This past Black Friday and Cyber Monday, I delved into the fascinating insights from our Black Friday Index, crafted from a vast pool of 400 million genuine conversations. It was enlightening to see which brands stood out as AI’s top recommendations, especially as so many of us relied on Answer Engines to hunt down the best deals.
As I explored the data, the impact of AI on shopping trends became crystal clear. The technology not only streamlined how we search for deals but also influenced brand visibility and consumer choices. The excitement of seeing how AI is reshaping shopping habits made this year’s Black Friday and Cyber Monday particularly intriguing for me.
The findings from the Black Friday Index are a testament to the growing importance of AI in retail, showing us how indispensable it has become for both consumers and brands. Being part of this evolution makes me look forward to what future shopping events will bring, especially as technology continues to advance.
You want AI systems to recognize and cite your expertise, but you don’t want a generated answer to replace the page, dataset, or original work that paid for it. A blanket allow-or-block decision cannot resolve that conflict.
The workable approach is to decide separately what should be discoverable, available for live answers, eligible for model training, or kept behind real access controls. Connect those decisions to business value and rights status before anyone edits a crawler directive.
Stop treating crawl access as one permission
Traditional search indexing, result previews, live retrieval for an AI answer, and model training are different uses. A platform may offer separate controls for some of them, combine others, or provide no control that matches the choice you actually want to make.
The European Commission’s antitrust investigation puts this lack of choice at the center of the dispute: publishers argue that they cannot meaningfully reject generative use without jeopardizing search visibility. The investigation does not settle what is lawful for your content, but it does expose the strategic mistake of treating search inclusion as consent to every downstream use.
For every important group of URLs, answer four separate questions:
Should an ordinary search crawler be allowed to index this content?
Should a search result be allowed to display a preview or snippet?
Do you want an AI system to retrieve this page when constructing a live answer?
Do you want the content used to train or improve a model?
Do not assume that one directive answers all four questions. Write down the desired outcome first, and then identify whether each platform provides a documented control for it.
A robots.txt rule is also not a security boundary. It communicates a preference to crawlers that honor it; it does not make public material confidential or prevent every form of copying. If disclosure of a dataset, licensed report, client deliverable, or proprietary method would cause serious commercial or legal harm, protect it with authentication or another genuine access control. If ownership or licensing terms are unclear, have intellectual-property counsel review them before changing access or reuse terms.
Build a rights-to-visibility matrix before changing directives
Make decisions at the URL-family level rather than applying one sitewide rule. A public glossary, a product page, an original investigation, and a licensed database do not carry the same discovery value or substitution risk.
Decision factor
What to record
How it should affect your posture
Business role
Discovery, authority building, conversion, support, or paid deliverable
Discovery content usually benefits from broader access; a paid deliverable needs a stronger boundary
Rights status
Owned, licensed, contributor-supplied, user-supplied, or uncertain
Uncertain or restricted rights require review before you authorize new uses
Substitution risk
Whether a generated answer could satisfy the need without a visit
High-risk pages may need a useful public summary with the full asset kept under access control
Visibility dependency
Search impressions, qualified visits, leads, sales, or assisted conversions
Do not restrict a high-dependency URL group without a baseline and rollback plan
Distinctive value
Original data, reporting, methodology, tools, templates, or expert analysis
The harder the asset is to replace, the more deliberate its public surface should be
Available controls
Crawler, directive, affected product, documented behavior, and owner
Implement only controls that match the intended use closely enough to justify the tradeoff
Turn that matrix into an implementable policy:
Group URLs by template and business function. Start with categories such as public reference content, commercial pages, original editorial work, licensed material, and authenticated assets.
Assign a default posture to each group: open for discovery, public but bounded, restricted, or licensed for specific uses.
Record which team owns the decision. SEO can explain visibility consequences, but it should not silently decide rights questions for editorial, product, or legal teams.
Inventory the current robots.txt rules, page-level directives, authentication boundaries, and contractual restrictions before changing anything.
For each crawler instruction, record the exact crawler and product behavior it is meant to affect. Do not infer behavior from the directive’s name.
Apply the first change to a non-critical URL family. Preserve the previous configuration, capture the baseline, and define the condition that would trigger a rollback.
The same caution applies to noai, nopreview, and similar emerging conventions. A label does not tell you which systems honor it, whether it affects training or live retrieval, or whether it changes ordinary search eligibility. Platform-specific documentation has to answer those questions.
Make the public layer easy to cite and hard to confuse
Protecting high-value material does not require making your whole brand invisible. A stronger architecture separates a public reference layer from the asset that contains the complete commercial value.
Build a useful public reference layer
The public page must contain enough substance to deserve selection. A vague teaser gives an answer engine little reason to cite you, while publishing the entire asset may let the generated response replace you.
Use descriptive headings and answer one recognizable question directly under the relevant heading. Follow the short answer with scope, exceptions, evidence, and the next action.
Name your organization, authors, products, and subject entities consistently. Make authorship, expertise, editorial responsibility, and update history visible rather than leaving authority to be inferred.
Add structured data that agrees with the visible content. Appropriate schema, complete metadata, and meaningful image alt text can help machines connect the page to the correct entities, but markup does not grant a license or compel an AI system to cite you.
Show provenance for consequential claims. Identify who produced original data, explain the method at a useful level, state important limitations, and distinguish an observed fact from your interpretation.
Give the reader a reason to continue beyond the extracted answer: an interactive tool, complete dataset, implementation workflow, downloadable resource, consultation path, or transaction that the summary cannot reproduce.
Generic explanations are especially vulnerable to substitution because the answer contains little that belongs distinctly to your entity. The public layer should carry something attributable: a clear framework, original evidence, a named expert’s analysis, a transparent method, or a maintained record of change.
Keep the irreplaceable asset behind a real boundary
Keep full proprietary datasets, premium templates, licensed archives, and account-specific outputs behind authentication when public exposure is not an acceptable cost of discovery.
Publish a useful summary only if you are comfortable with that summary being publicly accessible and potentially reused.
State ownership and permitted uses in clear terms, and provide a licensing or permissions contact for organizations that want broader access.
Do not publish confidential material and rely on a bot instruction to protect it. Remove it from public delivery or require authorized access.
This creates a deliberate exchange: machines can understand what you know and why your entity is relevant, while the complete experience or asset still requires a relationship with you.
Measure whether visibility creates value or merely extraction
Organic sessions alone no longer describe search performance. Many AI interactions end without a click, so referral traffic cannot capture every useful mention or every instance in which your material satisfies the user elsewhere.
Capture a baseline before changing access controls, then monitor five layers:
Answer visibility: Use a fixed set of important prompts and record whether your brand, product, expert, or content appears. Keep the prompt wording stable enough to compare observations.
Attribution quality: Record whether the answer names you, links to the correct page, represents the claim accurately, and distinguishes you from similarly named entities.
Discovery: Track ordinary search impressions, clicks, AI referrals that can be identified, landing pages, and changes by URL family.
Business value: Measure qualified conversions, assisted conversions, sales conversations, subscriptions, branded search, and other downstream outcomes that matter to the page’s assigned role.
Exposure: Review server logs for crawler activity and document cases where protected or distinctive material appears elsewhere without the attribution or use you expected.
Interpret combinations of signals instead of chasing a single metric:
If AI mentions rise and qualified conversions also rise, the public layer is probably supporting discovery even when direct clicks are limited.
If mentions rise but links and downstream value do not, inspect whether the answer reproduces too much of the page, the citation is missing, or the page lacks a compelling next step. Blocking should not be your automatic first response.
If visibility falls after a directive change, compare crawler logs, indexing, and the affected URL family against the recorded intent. Roll back when the lost discovery is more valuable than the use you prevented.
If an AI answer misstates your position, improve the page’s explicit definitions, entity relationships, evidence, and limitations. Preserve examples of the error so you can determine whether the problem changed.
If licensed, confidential, or access-controlled material is reproduced, preserve the output, URL, date, relevant access logs, and configuration. Escalate to the platform and qualified counsel rather than trying to settle the rights question through SEO settings alone.
Keep a change log with the affected URL family, intended behavior, implementation owner, prior configuration, observed result, and rollback condition. Without that record, a later traffic change will tempt the team to assign causation to whichever AI event is most visible.
Key takeaways
Search indexing, snippets, live AI retrieval, and model training are separate uses, even when a platform does not provide separate controls for all of them.
Google-Extended can address Gemini training without necessarily removing indexed content from AI Overviews or preventing live use in generated answers.
Make rights decisions by URL family and business role, not with one sitewide allow-or-block rule.
Schema and clear HTML improve machine understanding; they do not create access control, waive rights, or guarantee attribution.
Use authentication for assets that must remain protected. Crawler preferences are not a substitute for a security boundary.
Judge AI visibility by attribution, accuracy, qualified outcomes, and exposure as well as traffic.
Your next move is to choose one important URL family and complete the rights-to-visibility matrix before touching its directives. Capture the current configuration and performance, decide which uses you actually want, and change only the control that can credibly serve that decision. The durable strategy is neither maximum exposure nor total disappearance. It is a deliberately designed public surface with a defensible boundary around the value you cannot afford to give away.
