CrushPress.AI

Category: AI

  • Google Ads Developer AI Updates: A Practical Playbook

    Google Ads Developer AI Updates: A Practical Playbook

    You do not need another AI announcement in your backlog. You need to know whether Google’s direction changes what your advertising team should build, who should control it, and how much authority an AI agent should receive.

    The immediate answer is not to rebuild your Google Ads integration around agents. Treat the update as an architectural signal: prepare for AI systems to propose and invoke advertising actions, but keep permissions, validation, approvals, execution, and audit controls outside the model.

    The update is a learning channel, not an API release

    An engineer studies abstract signals from a studio beacon while a separate sealed production system remains unchanged on the workbench.

    Google has introduced Ads DevCast as a bi-weekly pilot hosted by Cory Liseno from its Advertising and Measurement Developer Relations team. Its technical scope includes Google Ads, Google Analytics, and Display & Video 360. Google is also inviting feedback while the pilot develops.

    That positioning matters. Ads Decoded, hosted by Ginny Marvin, addresses campaign strategy. Ads DevCast is intended for the people building, configuring, debugging, and governing the systems beneath that strategy. Subscribe the technical owner of your advertising stack, not only the person who manages campaigns.

    A new developer show does not, by itself, change an endpoint, schema, authentication flow, or deprecation date. Do not turn an episode into a production migration ticket merely because an idea sounds important. Use three separate lanes:

    • Discovery: Use Ads DevCast to notice technical themes, emerging capabilities, and the problems Google expects developers to encounter.
    • Verification: Confirm implementation details in the relevant official API documentation, release notes, schemas, and account controls before changing code.
    • Delivery: Create an engineering task only after you can name the affected platform, resource, operation, permission, test case, and rollback path.

    This distinction prevents two common errors. One is ignoring a directional signal until it becomes an urgent implementation problem. The other is treating a discussion of future architecture as though it were a released feature with stable production behavior.

    The agentic shift changes your control plane

    The first episode, titled “MCPs, Agents, and Ads. Oh My!”, presents an “agentic shift” in which AI agents become important users of advertising APIs. Treat that as Google’s direction of travel, not as evidence that every advertiser should give an agent unrestricted control of live campaigns.

    Model Context Protocol, or MCP, is relevant because it gives AI systems a common way to discover and invoke tools. A consistent tool interface can make an API easier for an agent to reach. It does not make the requested action correct, authorized, affordable, or reversible.

    The safest mental model is simple: the agent is a planner and operator working inside a control system. It is not the control system. A production workflow should separate intent from execution:

    1. Observe: Retrieve only the account and campaign data needed for the task.
    2. Propose: Produce a structured change showing the target resource, current value, proposed value, rationale, and expected scope.
    3. Validate: Check the proposal against the API schema, account state, internal policy, and allowed operations.
    4. Approve: Require the appropriate human or policy-based approval before any consequential write.
    5. Execute: Pass the approved action to deterministic code that calls the advertising API.
    6. Verify: Read the affected resource again, record the result, and surface any difference between the approved proposal and the final state.

    Put hard limits outside the prompt

    A prompt can tell an agent not to make risky changes. It should not be the only thing preventing them. The enforceable rules belong in the gateway between the agent and the ad platform.

    • Allowlist the accounts, resource types, fields, and operations the agent may access.
    • Use read-only access by default and grant write access per workflow rather than per agent.
    • Reject requests that omit the target account, current state, proposed state, or approval record.
    • Place budget, bid, scheduling, targeting, and deletion constraints in code or platform policy.
    • Use idempotency or equivalent duplicate protection where the operation supports it.
    • Log the request, tool call, actor, approval, API response, and resulting resource state.
    • Maintain a tested way to reverse mutable changes and a separate recovery procedure for actions that cannot be cleanly undone.

    This is a money-sensitive system. An agent with broad write access can alter live delivery before a person notices the mistake. For any action that can increase spend, narrow reach, pause revenue-producing activity, remove data, or change measurement, use a preview-and-approval flow until you have evidence that a more automated policy is safe for that exact operation.

    Turn each episode into an engineering decision

    A bi-weekly technical program can quickly become background noise unless someone owns the intake process. Give one person responsibility for converting each relevant item into a decision, including a deliberate decision to take no action.

    1. Capture the claim precisely. Write down the named product, capability, resource, or workflow. Avoid tickets such as “investigate AI for ads” because they have no testable boundary.
    2. Classify its status. Mark it as a concept, directional signal, pilot, documented capability, released change, or deprecation. Do not let enthusiasm silently upgrade its maturity.
    3. Map the affected surface. Identify whether it touches Google Ads, Google Analytics, Display & Video 360, or more than one system. Then name the relevant integration, credential, data flow, and owner.
    4. Verify implementation facts. Check the authoritative documentation for availability, supported operations, permissions, quotas, version requirements, and known limitations.
    5. Record the decision. Choose watch, prototype, adopt, migrate, or reject. Include the evidence needed to revisit that choice.

    Your decision record does not need to be elaborate. It should include the topic, status, affected system, documentation link, owner, next review trigger, test environment, approval requirement, and rollback method. That is enough to distinguish a useful technical signal from an unverified idea circulating in team chat.

    Use a prototype when the value is plausible but the operational risk is unclear. Start with a read-only workflow that answers one bounded question, then let the agent draft a change without executing it. Compare its proposal with the decision a qualified operator would make. Only after that should you test an approved write in a controlled account or environment.

