned, repair an inappropriate comparison, or turn correlation into proof of causation.

Four boundaries matter in every conversation:
- Account boundary: confirm which Profound account or workspace is connected. A polished analysis of the wrong property is still wrong.
- Field boundary: establish which records, metrics, dimensions, and identifiers Claude can actually access. Do not assume that every object visible in Profound is available through the connector.
- Filter boundary: record the market, language, AI platform, topic, brand, competitor set, and date range whenever those dimensions are present. A change in scope can create an apparent performance change.
- Interpretation boundary: separate returned measurements from explanations proposed by Claude. The former can be verified against Profound; the latter are hypotheses until checked.
Official connector status should not be interpreted as a promise of complete data coverage, live refreshes, write access, or a particular permission model. Verify those details in your own connected environment instead of building a workflow around assumptions.

Your first message should therefore be an inventory request:

Starter prompt: Inspect the Profound connection available in this conversation. List the accounts or workspaces, record types, fields, filters, date ranges, and identifiers you can access. Distinguish fields you can retrieve from fields you are inferring. Do not begin the analysis yet. Tell me which parts of the requested scope cannot be verified from the connection.

Save the answer with the analysis. It becomes a compact data contract: a record of what Claude could see when it produced the result. If Claude cannot identify the available scope clearly, resolve the connection or permissions question before asking for strategy.

Scope the decision before you scope the data

A useful connector workflow starts with a decision, not a dashboard tour. “Understand our visibility” is not a decision. “Choose which topic cluster should receive the next content update” is. The second version tells Claude what evidence to prioritize and gives you a clear way to reject irrelevant analysis.
1. Name the decision. State what will change if the analysis supports it: a content update, a new page, a technical investigation, a brand-entity correction, or continued monitoring.
2. Name the entity. Use the exact brand, product, property, or business unit you intend to evaluate. Add aliases only when you deliberately want them included.
3. Set the comparison. Supply an approved competitor list or ask Claude to analyze the brand alone. Do not let the model silently invent a comparison set.
4. Lock the scope. Specify the topic, audience, market, language, AI platform, and time window that matter. If a requested dimension is unavailable, require Claude to say so rather than substitute another one.
5. Define acceptable evidence. Require every conclusion to point to returned fields, records, citations, or other traceable identifiers. Anything else must be labeled as an inference or a proposed next check.
A reusable control prompt can carry those rules into the rest of the conversation:

Control prompt: Use only information returned through the connected Profound account and context I explicitly provide. Preserve the available date range and filters. For every finding, show the supporting field or record identifier. Put measured observations, interpretations, and recommended actions in separate sections. Mark missing data as missing; do not estimate it. Ask for clarification when a missing input would change the decision.

Before using connected business data, also confirm who is permitted to access the selected workspace, whether the conversation may be shared, and what information can be placed in prompts under your organization’s policies. A connector reduces manual transfer; it does not remove your responsibility to control sensitive data.

Three workflows that produce defensible AI visibility actions

1. Find a visibility gap without inventing its cause

The most useful gap analysis identifies where a brand underperforms within a defined set of prompts or topics. It does not immediately claim to know why. Visibility can differ alongside many variables, and the connector alone does not establish which variable caused the difference.

Diagnostic prompt: For [brand], analyze [topic] in [market and language] across [available time window]. Compare it with [approved competitors] only where equivalent comparison data exists. Rank the most consistent visibility gaps. For each gap, return: the observed result, the fields or records supporting it, the scope and filters, one or more plausible explanations labeled as hypotheses, and the next evidence needed to test each explanation. Do not present a hypothesis as a finding.

Review the output in that order. First decide whether the observation is supported. Then check whether all compared entities use the same filters and coverage. Only after those checks should you consider the proposed explanations. This prevents an appealing theory about content quality, authority, or entity recognition from outrunning the connected data.

2. Turn prompt and citation signals into a content brief

If the connection returns prompt-level answers, cited domains, URLs, or related records, Claude can organize those signals into editorial questions. Make the availability of those fields a condition of the task. A domain name in a generated explanation is not evidence that the domain appeared in Profound.

Content-opportunity prompt: From the records available through Profound, find recurring prompts about [topic] where [brand] is absent, represented weakly, or trails [approved competitors]. If citation fields are available, show the exact cited domains or URLs and their associated records. Group the prompts by user intent rather than by shared keywords. For each group, propose one content action tied directly to the observed gap. Label any claim about why another page was selected as a hypothesis unless its page content is also available for inspection.

Translate the result into a brief with five required fields:
- User question: the specific decision or problem represented by the prompt group.
- Observed gap: what the connected records actually show about the brand.
- Evidence: the record, metric, answer, citation, or identifier supporting the gap.
- Page action: update an existing answer, create a missing resource, clarify an entity relationship, or investigate a technical obstacle.
- Validation condition: what comparable Profound signal you will inspect after the action has had an opportunity to appear in the available data.
Do not treat every missing brand mention as a reason to publish another page. If an existing page already answers the intent, the next step may be to improve its clarity, structure, supporting evidence, or entity references. If the connected data cannot distinguish among those possibilities, use it to prioritize an investigation rather than to prescribe the edit.

3. Compare periods without turning movement into causality

Trend analysis is only defensible when the compared records use equivalent scope. A different prompt set, market, platform, competitor group, or coverage level can make two periods look comparable when they are not.