When I think about improving my website’s visibility, AI comes to mind as a crucial tool. It serves as a second pair of eyes, helping me evaluate intent signals, compare top results, and refocus pages that aren’t performing well.
Despite having well-written content, excellent layout, and robust backlinks, pages can still underperform in rankings. A frequent culprit is misaligned search intent, which can be more elusive than it seems.
Focusing on content optimization and usability sometimes makes it easy to overlook or misjudge intent. This is where AI shines as a reviewing tool, effectively steering things back on course.
Whether I’m working on a new page or revising an existing one, returning to the basics of search intent always sets me up for success.
Starting with a simple AI prompt to outline likely search intents for a keyword offers a solid framework for content creation or optimization.
This comprehensive list isn’t something I strive to cover completely on a single page. Instead, it highlights diverse user types, shifts in intent, and needs I might not have initially considered.
By considering these factors, I aim to create a more useful, well-rounded page that genuinely satisfies user needs.
Getting the intent right can be challenging. AI tools help me understand what’s already successful by examining top-ranking pages and what they excel at.
I utilize AI tools for a swift overview of a page’s primary intent. By evaluating this at scale, I can see if top-ranking pages meet the same intent.
It’s crucial to assess the intent of my page with the same rigor, be it a fresh draft or a page I’m optimizing. If the primary intent aligns with what’s succeeding, it’s a strong starting point. If not, it provides clear direction for improvement.
Again, consulting AI tools for improvement suggestions can yield valuable insights into refining intent. Key areas to focus on include:
The language I use can either reinforce or contradict the intended message. For commercial intent, persuasive wording is necessary, while for informational pages, clear and descriptive language is preferred.
The format of a page can also convey intent. For instance, in a sales page, details like product placement and accompanying information matter greatly. Similarly, guides need clear step-by-step labeling and possibly visual aids.
Clearly defined calls to action are essential. They align the user’s actions with the page’s intent, enhancing both engagement and ranking potential. Unclear or generalized calls to action dilute this effect.
Listing accurate pricing, VAT elements, and currency signals is vital in conveying commercial intent. They guide users accurately at critical decision points.
Availability of support is another crucial factor. I make sure that pre- or post-sale queries can be easily addressed by ensuring my contact details and support options are clearly visible.
Trust signals, like product guarantees, return policies, and customer reviews, make a big difference in user decisions. Including these details serves to strengthen user trust.
When clear comparisons are needed, laying out products side by side can assist users in their decision-making process, moving them closer to making a purchase.
In my experience with working pages centered around user intent, I’ve seen that excess information can sometimes bloat a page.
Previously, this depth might have worked, but now clarity and a focus on intent are what truly resonate.
I’ve learned to reassess where content performs best within the user journey, often seeking AI’s guidance to refocus content structure wisely.
For instance, if I notice my sales page for internal French doors isn’t performing, I consult AI, along with competitor analysis, to uncover key insights.
Competitors might be focusing on selling first, while my page addresses user concerns, which means I need to reposition my content priorities.
By reordering sales-driven content and addressing pain points concisely, I better align with user intent, letting supporting pages deal with detailed post-sale information.
AI isn’t here to replace expertise but to guide my strategic intent, enhancing my understanding of user behavior for better conversion.
You don’t need a separate content factory for this. Keep the foundations of SEO, but change the unit you optimize: move from isolated keywords to complete decision journeys. Then measure demand, retrieval, answer visibility, and business outcomes instead of treating clicks as the only proof that your work mattered.
Optimize for the decision behind the prompt
The meaningful change in AI-driven search isn’t simply that queries are getting longer. A detailed prompt can contain several jobs at once: define a problem, compare options, apply constraints, check current conditions, and recommend a next step. Google’s rollout of Gemini 3 Flash as the default model for AI Mode is designed around reasoning across those facets while incorporating web, real-time, and local information.
Think of the behavior as query decomposition. A request such as “Which platform should our international retailer use to manage SEO during a site migration?” may require answers about ecommerce features, regional requirements, migration workflows, integrations, cost considerations, and implementation risks. Ranking for the broad phrase “SEO platform” addresses only a fraction of the job.
Before revising a page, write down the complete decision it needs to support:
The core problem the user is trying to solve.
The constraints that could change the answer, such as location, business type, technical environment, or deadline.
The alternatives the user is likely to compare.
The criteria needed to make that comparison fairly.
The sequence of actions required after the decision.
The facts that must be current rather than generally true.
The follow-up question a careful user would ask before acting.
This exercise gives you a decision map rather than a bag of keyword variations. It also tells you how to structure the site. Keep closely connected facets on one page when they serve the same reader and require the same evidence. Create supporting pages when a facet has its own intent, evidence, or implementation path. Link those pages so a crawler, search engine, and person can follow the relationship without guessing.
The strategic foundation remains familiar because Google’s position is that SEO for AI is still SEO. AI visibility doesn’t excuse weak crawlability, vague writing, unsupported claims, or poor site architecture. It raises the cost of those weaknesses because an answer system can select a clearer passage from another site even when your page nominally covers the same topic.
Build a prompt map from evidence you already have
You probably can’t open a report that lists every prompt for which an AI system considered, retrieved, or cited your content. You can still build a useful model of that demand by combining several imperfect signals. The discipline is to label them as proxies rather than treating them as a complete record of AI-search activity.
Start with a commercially or strategically important topic, not with every URL on the site. Define the decision, action, or problem that makes the topic valuable to your audience.
Collect the related questions shown in Google’s People Also Ask results. These questions turn a broad keyword into the definitions, comparisons, objections, and follow-ups that people may express in a conversational prompt. A service such as AlsoAsked can extract People Also Ask relationships at scale.
Probe likely follow-up paths in an answer engine. Perplexity’s suggested follow-ups can reveal the next clarification a user may ask, but use them as ideation rather than proof of demand.
Group the collected prompts by shared intent and answer requirements. Prompt-tracking platforms can help at scale; for example, Semrush’s AI visibility workflow consolidates prompts into broader topics so teams can assess intent and brand mentions without managing every wording as a separate campaign.
Keep the original wording even after clustering. A cluster label such as “migration risk” is convenient for reporting, but the exact prompts preserve constraints that may change the answer. “How do I protect rankings during a migration?” and “Which migration mistakes prevent Google from finding a multilingual store?” belong near each other, yet they don’t require identical content.
A practical prompt-map record should contain:
The topic cluster and the user’s dominant intent.
The exact seed questions and long queries behind the cluster.
The constraints, entities, places, or products that alter the answer.
The best current URL for the intent, if one exists.
The missing evidence or explanation on that URL.
Whether the answer depends on current, local, or frequently changing information.
The business action you want the content to support.
Don’t publish a separate page for every prompt. That creates overlapping pages that repeat the same answer and compete for the same intent. Merge wordings when the reader needs the same decision and evidence. Split them only when the correct response, audience, or next action is materially different.
Make each page easy to retrieve, interpret, and cite
Once you have a prompt cluster, turn it into a page brief. The goal isn’t to mimic chatbot language. It is to make the correct answer, its boundaries, and its supporting evidence easy to identify.
State the central answer early. Include the condition that would make the answer change instead of burying qualifications near the end.
Use descriptive headings for genuine subquestions. A heading such as “When server-side rendering is necessary” carries more meaning than “Other considerations.”
Separate facts, recommendations, and uncertainty. If several options can be reasonable, give the decision criteria instead of manufacturing one universal winner.
Use explicit names for products, locations, audiences, and technical concepts. Pronouns and vague phrases may read smoothly, but they make a passage harder to understand when it is retrieved without the surrounding paragraphs.
Support comparisons with consistent criteria. A table is useful when every option can be evaluated on the same attributes; prose is better when the trade-offs aren’t symmetrical.
Connect the page to deeper supporting material with descriptive internal links. The primary page should answer the decision, while supporting pages can carry implementation detail, definitions, or evidence.
Keep time-sensitive claims maintainable. Identify the pages whose answer depends on current product behavior, local conditions, availability, or other changing facts, and assign them a review process.
Technical SEO still determines whether Google can reliably discover and understand the page. Confirm that the intended URL is crawlable and indexable, uses the correct canonical, is linked from the site, and exposes its important answer in accessible page text. If you use JSON-LD, make it a faithful representation of visible content and the entity on the page. Structured data can clarify meaning; it isn’t a switch that guarantees inclusion in an AI answer.
Write for selective retrieval as well as full-page reading. In retrieval-augmented generation, or RAG, a system finds external material and uses it to ground a response. That means a self-contained passage can shape an answer even when the user never opens the page. It also means unsupported, context-dependent copy is a poor candidate for reuse.
Not every prompt triggers retrieval. A system may answer from its existing training data without consulting a fresh page, especially when the question doesn’t require current information. You can’t force a citation by repeating a phrase or adding schema. Concentrate on queries where your content contributes something retrievable: current facts, specific comparisons, local information, original expertise, clear procedures, or a well-supported explanation.