    Because Ads DevCast is a pilot seeking community input, document where explanations leave an implementation gap. Useful feedback is specific: name the platform, operation, missing detail, and decision you could not safely make. That gives Google a clearer request than a general demand for more examples.

    Your ownership model must evolve with the integration

    An isometric AI advertising workflow routes action tokens through access controls, validation, human review, staging, and an audit vault while separate teams supervise their areas.

    Google is broadening the frame from a specialist Ads Developer Community toward a wider Ads Technical Community. That makes room for marketers to perform more technical work without waiting for a full development cycle. It does not erase the need for engineering ownership; it changes where the handoffs occur.

    Before connecting an agent to advertising tools, assign these responsibilities by name:

    • Business owner: Defines the campaign objective and decides which tradeoffs are acceptable.
    • Platform owner: Controls credentials, permissions, API configuration, and production access.
    • Workflow owner: Defines the agent’s tools, inputs, outputs, validation rules, and failure behavior.
    • Approver: Reviews consequential changes and has enough context to reject a technically valid but commercially poor action.
    • Incident owner: Can stop execution, assess affected resources, restore safe state, and preserve the audit trail.

    Do not collapse all five roles into “the AI team.” The business owner knows what should happen. The platform owner knows what can happen. The workflow owner controls how a request becomes an API call. The approver evaluates the actual change. The incident owner handles the moment when the system behaves differently from the plan.

    This division also makes low-code and agent-assisted work more practical. A marketer can describe or initiate a task without receiving unrestricted platform access. Engineering can provide constrained tools and reusable policies instead of implementing every request from scratch. The speed comes from a safer interface between roles, not from removing the roles.

    Key takeaways for your next working session

    • Use Ads DevCast as a technical discovery channel; verify every implementation detail in authoritative product documentation.
    • Treat Google’s agentic direction as a reason to prepare your architecture, not as permission to automate every campaign action.
    • Keep the agent focused on observation and structured proposals before granting narrowly scoped write capability.
    • Enforce permissions, spend constraints, approvals, logging, and recovery outside the model and its prompt.
    • Assign business, platform, workflow, approval, and incident ownership before connecting an agent to a live advertising account.
    • Convert each relevant update into a recorded decision: watch, prototype, adopt, migrate, or reject.

    Start with one existing Google Ads workflow that consumes too much operator time but has a clear input and output. Draw the six stages from observation through verification. Mark every place where a bad decision could affect spend, delivery, measurement, or data. Those marks define the controls your agent needs before it gets write access.

    Then build the smallest read-only version and require a structured proposal. That gives you a concrete way to evaluate Google’s agentic direction without betting a live account on an immature design.

    References

    crushpress.ai community screenshot
  • Google’s Universal Commerce Protocol: A Retailer Playbook

    Google’s Universal Commerce Protocol: A Retailer Playbook

    If you run ecommerce SEO, product feeds, or shopping infrastructure, your next visibility problem may not begin on a search results page. It may begin when an AI shopping agent tries to identify the right variant, confirm that it is available, calculate the correct price, and place it in a working basket.

    Google’s Universal Commerce Protocol, or UCP, is intended to connect those steps. Your practical task is to make product and customer data usable across discovery, selection, and checkout without assuming that protocol adoption will automatically produce rankings, recommendations, or sales.

    UCP moves product visibility closer to the transaction

    Traditional search optimization prepares a page for a person to discover and visit. Agentic commerce adds another route: software may evaluate products, assemble a purchase, and act for the shopper. UCP is an open, modular standard for connecting retailers with AI-driven shopping experiences.

    That does not make product pages irrelevant. It changes where accuracy has to survive. A persuasive description cannot compensate for an unavailable variant. Valid page markup cannot repair a cart that calculates the wrong price. A feed can expose a product, but the transaction can still fail if customer benefits disappear after identity linking.

    This gives you four connected layers to manage:

    • Page content and structured data explain the product in a crawlable, understandable form.
    • Catalog data supplies current commercial facts such as price, inventory, and available variants.
    • Cart logic turns selected items into a valid basket.
    • Identity and account logic determine whether the shopper receives eligible benefits.

    Keep these layers aligned, but do not treat them as interchangeable. UCP is not merely another name for JSON-LD, a product feed, or an ad format. It reaches into live commerce functions that page-level optimization alone cannot perform.

    Google has said it plans to use UCP capabilities in AI-enhanced experiences across Search and the Gemini app. That establishes a direction, not a promise that every retailer, market, capability, or product will receive the same access or exposure. Build readiness around documented availability and your own eligibility rather than an assumed rollout.

    Map each UCP capability to a real retail responsibility

    The useful way to evaluate UCP is capability by capability. Each one touches a different system, failure mode, and internal owner.

    CapabilityWhat it enablesWhat you should verifyLikely owner
    CatalogAccess to current product information, including pricing, inventory, and variantsStable identifiers, variant mapping, update freshness, and agreement between catalog, product page, and checkoutMerchandising, feed operations, or commerce platform team
    CartMultiple products from one retailer can be assembled into one basketAdd, update, remove, reprice, and out-of-stock behavior across a multi-item orderEcommerce engineering
    Identity linkingEligible benefits such as member pricing and free shipping can continue across connected experiencesAuthentication, consent, entitlement rules, session handling, and safe failure behaviorIdentity, security, loyalty, and legal or privacy teams
    Modular adoptionA retailer or platform can adopt selected capabilities instead of implementing everything at onceA rollout sequence tied to system readiness and a clear dependency mapCommerce product owner or program lead

    The capability names do not answer every implementation question. For example, knowing that an agent can create a cart does not by itself define how your taxes, promotions, substitutions, shipping restrictions, or returns work. Treat those as test cases that need authoritative documentation and validation in your own stack. Do not invent behavior from the protocol’s high-level description.