Monitoring prompt: If date-stamped Profound records are available, compare [period A] with [period B] using the same brand, topic, market, language, platform, prompt set, and competitor filters. Identify any dimension that is not equivalent before calculating or describing change. Report observed direction and magnitude only from returned values. Do not attribute movement to a content release, campaign, algorithm change, or competitor action. List those events separately as possible explanations that require additional evidence.

Use the same saved prompt for future checks, changing only the intended date window. If the accessible schema or coverage changes, note the break instead of joining the results into one uninterrupted trend. Consistency is what makes a connector-based monitoring workflow useful; a fluent narrative cannot compensate for mismatched inputs.

Build an evidence trail from conversation to action

Claude’s final answer should not become the only record of the analysis. Preserve enough structure that another person can reproduce the finding in Profound, challenge the interpretation, and understand why an action was approved.
1. Inventory the connection. Record the accessible workspace, fields, identifiers, filters, and coverage before analysis begins.
2. Run one decision-focused query. Keep unrelated brands, topics, and time windows out of the first pass.
3. Request counterevidence. Ask Claude which returned records weaken or contradict its leading interpretation. A robust finding should survive that check.
4. Verify the underlying records. Open the relevant Profound view or record where possible. Check values, labels, dates, filters, and citations rather than approving an action from the prose alone.
5. Create an evidence ledger. For each recommendation, save the observation, scope, supporting identifiers, interpretation, action owner, and validation condition.
6. Repeat with equivalent scope. At the next comparable data refresh, use the saved control prompt and document any change in coverage before comparing results.
Add a final quality-control request before sharing the work:

Audit prompt: Audit your previous response. Create three lists: claims directly supported by returned Profound data, inferences that require validation, and recommendations based on editorial judgment. For each supported claim, include the relevant field, filter, date range, and record or citation identifier. Remove any claim you cannot trace.

This audit will not guarantee correctness, but it exposes a common failure mode: a valid observation, a plausible explanation, and a recommended action being compressed into one sentence as though all three had equal evidentiary weight.

Key takeaways
- The Profound connector gives Claude a route to authorized Profound context; it does not make every Profound field available by default.
- Begin by inventorying accessible accounts, records, fields, filters, identifiers, and date coverage.
- Frame each conversation around one decision, one defined scope, and an explicit standard of proof.
- Require Claude to separate measured observations from hypotheses and recommended actions.
- Verify important findings in the underlying Profound records and save an evidence ledger before assigning work.
- Compare periods only when their scope and coverage are equivalent, and never treat movement alone as proof of causation.
Start with one narrow, diagnostic conversation. Inventory the connection, investigate a single visibility gap, and verify every consequential claim before converting it into a content ticket. Once that path is reproducible, save the prompts and evidence fields as a team workflow. The value of the Profound Claude connector will come from disciplined questions and traceable decisions, not from the volume of analysis it can generate.

References
- Try Profound – The Profound connector is live in Claude
July 20, 2026

AI-Driven Personalized Search: A Practical SEO Playbook

You check an important query and see your brand. A colleague runs what looks like the same search and gets a competitor. A prospect asks an AI assistant and receives a third answer. That variation is no longer just measurement noise: AI search can adapt its response to the person and the moment, even when the words in the query stay the same.

Your optimization target has to change with it. You still need technically accessible pages, clear answers, and credible evidence. But you also need to make your brand useful across the different contexts that can shape a recommendation. That means mapping audience situations, connecting evidence across channels, and measuring recommendation coverage instead of chasing one supposedly universal rank.

Why one ranking report can mislead you

Search results were never identical for everyone. Location, language, device type, search history, and geographic intent have influenced conventional search for years. AI-powered search expands the potential context. Depending on the product, settings, and permissions, that context can include previous conversations, current activity, preferences, images, voice, documents, app usage, calendar events, or connected email.

Do not assume that every search product can access every signal. A signed-out search, a logged-in AI assistant, and a private enterprise chatbot may have very different context. The important point is that the query text is only one part of the input.

A useful working model separates personalized search into four layers:

The expressed task: What did the person explicitly ask, and what constraints did they include?
The person: What location, language, preferences, prior questions, or recurring needs may be relevant?
The moment: What are they doing now, which device or medium are they using, and how far have they progressed toward a decision?
The available evidence: Which pages, profiles, videos, reviews, discussions, and structured facts can the system retrieve and reconcile?

This does not make rankings irrelevant. It makes a single observation incomplete. A conventional rank tracker can still tell you whether a page is discoverable for a query in a defined configuration. It cannot, by itself, tell you whether an AI system will consider your brand suitable for a returning customer, a first-time buyer, a local searcher, or a user whose earlier questions established a specific constraint.

Keep your clean, repeatable search as a control. Then add deliberately defined context scenarios. The control helps you detect broad visibility changes; the scenarios reveal whether your content survives personalization.

Key takeaways for personalized AI search

The same prompt can produce different answers because the system may consider context beyond the query text.
Your practical unit of optimization is a decision in context, not an isolated keyword.
Your website should provide the clearest version of your facts, while relevant third-party and social evidence corroborates them.
Images, video, audio, transcripts, profiles, reviews, and structured information can all contribute to discoverability.
Measurement should separate brand visibility, citation, factual accuracy, and recommendation fit.
A test result is a sample from a defined setup, not proof of what every user will see.

Build a context map before you rewrite content

The tempting response to personalization is to create more pages for more personas. That usually produces shallow variations of the same answer. Start with a context map instead. It will show you where a different situation genuinely requires different advice, proof, or content.