Measure AI visibility without confusing bots, citations, and people
If your reporting doesn’t isolate AI prompts and answer appearances, use a layered scorecard. No single metric tells you whether people wanted the information, a system retrieved it, the answer mentioned you, or the visibility produced a business result.
Measurement layer
Useful signals
What you can conclude
What you cannot conclude
Demand proxy
People Also Ask questions and long Google Search Console queries
Which needs, qualifiers, and conversational patterns deserve investigation
The total number or exact wording of prompts submitted to AI systems
Retrieval
Requests from identifiable user agents and URLs observed as citations
Which pages are accessible to, or selected by, particular systems
That every request represents a person, prompt, recommendation, or citation
Answer presence
Brand mentions, cited URLs, response context, region, and prompt cluster
Where and how the brand appears in sampled answers
Complete market visibility or guaranteed future inclusion
Business outcome
Visits, conversions, qualified enquiries, branded demand, and relevant offline outcomes
Whether visibility is associated with useful action
Perfect attribution when the answer satisfies the user without a click
If you control server or CDN logs, look for identifiable agents such as ChatGPT-User and Perplexity-User. Record the requested URL, response status, and time. These requests can reveal which pages AI services access or use, but they don’t reveal the full prompt by themselves. A bot request isn’t a human session, and it shouldn’t be counted as referral traffic.
Apply the same caution to unusual Search Console patterns. A long query with many appearances and no clicks may look like strong AI demand, yet some patterns can be generated by automated tracking rather than human behavior. Investigate repeated wording, improbable consistency, sudden unexplained volume, and mismatches with the rest of your demand data before building a content plan around it.
For answer monitoring, save more than a visibility score. Retain the exact prompt, location or market, model or surface, date checked, answer context, brand mention, cited URL, and competing domains. Then report at the topic-cluster level. Individual answers and phrasings vary; clusters show whether you consistently appear for a decision your business cares about.
Clicks remain useful, but they are no longer a complete measure of influence. A cited passage may answer the question without sending a visit. A recommendation can also lead to branded search, a later direct visit, or an offline action. Treat those as possible outcomes, not automatic credit. The defensible claim is that the brand appeared in the relevant answer; stronger attribution requires supporting behavioral or business data.
Key takeaways for your next optimization sprint
Map the whole decision behind a prompt, including constraints, comparisons, current information, and likely follow-ups.
Use People Also Ask, long Search Console queries, answer-engine follow-ups, and prompt tools as complementary proxies, not as a complete record of AI demand.
Cluster prompts by intent and evidence requirements. Preserve exact wording, but don’t create a separate URL for every variation.
Make answers self-contained, qualified, crawlable, internally connected, and easy to retrieve. Use JSON-LD to describe visible facts, not to manufacture relevance.
Track demand, retrieval, answer presence, and business outcomes separately. Never equate a crawler request with a person or a citation with a conversion.
Prioritize topics where fresh, local, comparative, or specialized information gives an AI system a reason to retrieve your page.
Start with one high-value decision your audience already brings to Google. Build its prompt map, audit the strongest existing URL, fill the specific evidence gaps, and create the four-layer scorecard before expanding the program. Your next round of work should follow observed gaps in retrieval and answer presence, not the temptation to generate more pages.
If your organic clicks fell after Google began showing AI Overviews, the obvious explanation is that the answer box took the visit. That can happen for a particular query, but it is not a safe diagnosis for your whole site. AI Overview coverage changed sharply during 2025, the mix of affected searches moved further down the funnel, and ads increasingly occupied the same results pages.
You need to separate three questions: Did your visibility change? Did the search results change around an otherwise stable ranking? Did the traffic change without reducing business value? The answers determine whether you should rewrite content, improve search-result presentation, defend branded queries, coordinate with paid search, or leave a page alone.
What changed in 2025 – and what it did not prove
AI Overview exposure was not a one-way rollout. In a Semrush analysis covering more than 10 million keywords, AI Overviews appeared for 6.5% of queries in January, rose to nearly 25% in July, and fell below 16% by November. A traffic change measured against the July peak could therefore look very different from one measured against January or November.
Treat those figures as evidence of volatility, not as a current coverage benchmark for every website. A page can lose AI visibility because Google stopped generating an overview for the query, because another domain replaced it inside the overview, or because the underlying organic result moved. Those are different events and require different responses.
The broad zero-click narrative also needs more care. AI Overviews tended to appear on searches that were already likely to end without a click. Yet when the same keywords were compared before and after an overview appeared, zero-click searches declined from 33.75% to 31.53%. That does not prove AI Overviews create clicks. It does show why you should not assume that every overview suppresses traffic.
Your sitewide organic total cannot tell you which mechanism is operating. Before changing a page, inspect the affected query cohort and the live result page. Otherwise, you may weaken content that still ranks and converts because a blended dashboard made a temporary search-feature change look like a content problem.
Diagnose the loss before changing your content
Start at the date the decline became visible. Export comparable query and page data from Google Search Console, keeping country and device filters consistent. Do not begin with the site’s average position or total clicks; averages mix branded searches, informational articles, product queries and pages with very different exposure to AI results.
Build the affected cohort. Identify the queries and landing pages responsible for most of the lost clicks. Keep unaffected pages as a comparison group.
Label search intent. Mark each material query as informational, commercial, transactional or navigational. Also separate branded from non-branded searches.
Record the result-page layout. Note whether an AI Overview appears, whether your domain is linked from it, where the organic result sits, which ads appear, and whether another search feature is competing for attention.
Compare the component metrics. Review impressions, clicks, click-through rate and average position for the same query-page combinations. Do not substitute a sitewide average.
Connect the cohort to outcomes. Compare leads, sales, sign-ups or another relevant conversion. A click decline matters differently when conversions fall with it than when low-value visits disappear while outcomes hold.
Use the pattern below as a diagnostic route, not as automatic proof of causation.
Observed pattern
Investigate first
Next check
Impressions are stable, average position is broadly stable, and CTR falls
Search-result presentation and crowding
Compare AI Overview, ad and other feature presence for the affected queries
Impressions fall while CTR is broadly stable
Search demand, query coverage or indexing
Separate lost queries from pages that still receive impressions
Clicks and average position fall together in a page-query cluster
Traditional organic visibility
Review relevance, competing pages, technical accessibility and content quality
Clicks fall but conversions remain stable
Traffic mix rather than raw volume
Calculate whether the lost cohort previously contributed meaningful outcomes
Branded-query CTR changes
Navigational result-page control
Inspect the overview, ads, official pages and third-party brand information together
This process prevents a common reporting error: treating ranking, AI inclusion and traffic as interchangeable. Track them as separate observations. A ranking report tells you where an organic result appeared; an AI visibility record tells you whether the brand or page appeared in the generated answer; analytics tells you what visitors did after clicking.
Rebuild your visibility map around search intent
AI Overview optimization can no longer be confined to informational blog posts. Informational searches represented 91% of AI Overview queries in January 2025 but 57% by October. Over the same period of expansion, the commercial share rose from 8% to 18% and the transactional share from 2% to 14%. Navigational exposure climbed from under 1% in January to more than 10% by November.
That shift changes which pages deserve monitoring. A blog-only dashboard will miss AI visibility around product evaluation, purchase decisions and direct brand searches. Add category pages, product or service pages, comparison pages, pricing information, support content and official brand pages to your query map.
For informational queries, answer the main question near the start, define important terms, show the reasoning or evidence, and give the reader a useful next step. Do not bury the answer beneath a long preamble written only to retain the visit.
For commercial queries, make evaluation criteria explicit. State who an option suits, where it does not fit, what constraints matter, and how alternatives differ. Generic claims give a search system little concrete information to represent.
For transactional queries, keep offer details, availability, requirements, limitations and the conversion path clear. The page should resolve purchase uncertainty as well as target a keyword.
For navigational queries, make official brand facts easy to verify. Keep names, product descriptions, contact details, location information and support destinations consistent across the pages you control. Monitor brand-plus-product and brand-plus-support searches, not only the bare company name.
The navigational increase deserves special attention because it turns AI visibility into a reputation and brand-representation issue. If an overview intercepts a destination search, the question is no longer only whether you rank first. You also need to know what Google says about the organization, which pages it links, and whether the answer helps the searcher reach the correct destination.
Prioritize by business value rather than overview frequency alone. A high-volume informational query may produce little commercial impact, while a smaller product or branded query may sit close to a decision. Your reporting should preserve that distinction instead of assigning every appearance the same visibility score.
Treat AI, ads, verticals and page quality as one system
AI Overviews increasingly shared the results page with paid placements. Ads appeared alongside roughly 3% of AI Overviews in January 2025 and about 40% by November. Roughly a quarter of AI Overview results pages placed ads at the bottom of the overview.
This matters when you interpret CTR. If an overview and additional ads appeared at the same time, you cannot attribute the entire change to the generated answer. Keep a shared SERP record for SEO, paid search and analytics teams: query, intent, device, AI Overview presence, domain inclusion, ad presence, organic position, landing page, clicks and business outcome. That record lets you distinguish feature crowding from an organic ranking loss and exposes cases where paid and organic teams are reacting to the same change independently.