    Modularity is especially important for planning. You do not need to frame UCP as an all-or-nothing rebuild. If your identity system is not ready, that does not erase the value of repairing catalog inconsistencies. If your catalog cannot reliably distinguish variants, however, adding an agent-facing cart simply moves bad data closer to checkout.

    Audit product data as if it were the storefront

    An unbranded jacket, variant swatches, packaging, inventory objects, and a magnifying lens are arranged for a detailed product data audit.

    An agent cannot walk a virtual aisle and infer that a stale price is probably wrong. It receives representations of your inventory and has to make decisions from them. Because the catalog capability is designed to expose real-time pricing, inventory, and variant information, conflicting product facts become a commercial problem, not merely a feed-cleanup task.

    Start with one product family that has meaningful variation. A product with size, color, configuration, or member pricing will reveal more than a simple item with one price and one stock state. Trace it through every system an agent-assisted purchase could touch.

    1. Resolve the identity chain. Confirm that the parent product, each purchasable variant, the catalog record, the product page, and the cart line resolve to the intended item. A parent identifier should not silently stand in for a specific variant at purchase time.
    2. Name the source of truth for each commercial fact. Decide which system owns price, sale price, inventory, variant attributes, and account benefits. If two systems can overwrite the same fact, document precedence and failure handling.
    3. Compare anonymous and authenticated states. Check whether public pricing, member pricing, shipping benefits, and eligibility rules remain distinguishable. The agent should not present a conditional benefit as universal.
    4. Test change propagation. Change a price or inventory state in the owning system and observe every downstream representation. Record your actual delay and failure points rather than relying on the intended architecture.
    5. Inspect contradictions. Compare the catalog, rendered product page, structured data, basket, and logged-in experience. Any disagreement can lead to a poor recommendation, a rejected add-to-cart action, or an unpleasant price change at checkout.
    6. Log failed and stale updates. A synchronization process that usually works is not enough. Your team needs a way to identify which products failed, when the last successful update occurred, and which downstream surfaces may still carry old information.

    This is also where SEO, GEO, and feed teams should coordinate. Keep descriptive content and structured data consistent with commercial systems, but do not add unsupported claims to markup merely to make the product look more complete to an AI system. The safest machine-readable answer is the same answer the shopper will receive in the cart.

    Do not call the audit complete because a sample record validates syntactically. A valid record can still identify the wrong variant, carry an old price, or point to inventory that cannot be purchased. Validation checks form; transaction tests check truth.

    Roll out the smallest capability you can verify end to end

    A coffee maker follows one illuminated path through catalog, inventory, basket, payment, and delivery modules while unused modules remain dark.

    Catalog readiness is usually the sensible first workstream because cart and identity experiences depend on accurate merchandise data. That is a sequencing recommendation, not a protocol requirement. Your architecture may justify a different order, but every pilot should have one defined capability, one accountable owner, and an observable pass or fail condition.

    1. Choose a bounded product set. Select products that expose the problems you need to solve, including variants or conditional benefits, while keeping the pilot small enough to inspect manually.
    2. Capture a baseline. Record current catalog mismatches, failed add-to-cart actions, unavailable variants presented as purchasable, and benefit-entitlement failures. Without a baseline, protocol activity can look like progress while customer-facing accuracy remains unchanged.
    3. Define acceptance tests before integration. Write expected results for price changes, inventory changes, variant selection, multi-item baskets, account linking, and entitlement loss. Include negative cases, not just a successful purchase.
    4. Test the cart as a changing object. The new cart capability is intended to let agents place multiple products from one retailer into a single basket. Verify what happens when quantity changes, one line becomes unavailable, a promotion expires, or the shopper switches variants.
    5. Isolate identity testing. Identity linking can preserve member pricing and free shipping, but it also touches account access and personal data. Use controlled test accounts and obtain security, privacy, and legal approval before exposing real customer identities. The specific downside of rushing this step is not just a broken discount; it can be unauthorized account access or inappropriate data sharing.
    6. Monitor outcomes by failure stage. Separate catalog retrieval, variant resolution, cart creation, cart mutation, authentication, entitlement, and checkout failures. A single conversion total will not tell you which capability needs repair.

    Your ownership model matters as much as the integration. Feed operations can correct a variant mapping but should not define authentication policy. SEO can identify contradictions visible to search systems but should not own checkout integrity. Ecommerce engineering can make a cart function without knowing whether member benefits are represented correctly. Put these teams behind one shared test plan rather than handing UCP to whichever team first notices it.

    Google has also indicated that it plans to simplify UCP onboarding through Merchant Center. Use that as a reason to prepare your data and test cases, not as a reason to assume that implementation is already automatic. When onboarding becomes available to you, confirm supported capabilities, required fields, market coverage, permissions, and reporting from the documentation presented in your account.

    Most importantly, do not report UCP adoption as an SEO win by itself. There is no basis here for calling it a guaranteed ranking factor or recommendation boost. Measure what you can actually observe: eligibility, accurate product representation, successful basket creation, preserved benefits, completed purchases, and the failure rate at each handoff.