Choose one decision where AI visibility matters. Write it as a complete sentence: a particular kind of person is choosing something for a stated use case under a meaningful constraint. If you cannot name the person, choice, use case, and constraint, the topic is still too broad to guide a useful page.

Define the base decision. Replace a loose topic such as reporting software with the actual decision, such as choosing a reporting platform for a distributed marketing team.
List explicit context. Capture details people are likely to state themselves: location, language, role, use case, required capability, existing workflow, or a restriction they cannot ignore.
List possible implicit context separately. Previous questions, current activity, device, preferred format, and search history may affect an answer even when they are not repeated in the prompt. Treat these as testing hypotheses, not facts you know about an individual.
Turn context into questions. Ask what would change the correct recommendation. A buyer and an implementer may need different evidence. A local service query may need location-specific facts. Someone comparing options may need tradeoffs that a first-time researcher does not yet know to request.
Assign evidence to every material claim. Decide whether the best support is a product page, demonstration, expert explanation, customer review, public profile, original analysis, or structured business fact.
Mark the content gap. Record whether the answer is absent, hard to find, unsupported, outdated, inconsistent across channels, or trapped in a format that is difficult to interpret.

A useful row in your context map contains the base query, audience situation, decision stage, decisive constraint, answer your brand can honestly support, evidence required, best publishing format, and current gap. That is enough detail to turn an abstract personalization strategy into an editorial brief.

Turn the map into page architecture

Build the main page around the stable part of the decision. Give the direct answer first, then explain who the answer applies to, what changes it, and what evidence supports it. Use distinct sections for meaningful context branches rather than hiding every variation in a generic paragraph.

State the decision clearly. The title and opening should identify the problem the page resolves, not merely the broad category it targets.
Define suitability. Say who the option is for, who may need something else, and which conditions change the recommendation.
Expose tradeoffs. A credible answer explains limitations and alternatives instead of treating every visitor as an ideal customer.
Place evidence beside the claim. Do not make the reader or a retrieval system hunt through an unrelated resources section to understand why a statement is credible.
Use descriptive headings. Headings should name the questions and constraints identified in the context map.
Give the next step. Match it to the decision stage: learn, verify, compare, inspect, configure, or contact.

Create a separate page only when the answer, evidence, or action changes materially. If two audience variants receive the same recommendation for the same reasons, one strong page with explicit subsections is more coherent than a collection of near-duplicate pages.

This is also where audience research and SEO meet. Search data can reveal recurring phrasing. Sales, support, community, and review language can reveal the conditions people omit from short queries but care about before acting. Convert those conditions into answerable sections, not a pile of persona labels.

Turn scattered channels into one corroborated brand record

An AI-generated response may synthesize information from a website, YouTube, LinkedIn, customer reviews, interviews, Reddit discussions, local business profiles, news coverage, and structured business information. At the same time, people use social and community platforms as search tools. Your brand is therefore encountered as an interconnected body of evidence rather than a set of isolated marketing channels.

You do not need to publish everywhere. You do need a deliberate role for every channel you use. Choose the places where your audience asks relevant questions and where the format can carry useful proof.

Start with an entity fact sheet that search, content, social, public relations, product, and support teams can share. It should contain:

The preferred organization and product names, including distinctions between similarly named offerings.
A concise, factual description of what the organization provides and for whom.
Official website, profile, support, and contact URLs.
Locations, service areas, or languages where those facts are genuinely relevant.
Named experts and authors, with accurate roles and biography pages.
The approved evidence behind important product, performance, compatibility, and expertise claims.
The owner and canonical location of each fact so outdated copies can be corrected.

Audit public assets against that sheet. Small differences in wording are natural. Contradictory names, obsolete descriptions, mismatched locations, and unsupported claims are not. When systems have to reconcile conflicting facts, you give them a reason to omit the brand or describe it incorrectly.

Give each channel a specific job. Your website should hold the canonical explanation and supporting detail. A video can demonstrate a process that is hard to understand in prose. LinkedIn can connect expertise to identifiable professionals. Reviews can provide independent evidence about customer experience. Local profiles can establish operational facts. Relevant community participation can answer real questions in the audience’s own language.

Do not try to manufacture consensus in forums or review platforms. Independent discussion is useful precisely because it is not another version of your landing page. Monitor recurring confusion, correct factual errors where participation is appropriate, and use the language of legitimate questions to improve the information you control.

Use JSON-LD to remove ambiguity, not manufacture authority

Structured data can make entities and relationships easier for machines to interpret. It cannot turn an unsupported assertion into a trusted fact. Treat JSON-LD as a consistency layer between visible content and your entity record.

Choose the Schema.org type that matches the actual entity or content, such as Organization, Person, Product, LocalBusiness, Article, or VideoObject.
Use stable names, canonical URLs, and identifiers across templates.
Connect an article to its real author and publisher rather than leaving those entities as unlinked text strings.
Use sameAs for authoritative profiles that represent the same entity, not for every page that happens to mention the brand.
Mark up facts that users can find on the page. Hidden or contradictory claims weaken the value of the implementation.
Validate generated markup and check it again when a template, plugin, author record, product record, or business fact changes.

Schema can clarify who published a claim, which product it describes, and how related entities connect. Authority still depends on the quality of the information and the wider evidence supporting it.