If your site operates in a lower-exposure category, do not copy the monitoring budget or traffic assumptions of a science publisher. If it spans several categories, do not assign one AI risk score to the entire domain. Build cohorts around your actual topics and query types, then prioritize the intersection of frequent AI exposure, meaningful traffic change and commercial value.
Once the diagnosis points to a page-level opportunity, improve the page for both extraction and human decision-making:
Give the primary question a direct, self-contained answer before expanding into nuance.
Use descriptive headings that reflect the decisions or subquestions a searcher actually has.
Keep claims, definitions, product attributes and comparisons internally consistent.
Support important assertions with evidence the reader can inspect, rather than repeating an unsupported consensus statement.
Make authorship, organizational responsibility and update context clear where trust affects the decision.
Remove sections that restate the same answer without adding evidence, criteria or a next action.
Use JSON-LD only when the schema type matches the page and the marked-up facts are visible to readers. Validate the markup, but do not treat valid schema as a guarantee of AI Overview inclusion.
At enterprise scale, AI visibility is an upstream acquisition signal, not the final outcome. It becomes operationally useful when SEO, content, paid media and analytics work from a shared visibility process. Assign an owner to the query set, define how SERP observations are recorded, and connect changes to conversions. A large visibility score without that chain can create activity without explaining business impact.
Key takeaways and your next move
Do not use a sitewide traffic decline as proof that AI Overviews took your clicks; isolate the affected queries and inspect their result pages.
Track organic position, AI Overview inclusion, ads, clicks and conversions separately. Each metric answers a different question.
Expand monitoring beyond informational content because commercial, transactional and navigational queries gained substantial AI Overview exposure during 2025.
Judge CTR within comparable query cohorts. Aggregate zero-click assumptions can conceal different behavior on the same keywords.
Prioritize pages where AI exposure, measurable performance loss and business value overlap; raw appearance counts are not a strategy.
Use clear answers, verifiable evidence and accurate structured data to improve machine readability without weakening the page for human visitors.
Begin with the highest-value query cohort where impressions held but CTR changed. Capture the current result-page layout, check AI and ad presence, and compare business outcomes before editing the page. That gives you a defensible baseline for the next change Google makes – and a way to respond without mistaking every traffic fluctuation for an SEO emergency.
If your Black Friday plan stops at rankings, feeds, paid media, and conversion rate, it now has a blind spot. A shopper can ask an AI system to narrow a category, compare products, judge whether a discount is worthwhile, and recommend where to buy – without following the search journey you designed.
Your job is not to make an AI repeat your promotion. It is to make your products easy to identify, compare, and verify while demand moves from early research to live deal hunting. That requires coordinated work across your own site, retailers, marketplaces, review coverage, video, and genuine customer discussion.
Black Friday creates two different AI demand states
Before Black Friday, shoppers are reducing a large market into a shortlist. Their questions tend to concern suitability: which product fits a use case, what features matter, which compromises are acceptable, and whether waiting for a sale makes sense. When the event begins, the task changes. Price, availability, seller credibility, current sentiment, and the quality of the deal become more important.
Those percentages do not establish a permanent formula for every category or model. They do expose a useful operating distinction: the content that builds a shortlist is not sufficient on its own when shoppers want current confirmation from other people.
Build your campaign around four information layers:
The identity layer explains what your brand sells, which categories it belongs in, and who its products are for.
The decision layer supplies specifications, use cases, limitations, compatibility details, and defensible comparisons.
The offer layer states the current price, discount terms, sale window, availability, fulfillment conditions, and applicable returns information.
The verification layer gives shoppers independent evidence through reviews, demonstrations, retailer listings, comparison coverage, and legitimate customer discussion.
The first two layers should be settled before promotional demand arrives. The offer layer must be updated whenever the commercial facts change. The verification layer takes longer to earn, so it cannot be manufactured credibly on launch day.
Make every offer answerable without reconstruction
An AI system should not have to combine a slogan on your homepage, specifications in a PDF, a discount in a banner, and shipping terms in a support page to explain your offer. Every extra reconstruction step creates another opportunity for omission, confusion, or a stale answer.
Start at the homepage because it is more than a navigational doorway. Within the examined brand-site citations, homepages accounted for 40%. Give that page a plain statement of what the brand is, the categories it serves, the customer problems it solves, and the main paths to product information. A clever campaign line can support that explanation, but it should not replace it.
Then audit each priority product or offer page in this order:
Use the exact product name and model consistently in the title, visible copy, structured data, retailer listings, and supporting content.
State what the product is and who it suits near the top of the page. Do not make the reader infer the category from branding language.
Present specifications as labeled facts. Include the dimensions, materials, capacity, compatibility, included components, or technical requirements that actually drive a decision in your category.
Explain the important tradeoffs. A page that identifies who should not buy the product can be more useful than one that describes every shopper as an ideal customer.
Place the live offer in visible text. Include the current price, reference price where applicable, conditions, start and end information, seller, stock state, and fulfillment details that a buyer needs to interpret the promotion.
Add concise questions and answers for real research intents: compatibility, setup, maintenance, warranty, returns, common alternatives, and differences between adjacent models.
Provide evidence close to the claim it supports. Demonstrations should show the use case, while reviews and technical documentation should be clearly attributable and reachable.
Keep stable product facts separate from volatile promotional facts in your content workflow. The product’s dimensions should not change because a sale begins, but price and availability might. Assign ownership accordingly: merchandising maintains the offer state, while product or content teams maintain the underlying facts.
Structured data can make those facts less ambiguous to machines, but it cannot rescue incomplete visible content. Product and Offer markup should agree with the page a shopper sees. If a price, availability value, model identifier, or seller differs between the markup and the page, the markup has added conflict instead of clarity.
Finish with a manual extraction test. Give someone who did not build the page the URL and ask them to answer: What is this product? Who is it for? Why would they choose it over the closest alternative? What exactly is the Black Friday offer? What restriction could change the decision? If any answer requires another tab or an assumption, the page is not finished.
Build comparison coverage before the promotion starts
Brand pages are good at establishing first-party facts. Shopping recommendations require a second job: organizing choices and reducing uncertainty. That is why AI systems repeatedly draw from retailers, review publishers, video platforms, and community conversations when they construct commercial answers.
Retail concentration matters too. Generalist retailers held 48% of retail citations, while electronics specialists held 23%. Large retailers have broad assortments, familiar identities, and enough product information to answer many different shopping questions. A smaller brand is unlikely to reproduce that footprint, but it can make its category knowledge and product distinctions much easier to reuse.
Create comparison pages around decisions, not around the phrase “best product.” A useful comparison should tell the reader:
Which products are genuinely comparable and which belong to a different use case.
What each option is best suited to, using a stated criterion rather than a vague superlative.
Which specifications materially change the experience.
What the buyer gives up by choosing the cheaper, smaller, faster, or more capable option.
Whether accessories, subscriptions, installation, or compatibility requirements affect the practical cost.
Which facts are stable product attributes and which are temporary Black Friday conditions.
Publish first-party comparisons even when an independent reviewer would be more persuasive. Your version establishes accurate entities, specifications, and distinctions that other people can check. It should disclose its perspective and link to the underlying product details rather than pretending to be neutral.
For third-party coverage, prioritize relevance over raw volume. Give suitable reviewers and publishers clean model names, current specifications, images, documentation, and access to products where your normal review policy allows it. Correct factual errors without trying to dictate conclusions. Inclusion in a trusted comparison is valuable because the comparison answers a real decision, not merely because it creates another brand mention.
Treat off-site evidence as part of product information
Your website can declare what a product does. It cannot independently establish how the product behaves in ordinary use or how buyers feel about its compromises. AI shopping answers often seek that corroboration elsewhere.
Each environment contributes a different kind of evidence:
YouTube can show setup, scale, sound, motion, results, and other experiential details that are difficult to communicate in a specification table. Use accurate titles and descriptions, identify the exact model, and make spoken explanations clear enough to stand without promotional visuals.
Retailer and marketplace listings connect the product to a category, seller, price, reviews, and comparable inventory. Keep identifiers, variants, specifications, and images consistent with your own site.
Review coverage organizes alternatives and makes tradeoffs explicit. Give reviewers enough factual material to distinguish models without forcing them to decode your catalog.
Community conversations reveal recurring questions, edge cases, frustrations, and unexpected use cases. Use those conversations to improve product information and support. Do not simulate participation or manufacture endorsements.
Consistency is the operational priority. If your site calls a product one name, a retailer shortens it, a video uses a family name, and marketplace variants omit the model number, you have created several weak identities instead of one strong one. Maintain a shared product record containing the approved name, model identifier, category, key specifications, variant labels, current imagery, and canonical URL. Give every channel owner access to it.