    Key takeaways

    • UCP connects product discovery with live commerce functions; it is broader than page markup, feeds, or advertising alone.
    • Catalog accuracy is foundational because price, inventory, and variant errors can follow an agent directly into the cart.
    • Cart, catalog, and identity linking should be treated as separate capabilities with separate owners and tests.
    • Modular adoption lets you start with a bounded capability instead of waiting for a complete commerce-stack rebuild.
    • Identity linking requires controlled testing and security, privacy, and legal review before real customer accounts are involved.
    • Protocol adoption does not establish a ranking or recommendation benefit. Evaluate transactional accuracy and measurable outcomes.

    Your best next step is concrete: take one high-value product family with variants, compare its catalog record, product page, structured data, cart, and logged-in benefits, then document every contradiction. That exercise will tell you whether your first UCP project is an integration project or, more likely, a product-data repair project that needs to happen before integration can deliver anything useful.

    References

  • Google’s AI Mode: Revolutionizing Ad Monetization

    Google’s AI Mode: Revolutionizing Ad Monetization

    As I explore the ever-evolving landscape of Google’s AI Mode, it’s fascinating to witness how ad formats, reporting, and control are taking shape. Google seems to have a master plan in place that competitors just can’t keep up with.

    I find myself intrigued by Google’s entry into this next phase of conversational search. It’s not just about user numbers but who can effectively monetize them. Google’s mature ad systems and extensive advertiser base offer a significant edge.

    The initial panic surrounding Google’s position is over. Google’s long-standing advantages and huge investments have leveled the playing field with ChatGPT in LLM search.

    Back in December 2025, when Google declared code red, it became clear that they were serious. Apple’s decision to partner with Google for its AI needs is indeed telling.

    Initially, it seemed plausible that Google would struggle against ChatGPT, but the market has since adjusted its views. The company’s valuation reflects renewed confidence, rivaling even Apple at a substantial $3.6 trillion.

    As I dive deeper into how monetization will shape this race, I’m struck by how Google’s recent advances have significantly boosted its valuation.

    ```json { "alt": "Alphabet Inc. (GOOG) stock performance chart over five years, showing growth of 190.88%.", "caption": "Alphabet Inc.'s (GOOG) stock chart reveals a significant upward trend over the past five years, with a marked growth of 190.88%.", "description": "This image displays a five-year stock performance chart for Alphabet Inc. (GOOG), highlighting a substantial gain of 190.88%. The chart features key stock prices at the market close on February 13, with a closing price of 306.02, reflecting a decrease of 1.08%. The after-hours price is 305.88, down by 0.05%. The chart tracks the stock's fluctuations, offering insights into significant trends and key events impacting performance in the NasdaqGS market." } ```

    It’s clear that the visibility of financial projections plays a massive role in how the company is perceived financially. Google’s approach to shifts in user behavior is crucial in maintaining its robust business model.

    From my perspective, much of your digital advertising budget likely goes to Google. Its prominence demands attention, not just in search but also in emerging AI platforms like ChatGPT and Claude.

    The competition in LLM conversations is intriguing. Google and ChatGPT are vying for different monetization models, a fascinating case study of differing strategies.

    For those of us in advertising, it’s essential to monitor developments like ad formats, rollout pace, and public reception to ads within these platforms.

    OpenAI’s current monetization model is intriguing but still nascent, reliant on a small group of major advertisers. We’ll see how they expand and fine-tune this model over time.

    ```json { "alt": "Weather forecast indicating rain in Sarasota on February 22, 2026, with a summary of rain chances over the next 14 days.", "caption": "Stay prepared, Sarasota! Rain is likely on February 22, with varying chances throughout the next two weeks. Know what's coming your way!", "description": "This image shows a weather forecast for Sarasota, highlighting expected rain on February 22, 2026, with a 40% to 70% chance of showers. The forecast includes a detailed 14-day rain outlook with additional chances of rain later in the week and into March. A summary table provides daily rain chances and expected conditions. A side panel lists various weather services providing localized forecasts." } ```

    Outsourcing inventory to programmatic partners is a smart move for OpenAI but highlights their early stage in building an ads business.

    For Google advertisers, the shift to AI Mode need not be alarming. I’m watching for the ways these LLM sessions are shaping user experiences and ad placements.

    One thing is for sure; the enhancements in AI Mode continue, promising more seamless and user-friendly interactions. The potential for ads remains, though their form is still evolving.

    Monitoring key areas like the extent of monetization, advertiser control, and campaign types becomes more important as we navigate this new landscape.

    Ultimately, the future of advertising in AI-driven search is one of adaptability and strategic planning, aligning closely with user and advertiser behaviors in this exciting yet challenging era.

    Inspired by this post on Search Engine Land.

    crushpress.ai community screenshot
  • AI Search Traffic Surges 180% in 2025: Key Trends and Insights

    AI Search Traffic Surges 180% in 2025: Key Trends and Insights

    As I look back on 2025, it’s astonishing to see the AI search traffic growth leap by an impressive 180% year-over-year. I’m diving into the data to better understand how this impacts our visibility strategies. We’ll explore insights on ChatGPT, Gemini, Perplexity, and Claude usage trends in this review.

    With AI technologies rapidly advancing, I’ve noticed how they continue to reshape how we think about search and brand visibility. The increased use of AI-powered tools signifies a pivotal shift in the way we approach digital marketing strategies.

    In 2025, ChatGPT saw a remarkable surge in use, closely followed by interest in platforms like Gemini and Claude. This data is crucial as we plan for future visibility tactics, ensuring that our brand remains competitive in an ever-evolving digital landscape.

    How does this data affect your brand’s approach? I believe understanding and leveraging these trends will be key to optimizing AI-driven search capabilities and visibility while crafting more personalized and effective content strategies.