Make multimodal evidence understandable outside its original format

Personalized search is also multimodal. Systems can work with text, images, audio, video, voice, documents, and live context. That means a product photograph may become relevant to a visual search, while a video transcript may support an AI answer. Discoverability is no longer confined to conventional webpages.

Place useful captions and surrounding copy near images so the entity, action, and context are clear.
Write accessible alternative text that describes meaningful visual information rather than stuffing it with target phrases.
Publish accurate transcripts for useful video and audio, identify speakers, and link the media to the relevant organization, person, product, or topic page.
Explain important diagrams and demonstrations in nearby prose. Do not make a crucial qualification available only as text embedded in an image.
Keep product, expert, and organization names consistent in titles, descriptions, transcripts, captions, and profile metadata.
Edit transcripts into readable material when they are intended to answer a search need; a raw wall of speech is technically available but difficult for people to use.

The goal is not to duplicate every page in every medium. It is to choose the format that proves the point best, then provide enough textual and entity context for that asset to be understood and connected to your brand.

Measure recommendation coverage, not an imaginary universal rank

A personalized answer is not well represented by one position number. Your dashboard should separate four outcomes that are often collapsed into a single visibility metric.

Outcome	Question to record	What failure looks like
Visibility	Was the brand, expert, product, or content present?	A relevant answer omitted the entity entirely.
Citation	Was your asset linked, named, or used as supporting evidence?	The answer contained your information without connecting it to you, or relied on other evidence.
Accuracy	Were the description, relationships, qualifications, and current facts correct?	The answer repeated obsolete, conflicting, or incomplete information.
Recommendation fit	Was the brand suggested for a context it can genuinely serve?	The brand appeared but was not matched to the relevant audience need, or was recommended for an unsuitable case.

Build the test set from the context map, not from a generic list of high-volume keywords. Include prompts for broad discovery, evaluation, a decisive constraint, branded verification, and the questions people ask immediately before acting. If follow-up conversation is part of the interface, capture the whole sequence; prior turns can alter what the next question means.

Create a controlled baseline. Use a repeatable configuration and record the platform, exact prompt, account state, language, location, and device conditions that matter to the test.
Create contextual variants. Change one meaningful variable at a time, such as role, location, use case, or stated constraint. If several variables change together, you will not know which one affected the answer.
Keep supplied and inferred context distinct. Record what you explicitly told the system. Do not claim that an unseen personal signal caused a result unless the interface makes that connection clear.
Save the complete output. Capture the answer, follow-up prompts, citations or links, brands mentioned, recommendation language, and any factual errors. A screenshot without the test conditions is not a reusable record.
Score the four outcomes separately. A citation is not automatically a recommendation, and a mention is not automatically accurate. Preserve those distinctions in reporting.
Repeat the same configuration after meaningful changes. Compare patterns across the set rather than treating a single response as a stable ranking.

Do not assign a conventional rank when the output is not an ordered list. Record where the entity appeared and what role it played instead: direct recommendation, considered option, supporting authority, cited page, passing mention, or omitted entity. That description is more faithful to the experience and more useful to the team deciding what to fix.

The pattern of failures tells you where to investigate:

Absent across relevant scenarios: inspect technical accessibility, topic coverage, entity clarity, and external corroboration.
Visible only in branded prompts: inspect whether your content and evidence establish a clear association with the broader problem or category.
Cited but rarely recommended: inspect whether the material resolves suitability, constraints, and tradeoffs, rather than merely defining the topic.
Recommended but described inaccurately: find conflicting or outdated facts on your site, profiles, structured data, and prominent third-party pages.
Visible in one context but absent in another: inspect the missing context branch and the evidence required for that audience situation.
Different results across platforms: inspect which formats and evidence each answer used. Do not assume that one system’s result predicts another’s.

These patterns are diagnostic leads, not proof of causation. Confirm the gap in the underlying pages, profiles, markup, and cited evidence before changing content.

Begin with one decision journey where an incomplete AI answer could cost you a qualified opportunity. Build its context map, reconcile the entity fact sheet, publish the missing evidence in the format that best carries it, and capture a controlled baseline. Let the observed gap determine the next change. Personalized search is too variable for a vanity ranking, but it is structured enough for a disciplined visibility strategy.

References

Search Engine Land — A practical guide to AI-driven personalized search

July 20, 2026

A Decision Guide to Eight Insurance GEO Agencies in 2026

Insurance companies evaluating generative engine optimization agencies face a specialized buying decision: a partner may understand AI search without understanding insurance, or know insurance marketing while offering little evidence of a mature GEO practice.

A comparison published by First Page Sage Blog highlights eight agencies with different combinations of AI visibility, sector knowledge, content capabilities, and channel coverage. Because First Page Sage evaluated the market and ranked itself first, buyers should treat the results as a vendor-produced shortlist rather than an independent industry benchmark.

How the reported comparison was constructed

First Page Sage Blog says its team assessed 38 agencies and selected eight. AI visibility carried 25% of the evaluation, while the depth of each GEO offering and aggregated client reviews each represented 20%. Leadership experience accounted for 15%, with media references and notable insurance clients contributing 10% apiece.

This framework rewards more than conventional search performance. It considers whether an agency can help a brand appear in answers from platforms such as ChatGPT, Perplexity, Claude, and Google Gemini, while also examining evidence such as GEO research, case studies, reviews, leadership credentials, media citations, and client portfolios. The source does not describe independent auditing of the scores, so the numbers are most useful as comparison points to investigate further.