Do not turn this into a backlink-counting exercise. A mention that does not help identify, compare, or verify the product contributes little to the shopping decision. Audit off-site presence by question instead: Where can a shopper see the product used? Where can they compare it with the nearest alternative? Where can they verify specifications? Where can they find credible discussion of its limitations?
Run a two-phase AI visibility operation
Black Friday AI optimization should operate in a preparation phase and a live phase. The preparation phase builds retrievable facts and comparison context. The live phase protects accuracy while offers, availability, and public conversation change.
Before the promotion, build a fixed prompt set from customer decisions rather than from your target keywords alone. Include category discovery, a constrained use case, a direct product comparison, a compatibility question, a value question, and a deal-verification question for every priority category. Keep the wording stable enough that later results are comparable.
Run those prompts separately on the AI platforms your customers are likely to use. Do not collapse their responses into one score. In the observed Black Friday sample, Gemini responses averaged 606 words, OpenAI responses averaged 401, and Perplexity responses averaged 288. Those are sample characteristics, not permanent product specifications, but they show why a citation or mention can play a different role on each platform.
Use one tracking row for each prompt and platform. Record:
The exact prompt, model or product name, and time of the check.
Whether the brand and correct product appear.
How the product is framed: recommended, compared, merely listed, or excluded.
Which URLs support the answer.
Whether the price, specifications, seller, availability, and promotion terms are accurate.
Which competitor or third-party page supplied information you did not make easy to find.
The correction required: page content, structured data, marketplace data, comparison coverage, video, or support documentation.
At sale launch, rerun the deal and verification prompts. Repeat the check after any material price, inventory, seller, or terms change. If an answer is wrong, correct the authoritative page and connected listings first. A prompt variation may produce a different answer, but it does not repair the underlying information conflict.
Judge progress by failure mode rather than by a single visibility number. A missing brand is a discovery problem. The wrong model is an identity problem. An incorrect price is a freshness problem. A competitor winning every comparison may indicate weak decision content or stronger independent corroboration. Each diagnosis leads to different work.
Key takeaways
Plan separately for pre-sale research and live deal verification because the source mix changes when Black Friday begins.
Give every priority offer a clear identity, complete decision facts, current commercial terms, and evidence a shopper can verify.
Build comparisons around use cases and tradeoffs, not unsupported claims that a product is “best.”
Coordinate product information across your site, retailers, marketplaces, video, review coverage, and community support.
Test the same customer decisions across AI platforms and classify failures before choosing a fix.
Before your next promotion, choose one prompt for each major customer decision in your highest-value category. Run the set when product pages are frozen, again when the sale launches, and whenever a material offer fact changes. The gaps you find will give your content, merchandising, SEO, marketplace, and communications teams a concrete Black Friday worklist – before shoppers ask AI to make the choice for them.
Today, I am thrilled to share that Search Engine Land is celebrating its 19th anniversary!
Nineteen years is an incredible milestone. For almost two decades, we have been diving deep into the ever-evolving world of search engines, always striving to make sense of the changes and challenges Google and the search industry present.
This year, 2025, has been one of the most transformative since our launch in 2006. The rapid pace of change has been exhilarating.
Through it all, our mission remains steadfast: to provide clear news, insightful analysis, and practical guidance to help you navigate the world of search.
Before we look to the future, I want to express my heartfelt thanks for your support and reflect on the past year with you.
Thank you for reading
Sincerely, thank you for being with us.
Every day, we focus on you: what you need to know, what really matters, and what changes will impact your work today or your strategy months down the line.
Our goals include:
Focusing on meaningful stories, not filler.
Delivering news clearly and quickly.
Providing essential context and expertise.
Being a dependable resource in a fast-changing industry.
Helping you anticipate where search is heading, even when it’s unclear.
If you haven’t yet, I encourage you to subscribe to our daily newsletter for a curated summary of all things search, helping you stay updated without feeling overwhelmed.
Thank you to the Search Engine Land team
Our team’s passion is what has driven our success for almost two decades.
Though small, our team accomplishes significant and impactful work because we are mission-driven and dedicated to search.
I extend my greatest thanks to:
Barry Schwartz. With 22 years of experience, Barry’s passion for search ensures complex topics become understandable. He is indispensable.
Anu Adegbola. Focusing on paid media, Anu offers clarity amidst constant changes with her insightful writing.
Angel Niñofranco. Angel plays a crucial role in our SME articles through his coordination and editorial oversight.
Kathy Bushman. Kathy’s behind-the-scenes expertise ensures SMX events are seamless and valuable.
And to the entire team at Third Door Media within Semrush, whether or not your name appears here, your contributions are invaluable.
Top highlights from the past year
Despite the uncertainties of this year, Search Engine Land thrived, thanks to the trust of our community.
SMX Advanced returned in person for the first time in 6 years
This was arguably the highlight of the year. SMX Advanced’s return in person after six years was electrifying.
With attendance surpassing expectations, the sessions were dynamic, and conversations felt like reunions for the search marketing community. It was clear that we all missed these face-to-face exchanges about AI, Google’s updates, and more.
We learned again that when great minds gather, extraordinary things happen. We eagerly await our next gathering in Boston, June 3-5.
Defining industry coverage of AI Overviews and the new era of search
This year, more than ever, transformed the search landscape. We’ve provided the clarity and reporting needed in this evolving environment.
Our readers rely on us for insights during times of change, and we take pride in shaping the industry’s future understanding of search.
Subject Matter Expert (SME) program growth
This year saw a surge of new and returning readers turning to us for insight into SEO and PPC shifts, from AI to SERP experiments.
Our growth owes much to our fantastic contributors, and I extend my gratitude for their impactful work.
Looking ahead: What’s next for Search Engine Land
As we embrace our 19th year, our resolution is steadfast: to offer unparalleled coverage of search-related topics.
This year, you can anticipate:
Continued breaking news on SEO, PPC, AI, and more.
In-depth analysis, guides, and contextual explainers on industry evolution.
SMX events tailored around the nuances of AI-driven search.
Enhanced expert viewpoints, data, and market clarity.
Mark your calendars for:
SMX Advanced: June 3-5
SMX Next: Nov. 18-19
We have much in store for you, with the aim of equipping you with the insights necessary for your best work.
A brief look back to where it all began
Launched on Dec. 11, 2006, Search Engine Land began with a vision of search as a vast community. A place of exploration, connection, and evolution. Over these years, it’s grown beyond our expectations.
The mission remains the same:
Search Engine Land is your destination to remain informed, educated, and connected within the world of modern search engines.
Thank you for 19 incredible years
From everyone here at Search Engine Land and Semrush, thank you for your readership, engagement, and passion for the evolving world of search.
Here’s to a promising rest of 2025 and a remarkable 2026.
If you are budgeting for AI visibility on the assumption that every product mention is close to a sale, stop and reclassify the opportunity. Commercial demand exists in AI chats, but much of it appears while people are framing a problem, weighing approaches, or trying to succeed with something they already bought.
Your job is to recognize those moments without forcing a sales funnel onto every conversation. That changes which pages you prioritize, how you structure an answer, where you place the next action, and what you count as success.
Commercial intent is a minority, but it is not one moment
Across a corpus covering 4.4 billion characters, 613 million words, and 3.9 million conversation turns, people used AI heavily for tasks such as planning, brainstorming, analysis, learning, transformation, and creation. Those activities may happen at work or mention a product, but that does not automatically make them commercial.
Within a categorized sample of 24,259 sessions spanning 42 intent categories, 64.6% did not fit a purchase funnel, while 35.4% showed some form of commercial intent. The useful correction is not that AI chats have no commercial value. It is that commercial value is distributed across several different jobs, most of which are not an immediate purchase request.
Awareness accounted for 10% of the categorized sessions and consideration for 8.5%. Together, those early stages represented 18.5% of all sessions and the largest block of commercial activity. Discovery accounted for 4.1%, decision support for 2.8%, transactional support for 4.8%, and post-purchase needs for 5.1%.
That distinction matters when you set priorities. If your AI strategy watches only prompts containing words such as buy, price, best, or demo, it will miss people who are still deciding what kind of solution they need. It will also miss existing customers asking how to configure, use, integrate, or repair what they own.
Do not treat the percentages as a universal forecast for every market. They describe the analyzed corpus, not the exact intent mix for your category. Use them to challenge an overly transactional strategy, then classify the questions that appear in your own sales, support, search, and customer research.
Classify the user’s job before choosing the content
A noun is not an intent signal. A user can mention your category while asking for writing help, summarization, technical instruction, product evaluation, or troubleshooting. Classify the job being done before deciding whether the conversation belongs in a commercial funnel.