    Inspired by this post on genmark.ai Blog.

    crushpress.ai community screenshot
  • Unlock Content Creation with Profound: Harness Prompt Volumes

    Unlock Content Creation with Profound: Harness Prompt Volumes

    I’ve found an incredible new way to streamline content creation, competitive analysis, reporting, and monitoring with the latest Profound Agents feature. We can now effortlessly integrate prompt volume data directly into any Profound Agent, bringing together all our workflows into a single platform. This innovation is perfect for marketers looking to enhance efficiency.

    Inspired by this post on Try Profound Blog.

    crushpress.ai community screenshot
  • A Marketer’s Playbook for Ads in AI-Assisted Discovery

    A Marketer’s Playbook for Ads in AI-Assisted Discovery

    Your next paid discovery brief may arrive before the format has a stable name. The ad might represent an entire store instead of a single product, while an AI assistant might capture useful engagement before the buyer ever visits your site. A campaign structure built around a keyword, a product, and a click will not give you enough control.

    You do not need to predict which interface will win. You need a preparation model that works across store-level placements, conversational environments, and whatever hybrid appears between them. That means strengthening the advertised object, the evidence around it, the routes a buyer can take, and the measurement required before you commit budget.

    The advertised object is getting larger

    Traditional shopping campaigns make the individual product the center of gravity. Google is testing Sponsored Shops, a Shopping block that groups several products from one retailer with the store name, ratings, and broader brand presence. The impression can therefore introduce an assortment and a merchant, not merely an item.

    Conversational discovery creates a different expansion. OpenAI has begun testing an Ads Manager dashboard with selected partners as it develops advertising around ChatGPT. The exact inventory, interaction model, and optimization system remain early. You should treat them as provisional rather than assume conversational ads will inherit the rules of paid search.

    The practical lesson is that the thing you advertise can sit at several levels. It might be a product, a coherent assortment, a store, or a solution to the need expressed in a conversation. Each level requires different proof and a different continuation after the impression.

    Add the following fields to your campaign planning before a new platform makes them mandatory:

    • User need: the problem, task, or buying situation that triggered discovery.
    • Advertised object: the product, collection, store, or solution path the unit represents.
    • Evidence: the ratings, product details, range, brand facts, and on-page claims that support the promise.
    • Possible interactions: product selection, brand selection, continued conversation, or a direct visit.
    • Continuation: the exact page or in-platform step that follows each interaction.
    • Business event: the observable action that would make the placement valuable.

    This prevents a common category error: treating a larger discovery unit as if it were merely a wider text ad. More visible products do not automatically create a coherent reason to choose the store. A conversational placement does not automatically produce a qualified visit. The advertised object must make sense as a whole.

    Build a discovery asset stack before you buy media

    A modular stack of storefront, product, evidence, inventory, and data elements connects to three abstract discovery interfaces.

    A store-level placement exposes the quality of the catalog as a portfolio. Sponsored Shops could favor merchants with stronger product feeds, useful assortment depth, and credible seller ratings, because several products and the retailer identity appear within the same unit. A weak item is no longer isolated; it can make the entire selection feel less relevant.

    Do not answer that pressure by putting more products into every group. Build an asset stack in which every layer has a defined job:

    1. Catalog facts establish what each product is, what it costs, whether it is available, and how it differs from nearby options.
    2. Assortment logic explains why a set of products belongs together for a particular need. Shared inventory is not enough; the group needs a shopper-facing reason to exist.
    3. Brand evidence gives the buyer a reason to trust the store behind the assortment. Ratings and consistent brand identity matter more when the merchant is part of the advertised object.
    4. Destination continuity carries the same promise from the ad into the next page. The buyer should not have to reconstruct the category, filter, or use case after clicking.
    5. Machine-readable agreement keeps feeds, visible page content, and structured data aligned. JSON-LD should repeat defensible facts shown to the user, not introduce a cleaner but contradictory version of the offer.

    Audit this stack by discovery theme rather than by campaign name. Write the buyer’s need in plain language, select the products that genuinely address it, and inspect every item in that set. Mark missing details, inconsistent naming, stale availability, weak images, unexplained variations, and claims that do not match the destination. Then decide whether the set deserves to be presented as a store-level recommendation.

    Keep product-level optimization intact while you do this. A broad assortment should not bury the strongest item or force unrelated products into the same story. You are adding a portfolio layer above the product layer, not replacing product relevance with brand reach.

    Give every interaction a deliberate next step

    A multi-element discovery unit creates more than one possible click. With Sponsored Shops, the split between clicks on the brand and clicks on individual products is an open measurement and usability question. If you only plan the final conversion page, you will miss the intent expressed by the element the buyer selected.

    Design a continuation for each route that the format exposes:

    • Store or brand interaction: use a focused storefront that confirms the range, positioning, and evidence shown in the unit. Avoid a generic homepage unless it already performs that job.
    • Collection interaction: preserve the discovery theme, relevant filters, and visible product set. Do not make the buyer rebuild the selection from a broad category page.
    • Product interaction: land on the exact item with its important facts, proof, availability, and next action easy to find.
    • In-assistant interaction: identify what the platform can report when the user continues the conversation without visiting your site. Treat unreported engagement as unknown, not as a click or a conversion.

    Put this destination map in the campaign brief before creative production. For every clickable element, record the likely intent, destination, page promise, and success event. If the platform allows distinct tracking parameters for different elements, use them. If it does not, record that limitation before deciding how much you are willing to spend.