The eight-agency scorecard at a glance

The following table preserves the source’s ranking and its four scored dimensions. A higher position reflects the complete weighted framework, not AI visibility alone.

Rank	Agency	AI visibility	GEO	Reviews	Leadership
1	First Page Sage	4.9	5.0	4.9	4.9
2	Genevate	4.6	4.8	4.8	4.3
3	Focus Digital	4.3	4.5	4.8	4.2
4	Amsive	4.3	4.4	4.7	4.4
5	BrightFire	4.2	4.2	4.8	4.4
6	EWR Digital	4.4	4.4	4.6	4.2
7	Neilson Marketing	4.1	4.0	4.7	4.3
8	Digital Logic	4.2	4.3	4.6	4.3

Match the agency model to the insurance buyer

For a GEO-led content program, the source places First Page Sage at the front of the field. It describes an in-house insurance content operation covering regulatory reports, interviews, compliance topics, and commercial landing pages. The publisher also reports that its insurance clients average $1.7 million in new net revenue annually, alongside a 1.7% landing-page conversion rate and 63% average engagement rate. Those are vendor-reported campaign claims and should be validated against comparable client references, attribution rules, and contract scope.

Genevate and Focus Digital represent two alternatives for organizations prioritizing GEO expertise over deep insurance specialization. The source characterizes Genevate as combining AI-focused optimization with public relations and reputation work, while Focus Digital emphasizes thought-leadership content for smaller and mid-market companies. It also cautions that both portfolios contain less insurance experience than those of sector-focused competitors. EWR Digital occupies related territory, combining B2B SEO, digital PR, and AI search visibility, but with a portfolio reportedly weighted toward other professional-services sectors.

Amsive is positioned for larger insurers that need data, paid media, email, direct mail, organic search, and programmatic execution under one relationship. First Page Sage Blog identifies USAA and Allstate as notable clients, but says GEO is one component of a broader performance-marketing operation rather than the agency’s defining specialty.

BrightFire, Neilson Marketing, and Digital Logic are more closely aligned with traditional insurance marketing needs. The source describes BrightFire and Neilson as insurance-focused specialists, with Neilson bringing more than 30 years of sector experience. Digital Logic is presented as a practical option for independent agencies and regional brokerages. In each case, however, the report finds less public evidence of a developed GEO methodology than it attributes to the higher-ranked GEO specialists.

Key takeaways

No single score captures both AI-search capability and insurance fluency.
First Page Sage leads its own published ranking, making independent validation especially important.
Genevate, Focus Digital, and EWR Digital emphasize GEO or AI visibility but reportedly have less insurance depth.
Amsive suits complex multichannel programs, while BrightFire, Neilson Marketing, and Digital Logic lean toward established insurance marketing services.

What to verify before selecting a partner

A useful procurement process should test the claims behind the scorecard. Buyers can ask each finalist to show insurance-specific work, explain how AI visibility is measured, distinguish citations from referral traffic, and identify which activities are handled in-house. Case studies should clarify baselines, time periods, attribution methods, and whether reported outcomes came from GEO, traditional SEO, paid media, or several channels working together.

Fit also depends on operating needs. A carrier coordinating multiple channels may value Amsive’s breadth, while an independent agency may prefer a managed insurance-marketing provider. An insurtech seeking stronger brand representation in AI answers may place more weight on GEO and digital PR. The most defensible choice will be the agency that can connect its proposed work to the buyer’s audience, compliance review process, distribution model, and measurable business objective.

As AI discovery develops, documented methodology and transparent measurement should matter more than labels alone. A short paid pilot with agreed reporting standards can reveal whether an agency’s claimed specialization translates into useful visibility and qualified demand.

Inspired by this post on First Page Sage Blog.

July 17, 2026

A Buyer’s Guide to eCommerce ASO Agencies for 2026

Choosing an agentic search optimization agency requires more than comparing who mentions AI most often. eCommerce teams need to decide whether they want a specialist in AI discovery, an analytics-led partner, or a broader marketing agency that can add ASO to an existing program.

First Page Sage Blog evaluated seven agencies for its 2026 shortlist. The comparison below reorganizes its findings around buyer fit while keeping the source’s scores and claims clearly attributed.

ASO extends product discovery into purchasing

Agentic search optimization, or ASO, prepares a brand to be found, assessed, and potentially acted on by AI agents. For an online retailer, that can involve clear product information, credible comparison content, consistent brand signals, and technical systems that machines can interpret.

This makes ASO broader than simply appearing in a generated answer. An agency may also need to address how an agent evaluates alternatives and whether product or checkout infrastructure can support a transaction. The right scope therefore depends on whether a retailer needs visibility alone or an end-to-end agentic commerce program.

How to interpret the reported ranking

According to First Page Sage Blog, its weighted model assigned 25% to ASO expertise; 20% each to AI visibility, leadership experience, and average reviews; 10% to notable eCommerce clients; and 5% to estimated media references. The visibility assessment covered platforms such as ChatGPT, Perplexity, Claude, and Google Gemini.

Capability signals: ASO expertise, AI visibility, and relevant leadership experience.
Market signals: review ratings, client portfolios, and estimated media citations.
Important limitation: the publisher evaluated and ranked itself first, so buyers should treat the table as a sourced shortlist rather than an independent verdict.

The reported scores can help narrow the field, but they do not reveal pricing, staffing, contract terms, implementation capacity, or results for a particular catalog. Those points still require direct verification.