Intent class
Observed share
What the user is trying to do
What your content should accomplish
Outside the purchase funnel
64.6%
Create, learn, analyze, plan, transform, or converse without making a product choice
Complete the requested task honestly; introduce a commercial path only when it is genuinely relevant
Awareness
10%
Name a problem, understand its causes, or learn what kinds of solutions exist
Define the problem, explain when it matters, and make the available approaches understandable
Consideration
8.5%
Compare approaches, requirements, or tradeoffs
Provide selection criteria, limitations, alternatives, and use-case fit
Discovery
4.1%
Find products, providers, or options in a category
Help the user build a defensible shortlist without hiding eligibility criteria or constraints
Decision support
2.8%
Choose among known options
Supply verifiable details about fit, evidence, implementation, cost factors, and risk
Transactional support
4.8%
Complete or manage a commercial action
Remove uncertainty about requirements, process, timing, and what happens next
Post-purchase
5.1%
Set up, use, improve, or troubleshoot something already acquired
Help the customer reach the intended result and recover from predictable failures
The percentages in the table are rounded shares of the categorized sample. The user-job descriptions and content responses are practical applications of those intent classes.
Context is decisive. Create a launch brief for this product is primarily a creation task. Which type of platform should our distributed team use to manage a launch? is consideration. Why did this feature stop working after setup? is post-purchase. The same category terms can appear in all three prompts, but only the latter two have an explicit relationship to choosing or owning a solution.
Use a strict operational rule: label a conversation commercial only when the user is making an economic choice, evaluating a solution, completing a transaction, or seeking help with something already acquired. Do not inflate your opportunity estimate by treating every workplace task as latent demand.
Build for exploration and ownership, not just selection
Early-stage content should make the decision legible
Awareness and consideration together accounted for 18.5% of all categorized sessions. This is where product-led content often arrives too early. A user who is still defining the problem does not need an unsupported claim that your product is the answer. They need enough structure to decide whether the category is relevant at all.
A useful awareness or consideration page should do the following:
Answer the initiating question immediately. State the practical answer before company history, positioning, or a lead form.
Define the decision context. Identify who the advice applies to, the conditions that change it, and any prerequisites the user may not have mentioned.
Separate symptoms from causes. Help the user avoid buying a solution for the wrong problem.
Expose the criteria that change the choice. Explain requirements, constraints, tradeoffs, and cases in which a simpler approach is sufficient.
Include credible alternatives. A comparison is more useful when it covers different approaches, including doing nothing yet, rather than presenting a disguised product pitch.
Provide a natural next question. Link the problem explanation to criteria, the criteria to options, and the options to decision evidence.
The first answer carries unusual weight. The median conversation in the corpus had two turns and 430 words, and more than 80% of chats stayed below 1,000 words. Many users therefore do not spend a long sequence teaching the assistant their context. Your page should state its audience, assumptions, constraints, and core answer clearly enough to survive a short exchange.
This is also where answer-engine optimization and conversion writing need to part company for a moment. The strongest opening is the one that resolves the question accurately. The commercial handoff comes after the user can see why a category, method, or product deserves consideration.
Post-purchase content belongs in the commercial strategy
Support content should be designed around the customer’s actual failure state, not your internal feature taxonomy. A page titled with the symptom a user can observe is more useful than one that assumes they already know which component caused it.
Name the symptom, task, or desired outcome in the title and opening.
State the applicable product state, configuration, prerequisites, and access requirements.
Put the resolution steps in the order the user must perform them.
Describe the expected result so the user can verify that each meaningful step worked.
Branch explicitly when different causes require different fixes.
Say when self-service should stop and what information support will need.
Connect the fix to related setup or usage guidance without turning the page into a sales pitch.
Where security and account privacy allow it, publish general help in accessible, indexable page content. Keep account-specific data and privileged actions behind authentication. An AI visibility goal never justifies exposing information that should remain private.
Audit AI demand by prompt, page, and outcome
You do not need to guess whether your opportunity is mostly awareness, decision support, or ownership. Build an intent inventory from questions people already ask, then connect each question to a page and a measurable next step.
Collect real questions. Pull wording from site search, sales conversations, support records, community discussions, product research, and known AI referrals. Preserve the original phrasing instead of rewriting everything as a target keyword.
Assign one primary job. Label each question as non-funnel, awareness, consideration, discovery, decision, transactional support, or post-purchase. Record a secondary intent only when it changes the answer the user needs.
Map the best existing page. Choose the page that should answer the question, not merely the page currently ranking for adjacent terms. A product page is not automatically the right destination.
Find coverage and answer gaps. Mark questions with no page, pages that bury the answer, unsupported claims, missing limitations, stale instructions, or no sensible continuation.
Repair the visible content first. Make the answer, scope, evidence, and next step explicit. Structured data should reflect what a user can actually see on the page; it cannot manufacture commercial intent or compensate for an evasive answer.
Run repeatable prompt checks. Log the exact prompt, assistant, exposed model or version, date, language or market, answer, brand representation, and cited URLs. A single response is an observation, not a stable visibility benchmark.
Measure the outcome appropriate to the stage. Evaluate awareness content by accurate inclusion and progression to deeper evaluation. Evaluate decision content by qualified actions. Evaluate post-purchase content by successful task completion and reduced escalation where those signals are available.
Keep visibility and progression as separate measures. Visibility asks whether the assistant represents the right answer, entity, or page. Progression asks whether the user then reaches a useful next step. Combining them into one score hides whether you have a retrieval problem, an answer-quality problem, or a conversion-path problem.
Referral traffic is also incomplete by definition. You can observe a visit only when a user follows a link; an interaction that ends inside the chat produces no referral session. Use AI referral data as evidence of visits and downstream behavior, not as a complete count of AI influence.
Finally, compare like with like. Do not blend troubleshooting prompts and product-selection prompts into one visibility rate, then judge both by purchases. Segment the prompt set by intent, page type, market, and user state. The resulting report will tell you which content is failing and what kind of repair it needs.
Key takeaways
Commercial intent appeared in 35.4% of the categorized AI chat sessions, while 64.6% did not fit a purchase funnel.
Awareness and consideration formed the largest commercial block, so problem framing and selection criteria deserve more attention than purchase language alone.
Post-purchase demand exceeded both discovery and decision support, making setup and troubleshooting content part of AI commerce strategy.
Classify the user’s job, not the presence of a product or business keyword.
Because the median chat was short, make the first answer self-contained, scoped, and useful before asking the user to take a commercial action.
Measure visibility, answer accuracy, progression, and business outcomes separately for each intent stage.
Start with your own prompt inventory. Find an early-stage cluster and a post-purchase cluster with weak coverage, repair the answers and their handoffs, and retest them consistently. You will see where AI visibility can support demand and where usefulness should stand on its own.
Your analytics can show no traffic at the exact moment an AI answer starts putting your brand into a buyer’s consideration set. The inverse happens too: a citation looks impressive in a visibility tracker but sends no qualified visitor and supports no observable decision.
The fix is not to choose between citations and traffic. You need a measurement chain that separates presence, citation, referral, and commercial value. Once those signals have distinct definitions, you can see where your visibility is working, where the journey stops, and what to improve next.
A citation is not a click, and a mention is not a citation
AI search visibility is often compressed into one score. That hides four different events:
A mention occurs when an answer names your brand, product, expert, or other identifiable entity.
A citation occurs when the answer attributes information to your domain or links to one of your URLs.
A referral occurs when a person follows an AI-generated link and reaches your site in a way you can observe.
An outcome occurs when that visitor completes a meaningful action, such as starting a trial, requesting a quote, buying a product, subscribing, or entering a qualified sales process.
These events do not always happen in sequence. An answer can mention your brand without linking to it. It can cite a supporting page without naming the brand prominently. A person can encounter your brand in an answer, return later through branded search, and leave no direct AI referrer. A crawler or agent can also retrieve a page without producing a human visit.
Choose the primary metric from the job you expect the content to do. For discovery content, measure whether the brand appears accurately in relevant answers. For evidence-led content, measure citation coverage and the contexts in which the page is used. For decision pages, measure qualified referrals and outcomes. Do not grade all three content types against the same click target.
This distinction matters because generative systems can handle much of the early research journey before a person reaches a website. Traditional impressions, sessions, and click-through rates therefore describe only part of the path. Pricing, comparison, product, and validation pages may receive the eventual visit, while explanatory content did the earlier work of making the brand visible.
Build a visibility scorecard with separate denominators
A useful scorecard starts with a fixed set of prompts that represents the decisions your audience actually makes. Include non-branded prompts. A test set dominated by your company name will measure retrieval of a known entity, not discovery among alternatives.
Group prompts by intent before running them:
Discovery prompts ask what a problem is, why it occurs, or how to approach it.
Evaluation prompts ask about criteria, methods, categories, risks, or suitable options.
Comparison prompts weigh named alternatives, features, costs, or trade-offs.
Validation prompts look for reviews, evidence, limitations, implementation details, or compatibility.
Transaction prompts ask where to buy, what something costs, or how to begin.
Run the same prompt set separately in each engine. Preserve the wording and record the date, engine, answer, brand mentions, cited URLs, cited domains, source type, and intended landing page. If language, location, account state, or another test condition changes, record that as well instead of mixing the results into one trend line.
One industry analysis covered 250 million AI-generated responses. That scale is a useful warning against treating a few favorable screenshots as a baseline. Generative answers can vary, so repeat the same test design and compare like with like.