    The first visible part of each destination should close the loop opened by the ad. A store-level promise about range should reveal that range. A product promise should show the exact product. A solution-oriented message should answer the need before introducing unrelated navigation. That continuity is more useful than repeating the ad headline word for word.

    Keep paid visibility separate from organic AI visibility in your reporting. Buying placement does not make an unclear page easier for an answer engine to understand elsewhere. Your AEO and GEO work still needs clear naming, consistent facts, direct answers, accessible evidence, and structured data that agrees with the visible page. Paid discovery adds distribution and control; it does not repair weak information architecture.

    Make measurement and budget pass the same gate

    A glowing interaction moves through a branching journey toward a product shelf, consultation doorway, or parcel while paired measurement and budget tokens pass through one gate.

    Use a measurement ladder, not a click counter

    Early ChatGPT advertisers have reportedly received weekly CSV reports containing impressions and clicks, while initial click-through rates have trailed Google Search. Delivery and click data can confirm that an ad ran. They cannot, on their own, tell you whether conversational discovery created valuable demand.

    Measure emerging discovery formats as a ladder:

    • Delivery: impressions, placement, advertised object, unit variant, and any available context about where the ad appeared.
    • Interaction: clicks by element, product selections, brand selections, or reported continuation inside the interface.
    • Progression: meaningful visits to product or collection pages, deeper product exploration, cart activity, lead starts, or another relevant journey event.
    • Outcome: completed purchases, qualified leads, revenue, or the business result attached to the campaign.
    • Incremental value: evidence that the new channel added outcomes rather than taking credit for demand another channel had already created.

    Mark unavailable fields as unavailable. Do not enter zero, because zero means the platform measured the event and found none. Missing element-level interaction data is itself a decision signal: it limits what you can learn about creative, assortment, and destination performance.

    Your tracking taxonomy should identify the platform, placement, advertised object, unit variant, and destination wherever the platform exposes those controls. Keep those dimensions separate. Otherwise, a store click and a product click can collapse into the same campaign total even though they represent different user decisions.

    Write the test decision before launch. State the hypothesis, the variable being changed, the primary business outcome, the supporting engagement signals, the acceptable downside, and the condition that will stop or expand the test. A low click-through rate is not automatically failure for an upper-funnel discovery unit, but it cannot be excused by vague claims about awareness. The downstream evidence must carry the argument.

    Set a budget gate that reflects platform maturity

    Some early ChatGPT advertisers have reportedly been asked for a minimum commitment of $200,000. That creates material financial exposure while reporting and optimization capabilities are still developing. Early access is not valuable merely because access is scarce.

    Before accepting a pilot, require clear answers to these questions:

    • Where can the ad appear, and how is sponsorship disclosed to the user?
    • Which audiences, contexts, placements, products, and destinations can you include or exclude?
    • Which delivery, interaction, conversion, and cost fields can you export, and at what reporting cadence?
    • Can you distinguish a brand interaction from a product interaction?
    • How will conversion measurement work when part of the journey remains inside the assistant?
    • Which campaign changes can you make during the pilot, and what are the stop conditions?

    Ring-fence money you can genuinely treat as experimental. Do not pull budget from a proven acquisition channel simply to claim first-mover status. If the minimum commitment is too large to absorb as a learning cost, or the reporting cannot connect delivery to business outcomes, observing the format is the disciplined choice.

    Move from observation to a pilot when destinations are traceable, controls are understandable, disclosures are clear, and the downside fits the approved test budget. Move from pilot to scale only when the outcome is repeatable and the reporting explains why it happened. Impressions and novelty are not scale criteria.

    Key takeaways for your next planning cycle

    • Plan around the advertised object, which may be a product, assortment, store, or solution path.
    • Treat catalog quality, assortment logic, brand evidence, landing pages, and structured data as one discovery asset stack.
    • Map separate continuations for brand, collection, product, and in-assistant interactions.
    • Measure delivery, interaction, journey progression, business outcomes, and incremental value as distinct layers.
    • Do not fund a large early pilot without exportable reporting, usable controls, explicit stop conditions, and a tolerable downside.

    Your next move is to choose a commercially important discovery theme and complete the advertised-object and destination map for it. Audit the supporting catalog, page evidence, and machine-readable facts before a platform representative puts a media proposal in front of you.

    When access becomes available, ask the platform to map every promised metric and control to that plan. If the gaps prevent a business decision, keep observing. If the path is traceable and the risk is bounded, run a focused pilot with written stop conditions. Emerging discovery inventory should earn its budget on evidence, just like any established channel.

    References

  • How LinkedIn’s LLM-Powered Feed Ranks Your Content

    How LinkedIn’s LLM-Powered Feed Ranks Your Content

    If your LinkedIn reach feels erratic, stop treating the feed like one global leaderboard. The platform is trying to predict relevance for each person, so two professionals with similar networks can still receive different candidates in a different order.

    The useful question isn’t, “How do I please the algorithm?” It is, “Can the system understand who this is for, and will the right readers behave as though it was worth their time?” LinkedIn’s new architecture gives you a practical way to improve both sides of that equation without pretending there is a secret score you can reverse-engineer.

    LinkedIn now makes two separate feed decisions

    Abstract content tiles pass through a broad selection gateway and then a second prism that orders different feeds for three viewers.

    Feed visibility begins with two distinct jobs: retrieval and ranking. Retrieval decides which posts could appear. Ranking decides which of those candidates should appear first. A post that fails the first decision never reaches the second, while a retrieved post can still lose its position to something that better matches the viewer’s current interests.