The seven-agency shortlist at a glance

The following table preserves the source’s order and two principal scores while translating each profile into the type of engagement it appears designed to support.

Rank	Agency	ASO expertise	AI visibility	Positioning reported by the source
1	First Page Sage	5.0	4.9	Full-stack ASO, GEO, SEO, and thought leadership
2	Genevate	4.8	4.6	Specialist work across GEO, ASO, and emerging AI platforms
3	Focus Digital	4.5	4.5	Conversion-focused programs for small and mid-market retailers
4	Driven Metrics	4.4	4.4	Attribution modeling and agent-conversion diagnostics
5	Tinuiti	4.3	4.2	Full-funnel performance marketing with an AI SEO offering
6	SmartSites	3.9	4.0	Traditional eCommerce marketing with developing ASO services
7	Aumcore	3.7	3.7	Voice and AI search optimization with emerging ASO capabilities

First Page Sage describes its own program as spanning AI representation audits, comparison content, and machine-actionable checkout readiness. It also reports that research led by its president, Evan Bailyn, analyzed 2,417 agentic search commands and organized ASO into retrieval, evaluation, and action stages. Because these claims come from the agency itself, prospective clients should request supporting methodology and relevant case evidence.

The other profiles suggest several distinct choices. Genevate is presented as an AI-search specialist, although the source flags its smaller scale. Focus Digital may suit cost-conscious small or mid-market brands seeking conversion support, while Driven Metrics emphasizes measurement and diagnostics. Tinuiti and SmartSites offer broader marketing coverage, but the source characterizes their ASO practices as less specialized. Aumcore may be relevant when voice and conversational search are also priorities.

Questions to resolve before selecting a partner

What will the agency optimize? Confirm whether the scope covers discovery, product evaluation, structured product information, and transaction readiness.
How will progress be measured? Ask for platform-level visibility reporting and a defensible connection between agent activity and commercial outcomes.
Can the team handle the catalog’s complexity? Multi-SKU, multi-market, or enterprise programs may require different staffing and technical capacity than a smaller direct-to-consumer store.
Which claims can be demonstrated? Request relevant case studies, references, sample deliverables, and an explanation of how reported improvements were attributed.

Key takeaways

First Page Sage Blog ranked First Page Sage, Genevate, and Focus Digital in the first three positions.
The list spans dedicated AI-search specialists, conversion and analytics firms, and full-service performance agencies.
Published scores are useful for screening, but the source’s self-ranking and estimated inputs make independent due diligence essential.
The strongest choice is the agency whose scope, measurement approach, and delivery capacity match the retailer’s actual operating needs.

As agent-assisted shopping develops, retailers will benefit from treating ASO as an operational capability rather than a one-time visibility campaign. A tightly defined pilot can expose whether an agency can connect content, product data, measurement, and commerce infrastructure before the relationship expands.

Inspired by this post on First Page Sage Blog.

July 17, 2026

A Practical Framework for Auditing Local AI Visibility
A strong Google Maps presence does not reveal whether an AI assistant will recommend a local business, describe it accurately, or favor a competitor. A local generative engine optimization (GEO) audit measures those outcomes directly.

The goal is to establish a controlled baseline before changing content, citations, reviews, or technical settings. That baseline turns an uncertain visibility problem into a set of errors and opportunities that can be tracked.

Why local AI visibility needs its own benchmark

Traditional local rankings and AI recommendations are related, but they are not interchangeable. Search Engine Land cites SOCi’s 2026 Local Visibility Index, which analyzed nearly 350,000 business locations. ChatGPT reportedly recommended 1.2% of those locations, compared with a 35.9% appearance rate in Google’s local three-pack. The reported recommendation rates were 11% for Gemini and 7.4% for Perplexity.

The source also reports that business information was about 68% accurate on ChatGPT and Perplexity, while Gemini reached 100% accuracy in that analysis and relied entirely on Google Maps data. These findings illustrate why map rankings alone cannot serve as an AI visibility scorecard: different systems can select different businesses, consult different sources, and reproduce business facts with different levels of accuracy.

Key takeaways
- Test discovery, comparison, trust, and logistics questions across the AI platforms customers may use.
- Record whether the business appears, where it appears, how it is framed, whether its details are correct, and which sources support the answer.
- Separate visibility failures from factual errors and weak competitive positioning.
- Resolve crawl access and business-data inconsistencies before investing heavily in new local content.
- Repeat the same test set over time so changes can be compared against a stable baseline.
Build a test that produces comparable evidence

Begin with a spreadsheet and a fixed set of prompts. The prompt set should represent four kinds of customer questions: discovery queries such as the best service in a city, comparisons between the brand and a competitor, trust questions about reviews or reliability, and logistics questions covering hours, address, parking, or phone number.

Run the same questions in the relevant interfaces, which may include ChatGPT, Perplexity, Gemini, and Google AI Overviews. For every response, log the prompt, platform, date, test location, and session state. Search Engine Land recommends comparing logged-in and clean logged-out sessions to help identify personalization noise. The city or ZIP code must also remain explicit because local context can change the answer.

Each result should capture five observations: whether the brand was mentioned, its order in the answer, the positive, neutral, or negative framing, the accuracy of operational facts, and the cited sources. Competitors should be recorded in the same rows, including their position and supporting sources. This makes the audit useful for both brand diagnosis and competitive analysis.

Translate results into three types of failure

An aggregate visibility percentage shows how often the business appears, while an accuracy percentage shows how often its details are correct. Those summary figures are useful, but the underlying problem determines the appropriate response.