Signal
How to calculate it
What it tells you
Common misreading
Mention coverage
Eligible prompt runs containing the entity divided by all eligible prompt runs
Whether the brand enters relevant answers
Treating any mention as positive without checking context or accuracy
Owned citation coverage
Eligible prompt runs citing an owned domain divided by all eligible prompt runs
How often your site supplies answer evidence
Calling a citation a visit
Citation share
Unique citations to your domain divided by all unique citations in the tested answers
Your presence within the observed source set
Presenting test-set share as market-wide share
Qualified referral rate
AI-referred visits meeting your quality criteria divided by all tracked AI referrals
Whether arriving visitors fit the page’s intended audience
Judging value from raw sessions alone
Outcome rate
Desired outcomes divided by tracked AI referrals
How observable AI traffic contributes to the business
Crediting every later direct or branded visit to AI
Define a unique citation consistently. Counting the same URL several times inside one answer can inflate the result, so a practical default is one occurrence per unique URL per response. Keep domain-level and URL-level views. The domain view shows authority concentration; the URL view reveals which content actually earns the citation.
Do not roll every prompt into a single average too early. A brand may be absent from discovery prompts but dominant in transaction prompts. That is a very different problem from broad underperformance. Report by engine, intent, topic cluster, market, and source role first. Use an overall score only as a navigation aid.
Match your source strategy to the engine and the prompt
AI engines do not necessarily choose the same kinds of evidence for the same request. In a 2025 holiday-season analysis of tens of thousands of identical ecommerce prompts, retailer sources appeared in about 4% of Google AI Overview results and 36% of ChatGPT results. Google leaned more heavily on YouTube, Reddit, Quora, and editorial sources, while ChatGPT more often surfaced retailers, brand pages, and manufacturer pages.
That finding is specific to ecommerce prompts from that holiday period. It is not a universal rule for B2B software, healthcare, local services, finance, or every future version of either engine. The actionable lesson is narrower: segment your citation strategy by platform and query type instead of assuming one source profile applies everywhere.
Build a source-role map before creating more content
For each important prompt cluster, label every recurring citation as an owned brand source, retailer, editorial publication, community discussion, video source, or another relevant category. Then look for the missing role.
If owned pages are repeatedly cited, identify the exact passages and page formats supporting the answers. Maintain those facts instead of replacing a successful page simply because it is old.
If editorial and video sources dominate, give legitimate reviewers accurate specifications, evidence, and access to the material they need. Independent coverage cannot be replaced by publishing another self-authored claim.
If community discussions recur, improve the underlying product information and customer experience that people can discuss. Manufactured participation creates reputation risk and does not provide durable corroboration.
If retailer pages dominate, make product names, variants, attributes, and purchasing details consistent across the manufacturer site and authorized listings.
If competitors appear through a source type you lack, close that source-role gap rather than copying the competitor’s wording.
For retail research prompts following the observed Google pattern, an owned product page alone may not cover the sources the answer prefers. You may also need accurate independent reviews, useful demonstrations, and authentic community evidence. For ChatGPT prompts following the observed retail pattern, complete brand, manufacturer, and retailer pages deserve closer attention because those sources appeared much more often.
Validate both patterns against your own prompt set. Platform averages are a starting hypothesis, not a substitute for sector-specific observation.
Keep discovery content even when its clicks decline
Across an analysis of more than 7.2 million sessions to industry blog content, pricing and cost pages showed the strongest growth, comparison content also gained, and traditional guides declined. The scope matters: this was blog performance, not every content format, and the pattern does not by itself prove that AI caused the changes.
Deleting top-of-funnel content would still be the wrong response. Discovery material can supply the definitions, criteria, and explanations that generative engines use before a person is ready to visit. If you remove it because direct sessions fell, you may also remove the material capable of earning early mentions and citations.
Give each content layer a clear job:
Discovery pages should answer a narrow question directly, state their scope, distinguish easily confused concepts, and lead to the next decision.
Evaluation pages should provide criteria, trade-offs, limitations, and evidence a buyer can use to narrow the field.
Decision pages should expose pricing, comparisons, compatibility, availability, implementation requirements, or another concrete next step appropriate to the offer.
Product and service pages should keep names, claims, attributes, and calls to action consistent with the supporting content that introduces them.
Connect these layers explicitly. A cited explainer should link to the relevant comparison or decision page, while the decision page should link back to the evidence behind its claims. This gives a human visitor a coherent path even when the AI engine exposes only one page.
Use JSON-LD as a consistency layer, not as a citation counter. Mark up entities and attributes that are visible on the page, and keep names and relationships consistent with the readable content. Deployment is not the result. The result is whether the intended entity is understood accurately, cited in the right context, and connected to a useful next action.
Turn sparse AI referrals into commercial evidence
AI referral volume can be small while the visitors who do arrive are close to a decision. Generative systems may complete much of the discovery and evaluation work before sending a person to a pricing, comparison, calculator, retailer, or product page. Measure the quality of that arrival before deciding the channel has little value.
Build attribution in layers:
Create an analytics channel for observable AI referrers. Keep the underlying source visible so you can compare engines instead of hiding them under one label.
Record the landing page, content type, engagement events, and business outcome. A visit to a decision page should not be evaluated like a visit to an explainer.
Separate human referrals from bot and agent retrievals in server-side reporting. A fetch can indicate access or use, but it is not a human session and should not be counted as one.
Pass the original source into your CRM or lead system when your setup allows it. This lets you inspect lead quality, pipeline progression, and revenue instead of stopping at form completion.
Add a short self-reported discovery field where the value of the decision justifies the extra question. Treat the answer as complementary evidence because memory and channel overlap make it imperfect.
Not every AI-influenced journey will carry a usable referrer. A person may see a mention, open a separate tab, search the brand, or return later. Branded search growth, direct navigation, and self-reported discovery can help you notice that spillover, but they do not prove that a particular answer caused a particular visit.
Keep direct attribution and assisted evidence in separate columns. The first contains observable referrals and outcomes. The second contains correlated signals such as stronger branded demand following improved answer visibility. Combining them produces an impressive number but a weak decision tool.
Evaluate referral value with metrics that reflect your business:
Qualified visit rate: the share of tracked AI visits that meet your engagement or audience criteria.
Decision-action rate: the share that completes the action the landing page was designed to support.
Lead acceptance or sales progression: whether AI-sourced leads remain useful after the initial conversion.
Observable pipeline or revenue: the commercial result tied to tracked referrals under your normal attribution rules.
Landing-page concentration: which pages and intent stages receive the traffic, even when total volume is limited.
Compare equivalent journeys. An AI referral landing on a pricing page should be compared with other channels entering that pricing page or the same intent stage, not with the sitewide average. Otherwise, differences in landing intent can be mistaken for differences in channel quality.
Use the same discipline when evaluating citations. A citation on a broad educational prompt and a citation on a named comparison prompt have different commercial proximity. Report both, but do not assign them the same expected referral value.
Key takeaways
Measure mentions, citations, human referrals, machine retrievals, and outcomes as separate events.
Use a stable, non-branded prompt set grouped by intent, then report results by engine before calculating an overall score.
Count citation coverage against eligible prompt runs and define duplicate handling before collecting data.
Audit the source roles each engine favors. Improve owned pages where owned sources win, and earn legitimate independent evidence where editorial, video, or community sources dominate.
Maintain discovery content for mentions and citations while strengthening pricing, comparison, and decision pages for the visits that arrive later.
Judge AI referrals by qualified actions, pipeline, and revenue, while keeping unproven assisted effects in a separate evidence column.
Start with your highest-value prompt cluster and one engine. Freeze the prompt wording, capture the current answers and citations, map each cited source to its role, and connect every owned landing page to a measurable action. Change one content or source gap, repeat the same test, and let the movement in the correct signal determine the next change.
Your rankings can hold steady while your brand quietly falls off the buyer’s shortlist. A prospect may ask ChatGPT, Gemini, or Claude for options, encounter you in a comparison without visiting your site, see a social post, and convert long after the first interaction. Traffic and last-click conversions record only fragments of that journey.
You don’t need another all-purpose visibility score. You need a measurement system that separates business results, early intent, channel reach, AI perception, and volatility. That separation tells you whether to fix discoverability, positioning, conversion, or the metric itself.
Key takeaways
Keep business outcomes, validated proxy events, channel visibility, and AI perception in separate layers. They answer different questions.
Measure AI visibility as a current state, a change from the previous baseline, and a pattern of stability over time.
Use a fixed prompt library and consistent test conditions. Otherwise, changes in your test can masquerade as changes in brand perception.
Promote a micro-conversion into reporting or bidding only after it predicts a downstream outcome, occurs early enough to be useful, and remains dependable.
Treat every unusual metric pattern as a diagnosis to test, not an automatic instruction to publish more content or increase spend.
Build a layered scorecard instead of one blended score
A single score is attractive because it makes reporting look simple. It also hides the reason performance changed. An increase in AI mentions cannot compensate for declining qualified pipeline, just as revenue alone cannot tell you whether a recent visibility initiative is starting to work.