    Retrieval matches meaning, not just identical wording

    LinkedIn has consolidated previously separate discovery routes into a unified retrieval model. Large language models create embeddings: numerical representations that capture the meaning and context of a post. Those representations can be compared with a member’s professional interests even when the wording isn’t identical.

    Someone engaging with small modular reactor content, for example, may also receive material about renewable energy or a related professional field that uses different terminology. This semantic matching across related concepts matters more than repeating one phrase in every paragraph.

    The GPU-backed system processes millions of posts, can refresh content embeddings within minutes, and can retrieve candidates in less than 50 milliseconds. That speed means a fresh post can become semantically retrievable quickly. It does not guarantee that the post will be selected, ranked highly, or distributed widely.

    Ranking uses a sequence of viewer behavior

    After retrieval, a transformer-based sequential model orders the candidates. It doesn’t evaluate each post in isolation. It examines patterns in a member’s previous behavior, including likes, comments, and time spent viewing content, so the feed can adapt as professional interests change.

    This is an important limit on algorithm advice. A post does not have one universal rank. Its position depends partly on the person receiving it and the sequence of behavior that preceded that feed request. Strong results with one audience segment do not prove that the same post will rank the same way for everyone else.

    LLM-powered also doesn’t mean a chatbot is reading your prose like an editor and awarding points for style. One model represents meaning for retrieval; another uses interaction history to rank candidates. Human-readable quality still matters, but it matters because clear, useful content is easier to match and more likely to hold the right person’s attention.

    Make each post semantically legible

    A blank content card emits a focused constellation of topic symbols that connects with a matching group of professional readers.

    A vague post forces both the model and the reader to guess. A semantically legible post names the professional context, the problem, the affected audience, and the relationship between its main ideas. You can create that clarity without turning the copy into a keyword list.

    1. Write a private audience sentence before drafting: “This is for [role] deciding [specific decision].” If you can’t complete it cleanly, the topic is still too broad.
    2. Name the subject early. Don’t spend the opening on a generic tease that could introduce leadership, software, hiring, finance, or any other field.
    3. Explain the mechanism. State why the change happens, what it affects, or which constraint creates the problem. Adjectives such as “transformative” and “important” don’t supply that context.
    4. Connect the core topic to one relevant adjacent concept. Make the relationship explicit instead of dropping related terms into the copy without explanation.
    5. Show expertise through a process, tradeoff, decision rule, or concrete distinction. Claiming expertise is weaker than making knowledgeable reasoning visible.
    6. End with a question only when the answer can deepen the professional discussion. Ask about a decision, constraint, or experience, not whether readers agree.

    Compare “Big changes are coming. Thoughts?” with this structure: “For [role] deciding [decision], [named development] changes [specific constraint] because [mechanism].” The second version tells the retrieval system what the content concerns and tells the reader whether it deserves attention.

    Semantic retrieval is not permission to stuff a post with synonyms. Use the standard term your audience recognizes, explain it in plain language where necessary, and introduce adjacent terminology only when the relationship adds meaning. A keyword dump can mention everything while communicating almost nothing.

    A coherent series can help you explore a semantic neighborhood: the primary problem, its causes, its operational consequences, and the decisions around it. That does not prove LinkedIn grants account-level authority merely for repeating a topic. It does give each installment a clear chance to match similar professional interests, and it gives you a cleaner way to learn which angle resonates.

    Your network size is not the entire distribution story. Posts that demonstrate expertise and contribute to relevant professional conversations can travel beyond an author’s established connections. The practical move is not to chase every trending subject. It is to contribute when you have a specific connection between the timely topic and the work your intended audience actually does.

    Earn ranking signals without manufacturing them

    Because ranking considers likes, comments, and viewing time, it is tempting to treat every interaction as a lever. Resist that simplification. LinkedIn has not supplied a usable formula that tells you how much each action is worth in every context, and a pause on a post does not necessarily mean approval.

    Design for a meaningful reading experience instead. Give the opening enough information to qualify the audience. Build the body in a logical sequence. Make the promised point before asking for a response. If the subject needs depth, use depth; making a post artificially long in pursuit of viewing time only gives readers more opportunities to leave.

    • Use an opening that identifies the professional issue instead of withholding it behind suspense.
    • Break a complex explanation into distinct decisions, causes, or steps so the reader can follow the reasoning.
    • Ask for a response that requires professional judgment, such as which constraint changes the decision.
    • Reply manually and specifically when someone contributes. Continue the subject they raised instead of posting a generic thank-you.
    • Keep the text and any accompanying media on the same subject. An unrelated video may attract attention while weakening the content’s meaning.
    • Remove prompts whose only purpose is to inflate activity, including requests for a one-word comment with no substantive reason to answer.

    Automated comments and engagement pods are not clever shortcuts. LinkedIn has identified them as policy violations that create artificial discussion. The platform is also deprioritizing engagement bait, irrelevant text-and-video pairings, and generic recycled thought leadership.

    Don’t stretch that policy into a claim that every AI-assisted draft is automatically suppressed. The documented targets are automated engagement and low-value publishing patterns. Judge any drafting tool by the resulting content: Is the reasoning specific? Is the point accurate? Does the copy express a real professional distinction? Would the post still be worth reading if no engagement counter were visible?

    Test audience-topic fit instead of algorithm folklore

    A personalized feed makes casual testing unreliable. When one post performs better than another, the difference could involve the topic, the opening, the audience that received it, those viewers’ recent behavior, or the quality of the discussion. Changing several elements at once leaves you with a result but no useful explanation.