Organic visibility, paid media and artificial intelligence merge into one connected search ecosystem, where vivid data streams link a creative website with a powerful analytics dashboard.
- Invisible: The business is absent from relevant answers. Possible causes identified by the source include crawler restrictions, insufficient citable material, or limited third-party mentions.
- Inaccurate: The business appears with an obsolete address, incorrect hours, or outdated services. On-site errors and inconsistent name, address, and phone data across directories should be investigated.
- Misframed: The business is mentioned but placed below competitors or presented as a weaker choice. A limited review profile or weaker authority signals may be contributing factors.
This classification prevents a common planning mistake. Publishing another city page will not correct blocked access, and adding schema will not by itself overcome weak third-party validation. The audit should connect each observed symptom to the most plausible layer of the problem.

Prioritize access, trust, and then relevance

Remediation should follow the dependency chain. First, confirm that relevant crawlers can reach the site by reviewing robots.txt and applicable security or Cloudflare controls. Search Engine Land notes Cloudflare’s announcement that AI crawlers would be blocked by default on sites using its network, making the site’s actual configuration worth checking rather than assuming access.

Next, align the business name, address, and phone number across the website and external profiles. Validate appropriate structured data, including LocalBusiness, Organization, FAQ, and Service markup where the page content supports it. Then strengthen trust through accurate profiles, reviews, responses to customer questions, and a consistent description of the business across directories, social accounts, and coverage.

Content becomes the priority after those foundations are sound. Useful local pages should contain genuine city-specific information, concrete service examples, and practical details rather than repeating a template with a different place name.

Turn the baseline into an operating metric

Search Engine Land suggests a quarterly audit for most local businesses. Reuse the same core prompts and controls, then compare mention rate, position, factual error rate, citation count, and competitor share of voice with the previous run. Changes in cited sources or answer wording may indicate model drift and should be documented rather than treated as isolated anomalies.

Clicks are not the only relevant outcome because an AI answer may influence a decision without producing a website visit. Branded search activity, calls, and direction requests can provide additional business context. The next audit should then test whether the chosen fixes improved the specific weakness originally observed.

Inspired by this post on Search Engine Land.
July 16, 2026

Google AI Mode Visibility Is Splitting Into Three Channels

Commercial visibility in Google AI Mode is developing along several paths at once. Advertisers can buy placements, publishers and brands can earn citations, and businesses can appear through Google-hosted profiles or product panels.

Two separate reports show why these surfaces should not be treated as one ranking system. Paid coverage is expanding across commercially valuable queries, while Google is also becoming a more prominent source inside its own AI-generated answers. The practical payoff is a clearer way to assign budgets, ownership and measurement.

Key takeaways

CrushPress.AI’s summary of an SE Ranking study reported text ads on 29.45% of the commercial AI Mode queries examined.
Ad incidence rose with keyword cost, but the study did not find the same relationship with search volume or keyword difficulty.
Paid placement provided little overlap with cited or traditionally ranked URLs, indicating that advertising, citations and organic search require separate strategies.
A second CrushPress.AI report said google.com became AI Mode’s second-most-cited domain in Profound’s tracking, driven mainly by Google Business Profiles and Product Knowledge Panels.
Commercial visibility therefore depends on both website performance and the quality of information presented on Google-controlled surfaces.

Commercial visibility now has three distinct layers

The reports describe complementary changes rather than competing explanations. The SE Ranking analysis covered paid text placements, whereas Profound tracked the domains AI Mode cited. Together, the findings suggest that appearing near an AI-generated response can happen through three substantially different mechanisms: an ad auction, selection as a cited source, or a Google-hosted information surface.

The distinction matters because each layer answers a different business need. Advertising can provide purchased exposure on a relevant query. A citation can establish a website as supporting material for the generated answer. A Google Business Profile or Product Knowledge Panel can present decision-making information without requiring the user to reach the company’s website first.

Profound’s tracking, as summarized by CrushPress.AI, found that citations to google.com increased 8.4-fold in roughly two months, making it the second-most-cited domain in the data. The report attributed almost all of that increase to Google Business Profiles and Product Knowledge Panels. Its analysis ran from April 15 through June 30 and covered more than 32 million google.com/searchviewer instances.

The reported shift was especially relevant to local searches in hospitality and travel, home services, restaurants and dining, real estate, and healthcare. For product-oriented queries, panels appeared more often around comparisons, compatibility and specifications. These are situations in which structured facts can influence consideration before a conventional website visit.

Paid reach follows commercial value, not general popularity

CrushPress.AI’s account of the SE Ranking study reported ads across 14,733 queries, or 29.45% of the commercial searches analyzed. The study examined 50,032 U.S. keywords across 20 niches, using results collected on June 30. It focused on queries eligible for text ads and excluded product carousels.

Cost per click was the clearest reported indicator of whether an ad appeared. Ad incidence was 24.33% among keywords with CPCs below $2, 32.45% in the $2-to-$10 group and 53.56% for keywords at $10 or more. Search volume and keyword difficulty did not show the same relationship in the study. This pattern supports a cautious interpretation: AI Mode ad deployment appears more closely aligned with the economic value of a query than with its popularity or organic competitiveness alone.

When an ad block appeared, it usually included more than one advertiser. The study found two ads in 71.1% of ad-triggering responses and one ad in the remaining 28.9%. Category results varied sharply, from a reported 72.38% ad rate for pets to 2.64% for healthcare. Those differences warn against using the overall 29.45% rate as a forecast for every market.