Build the dashboard in layers. Let each layer retain its own denominator, time horizon, and decision owner.
Measurement layer
What to track
Question it answers
Decision it supports
Business outcomes
Qualified opportunities, pipeline, revenue, or the final outcome your organization accepts
Did marketing contribute to valuable demand?
Budget allocation and commercial priorities
Validated leading indicators
Events shown to precede the business outcome, such as a qualified demo request or meaningful product evaluation
Are high-intent behaviors moving before revenue appears?
Campaign optimization and faster testing
Search and social discovery
Impressions, query coverage, clicks, referrals, and channel-specific engagement
Where can people encounter the brand?
Distribution, content coverage, and channel investment
AI perception
Mentions, recommendations, prominence, category associations, factual accuracy, and cited support
How do AI systems recall and represent the brand?
Entity clarity, positioning, documentation, and third-party evidence
Signal stability
Changes in inclusion, recommendation, position, and associations across comparable snapshots
Is visibility persistent or fragile?
Investigation, monitoring, and risk prioritization
The business-outcome layer remains the truth layer. The other layers shorten your feedback loop or explain how the outcome developed. Calling an AI mention, a scroll, or an impression a conversion erases that distinction and encourages the team to optimize activity instead of value.
Channel data is also becoming less isolated. Google has begun integrating social channel data into Search Console Insights. That can make discovery reporting more convenient, but placement in one interface doesn’t turn social exposure into search performance or revenue. Preserve the channel label and follow the signal downstream.
Make AI visibility a repeatable measurement
AI visibility deserves its own layer because buyers are using generative systems during vendor discovery. A Responsive survey found that 80% of tech buyers use generative AI to research vendors as often as traditional search. That figure describes one surveyed market rather than every buyer, but it is strong enough to make AI recommendations relevant to B2B measurement.
The difficult part is that an AI answer isn’t a fixed search result. Output can vary with the model, prompt, access mode, available context, underlying data, and model updates. A screenshot proves what appeared once. It does not establish durable visibility.
Freeze a core prompt library
Start with the decisions a buyer asks an AI system to help make. Keep a frozen core for period-over-period measurement and a separate exploratory set for new questions. Your core can cover:
Non-branded category discovery: which products address a defined problem or use case?
Shortlisting: which options fit a specified company type, constraint, or workflow?
Comparison: how do named alternatives differ on criteria buyers actually evaluate?
Risk and suitability: when is a product a poor fit, and what limitations should a buyer consider?
Implementation: which products integrate with the relevant ecosystem or operating environment?
Record the exact prompt, model, date, access mode, language, location when relevant, repeat count, and full response. Keep these conditions consistent across snapshots. If you revise a prompt, preserve it as a new series instead of splicing its results into the old one.
Run the same prompt more than once within each measurement window. Repeated runs help you distinguish answer variability from a broader shift. Keep the number of runs consistent so that a larger sample in one period does not create an artificial change.
Score representation, not just mentions
Define an eligible prompt before calculating any rate. A prompt is eligible when your offering could reasonably satisfy the stated need. Counting irrelevant prompts in the denominator suppresses the score and encourages category sprawl.
Mention rate: the share of eligible responses that name your brand.
Recommendation rate: the share that presents your brand as a suitable option rather than mentioning it incidentally.
Prominence rate: the share that places the brand in the opening recommendation set or another consistently defined prominent position.
Category-association rate: the share that connects the brand to the category, use case, audience, or capability you intentionally target.
Representation accuracy: the share of evaluated claims that match your current, verifiable product information.
Source-support rate: among answers that provide citations, the share that supports the brand description with an appropriate first-party or credible third-party page.
A commercial AI brand score may combine visibility and rank in one number. Keep the underlying components accessible. A brand can be mentioned more often while becoming less prominent, or remain prominent while being associated with the wrong use case. Those situations demand different fixes.
Separate state, drift, and stability
Your current score is the state. The change between comparable snapshots is drift. The persistence of the signal across several snapshots is stability. Report all three.
Express rate changes in percentage points so the size and direction of movement remain visible.
Track which brands entered or left the recommendation set, not merely the average number mentioned.
Log association gains and losses. A brand may remain visible while moving from a core category into an adjacent one.
Compare models separately before calculating any aggregate. Agreement across models is stronger evidence than a gain confined to one system.
Measure persistent inclusion by checking which core prompts continue to mention or recommend the brand in adjacent periods.
A September-to-October 2025 project-management snapshot recorded Atlassian gaining prominence while Slack declined. The same dataset showed category boundaries extending into operations, digital transformation, workflow orchestration, enterprise productivity, and IT consulting. This is one case, not a universal benchmark or proof of causation. It demonstrates why rank alone is insufficient: the conceptual neighborhood around a category can move along with the brands inside it.
When an association changes, audit the evidence available across your site, technical documentation, integration material, reputable directories, GitHub repositories where relevant, reviews, and community discussions. These environments can reinforce different parts of an entity’s identity. The goal is not to manufacture mentions. It is to make the same accurate category, audience, capabilities, and limitations legible wherever people genuinely evaluate the product.
Validate proxy metrics before algorithms optimize them
Long B2B sales cycles create an uncomfortable gap: the team needs feedback before enough opportunities or revenue mature. Proxy metrics can fill that gap, but only if they predict the result you care about. A frequent event isn’t automatically a useful signal.
Use four tests when deciding whether a candidate event belongs in your scorecard:
Correlation strength: people or accounts that complete the event should reach the downstream outcome more often than comparable ones that do not.
Timeliness: the event must occur early enough to change a live campaign, audience, message, or budget decision.
Actionability: your team must know which lever to adjust when the metric changes.
Stability: the relationship should persist across reporting periods and relevant audience segments rather than appearing in one temporary spike.
Validate the event in a defined sequence:
Name the downstream outcome precisely. Do not mix raw leads, accepted opportunities, and revenue in one target.
Identify candidate events that happen before that outcome and can be joined to the same person or account without breaking your consent and data-governance rules.
Compare downstream outcome rates for entities that completed each event with suitable entities that did not.
Check the lead time. A strongly related event that occurs immediately before the final outcome may explain performance but still arrive too late for optimization.
Repeat the comparison by period, channel, and meaningful audience segment. Promote the proxy only when its direction remains dependable.
Keep events in three operational tiers. Business outcomes belong in executive reporting. Validated proxies can support campaign learning and, when appropriate, bidding. Diagnostic engagement events such as time on site or scroll depth should remain investigative until you demonstrate a downstream relationship.
This matters when supplying early signals to Google or Meta optimization systems. Micro-conversions can help an algorithm learn when final-conversion volume is sparse, but the system will pursue the behavior you define. If scroll depth is cheap and loosely related to qualified demand, optimizing for it can produce more scrolling rather than more customers.
Context changes the quality of a proxy. A newsletter signup may indicate continuing interest, while an add-to-cart event can mislead when abandonment is common. Neither event should inherit value from its name. Let its observed relationship with your own accepted outcome determine how you use it.
Read cross-metric patterns before choosing a fix
The scorecard becomes useful when you read movement across layers. The combinations below are working diagnoses, not conclusions. Use the next check to confirm or reject each interpretation.
Observed pattern
Working diagnosis
What to check next
AI mentions fall while search visibility holds
Brand perception, model behavior, or category association may have shifted without a traditional ranking loss
Compare models, inspect lost prompts, review association changes, and verify that the test conditions stayed constant
AI mentions hold but recommendation rate falls
The brand remains known but appears less suitable or less prominent
Examine stated limitations, comparison criteria, audience fit, and the brands now recommended ahead of it
Search impressions fall while AI visibility holds
The problem may sit in traditional search demand, coverage, ranking, or technical visibility
Segment branded and non-branded queries, inspect affected pages, and keep the AI series separate
A proxy rises while qualified outcomes remain flat
The proxy may have weakened, the audience mix may have changed, or a later handoff may be failing
Recalculate the proxy-to-outcome relationship and trace the journey after the event
AI visibility rises while referral traffic stays flat
The gain may represent exposure rather than visits
Check recommendation quality, branded demand, assisted journeys, and downstream outcomes before declaring success or failure
Social discovery rises while search remains flat
Distribution may be broadening in one channel without changing search demand
Preserve channel attribution and test whether the added audience reaches a validated proxy or business outcome
Discovery improves across channels but pipeline does not
The constraint may be message fit, offer fit, conversion, qualification, or the sales handoff
Inspect landing behavior and stage-to-stage progression before buying more reach
At each reporting review, identify the largest meaningful movement, write down the most plausible explanations, and assign a check that can distinguish among them. Record the decision and its expected effect in the next comparable snapshot. That decision log prevents the team from retrofitting a success story to whichever metric happened to rise.
Start your next dashboard revision by adding the missing layer, not by adding more charts. If you already report revenue and search traffic, build a fixed AI prompt baseline. If you already monitor AI mentions, add representation accuracy and stability. If micro-conversions drive optimization, revalidate their relationship with qualified outcomes. The next useful metric is the one that resolves a real decision your current reporting leaves ambiguous.