The study also noted that AI Mode results can vary between sessions. Its percentages should therefore be read as observations from the stated collection date and methodology, not permanent delivery guarantees. The source further cautioned that the pattern could change as Google introduces more AI-specific advertising formats.

Buying an ad does not secure the other layers

The most consequential finding for planning was the limited overlap between advertisers and unpaid visibility. According to the SE Ranking analysis summarized by CrushPress.AI, only 11.53% of advertiser domains appeared among cited sources for the keywords on which they advertised. At the individual URL level, overlap fell to 1.95%.

Traditional organic results showed a similar separation. Just 2.32% of advertised URLs also ranked organically for the corresponding queries, while domain-level overlap reached 15.35%. In other words, approximately 85% of advertisers did not appear in organic results for the same keywords, according to the source.

Visibility comparison	Reported overlap	Planning implication
Advertiser domain and cited domain	11.53%	Paid reach is not a substitute for earning citations.
Advertised URL and cited URL	1.95%	The landing page is rarely the exact source selected for the answer.
Advertiser domain and organic domain	15.35%	Advertising and domain-level organic visibility remain largely separate.
Advertised URL and organic URL	2.32%	Buying exposure does not ensure that the same page ranks.

The researchers reportedly compared advertisers with similar non-advertising domains while accounting for domain strength, backlinks, referring domains and organic visibility. Even so, the findings are observational. They do not establish that advertising causes or prevents citation and ranking outcomes. They do show that purchasing an AI Mode placement should not be assumed to improve either one.

A practical operating model for AI Mode visibility

Organizations can respond by assigning each visibility layer a distinct job. Paid-search teams can evaluate AI Mode ads according to query economics, placement availability and conversion performance. SEO and content teams can monitor whether the brand’s pages are cited or ranked, then improve the relevance and usefulness of the pages intended to earn that exposure.

Local and commerce teams need a third workstream for Google-hosted information. Business hours, locations, photos and reviews can become part of the AI Mode experience through Google Business Profiles. Product specifications and compatibility information may surface through Product Knowledge Panels. Because those details can be encountered before the website, maintaining them is part of commercial presentation rather than a secondary listing task.

Reporting should preserve the same separation. A combined visibility score can conceal whether progress came from spending more, earning stronger source selection, improving organic rankings or maintaining a more complete Google-hosted profile. Channel-specific reporting makes it possible to connect each outcome to the team and investment responsible for it.

The next useful evidence will be longitudinal: whether ad incidence continues to rise, whether new formats change advertiser competition, and whether Google’s share of citations remains concentrated in its own local and product surfaces. Until those patterns are clearer, the sound approach is to manage AI Mode as a portfolio of paid, earned and platform-hosted visibility rather than as a single search position.

References

July 15, 2026

Tag: AI Visibility

Capture the claim before anyone starts correcting it

Triage inaccurate claims by harm, evidence, and correctability

Choose the evidence layer that can actually be corrected

Run the correction as a controlled handoff

Verify the result without overreading one clean answer

Key takeaways

References

Key takeaways

First, determine whether you have a recognition or category problem

Define one primary category and a small set of adjacent frames

Build an entity map that gives every page a job

Turn the map into a content system, not one oversized page

Use JSON-LD to declare truth, not manufacture relevance

Create off-site category evidence, then measure the whole system

Build corroboration in the category you want to earn

Measure recognition, eligibility, recommendation and evidence separately

References

Use Brainstorm Mode to close a decision gap

Give Aim a broad goal without giving it an empty one

Review every proposed Agent as a research brief

Break apart bundled Agents

Separate observations from explanations

Turn the first Agent into a controlled decision loop

Key takeaways

References

Key takeaways for choosing a GEO agency

What a real GEO agency should actually own

Use the rebranded-SEO test

Demand an AI-visibility measurement system you can audit

Score every agency proposal against the same evidence

Ask questions that force the method into the open

Treat guarantees as a control problem, not a bonus

Structure the first engagement so you can inspect the work

Protect the assets and access your team will need later

References

What the Profound connector changes – and what it does not

Scope the decision before you scope the data

Three workflows that produce defensible AI visibility actions

1. Find a visibility gap without inventing its cause

2. Turn prompt and citation signals into a content brief

3. Compare periods without turning movement into causality

Build an evidence trail from conversation to action

Key takeaways

References

Why one ranking report can mislead you

Key takeaways for personalized AI search

Build a context map before you rewrite content

Turn the map into page architecture

Turn scattered channels into one corroborated brand record

Use JSON-LD to remove ambiguity, not manufacture authority

Make multimodal evidence understandable outside its original format

Measure recommendation coverage, not an imaginary universal rank

References

How the reported comparison was constructed

The eight-agency scorecard at a glance

Match the agency model to the insurance buyer

Key takeaways

What to verify before selecting a partner

ASO extends product discovery into purchasing

How to interpret the reported ranking

The seven-agency shortlist at a glance

Questions to resolve before selecting a partner

Key takeaways

Why local AI visibility needs its own benchmark

Key takeaways

Build a test that produces comparable evidence

Translate results into three types of failure

Prioritize access, trust, and then relevance

Turn the baseline into an operating metric

Key takeaways

Commercial visibility now has three distinct layers

Paid reach follows commercial value, not general popularity

Buying an ad does not secure the other layers

A practical operating model for AI Mode visibility

References