    1. Choose one business-relevant question that a recognizable professional audience needs to answer.
    2. Map the question into a core angle and adjacent angles, such as the cause, implementation constraint, common misreading, and decision tradeoff.
    3. Publish a coherent sequence in which every post stands on its own and names its subject clearly.
    4. Change one structural variable when you want to learn from a comparison: the opening, explanatory depth, example type, or closing question.
    5. Record more than reach. Note whether the people responding appear connected to the intended professional context and whether their comments engage with the actual issue.
    6. Use those observations to choose the next adjacent angle. Don’t turn one strong or weak result into a universal rule about length, timing, hashtags, or a supposed favorite interaction.

    Keep a simple brief beside each draft with these fields: intended reader, decision or problem, core concept, adjacent concept, mechanism or tradeoff, and response prompt. After publication, add what the discussion revealed. This turns a feed result into editorial information you can use rather than a number you can only admire or resent.

    Your own feed is also personalized evidence, not a neutral sample of LinkedIn as a whole. If you use it for topic research, remember that your likes, comments, and viewing behavior help shape what you see next. New members can make that preference-building more deliberate by choosing topics through the Interest Picker during signup. That helps customize the feed from the beginning, but it still does not reveal what every other audience sees.

    Key takeaways

    • Retrieval decides whether a post belongs in the candidate set; ranking decides where that candidate appears for a particular member.
    • Semantic embeddings make clear meaning and related concepts more important than exact-phrase repetition.
    • Ranking uses sequences of behavior, including likes, comments, and viewing time, but there is no dependable public formula for turning those actions into a universal score.
    • Expertise becomes visible through mechanisms, tradeoffs, processes, and useful distinctions, not through generic claims of authority.
    • Automated engagement, pods, bait, mismatched media, and recycled thought leadership create policy or quality risks instead of durable distribution.
    • The cleanest test is audience-topic fit: keep the subject coherent, change one structural variable at a time, and inspect who responds and what they discuss.

    Before your next LinkedIn post, write the private audience-and-decision sentence, rewrite the opening so the subject is unmistakable, and remove any question that can be answered without thought. Then use the quality of the resulting discussion to select the next relevant angle. That is a better compounding system than chasing a secret ranking trick.

    References

  • Google AI Advertising Strategy: What Marketers Should Do Now

    Google AI Advertising Strategy: What Marketers Should Do Now

    If you are waiting for a Gemini campaign type before changing your Google advertising plan, you are waiting for the least important part. Google is already learning how ads behave in AI-generated experiences, while its campaign systems are becoming more willing to assemble, resize and select creative for you.

    The useful response is not to move budget into an unannounced product. It is to make your offers easier to match to conversational intent, clean up the assets Google can reuse and put measurement guardrails around automation. Those changes improve campaigns you can run now and leave you ready if Gemini becomes advertising inventory later.

    Read Google’s AI ad strategy as two connected systems

    Google’s direction contains two different levels of certainty. The current testing ground is AI Mode, a Gemini-powered Search experience where ads are kept distinct from organic results, clearly labeled and shown only when Google considers them relevant. If an appropriate ad is not available, the experience can proceed without one.

    Gemini advertising belongs in the possible-later column. Google has indicated that lessons from AI Mode could eventually inform ads in the Gemini app, but it has not committed to a launch date, buying workflow or dedicated campaign format. Treat that as strategic direction, not media inventory you can forecast.

    At the same time, Google is expanding the creative work its existing systems can perform. Demand Gen’s Asset Optimization controls now group shorter auto-generated videos, automatic video resizing and images pulled from landing pages into a simpler set of toggles. This is operationally important: Google can test more combinations and placements when the advertiser supplies reusable source material.

    Strategic layerWhat is establishedWhat you should do
    Conversational deliveryGoogle is testing labeled, relevance-dependent ads in AI Mode.Map campaigns to the decisions users describe, not only the keywords they type.
    Creative assemblyDemand Gen can shorten videos, resize them and pull images from landing pages.Govern the original assets and inspect automated variants before relying on them.
    Gemini inventoryGoogle has left the possibility open, without announcing a buying product.Prepare reusable inputs, but do not assign a speculative Gemini budget.
    Personalized contextGoogle sees personalization as important, while broader Search integration remains prospective.Track product and data-policy announcements instead of assuming new targeting access.

    Use a no-regret test for every preparation project: would it still improve your current Search or Demand Gen operation if Gemini never carried ads? Clearer landing pages, better asset governance and stronger conversion measurement pass that test. A Gemini-only media plan does not.

    Build campaigns around the decision behind the query

    A strategist examines visual intent clues as shoppers follow branching paths toward different products and services.

    Keyword intent still matters, but conversational interfaces let a user express the situation around a purchase: who the product is for, what constraint matters and what must be true before they act. An ad can be relevant to the topic while being irrelevant to that decision. Your campaign brief should expose the difference.

    For each important offer, create a decision map with the following fields:

    1. Decision: Write the choice the user is trying to make, not the keyword category. A software buyer may be choosing a platform for a distributed team rather than searching for software in the abstract.
    2. Context: Record the audience, use case and stage of consideration that make the offer appropriate.
    3. Constraint: Identify the condition that can disqualify the offer, such as compatibility, geography, deployment model or required feature.
    4. Claim: State the promise the ad can make without exceeding what the landing page supports.
    5. Proof: Point to the specification, demonstration, policy, price information or other evidence that substantiates the claim.
    6. Destination: Choose the page that resolves this particular decision rather than sending every variation to a generic homepage.

    This map gives paid, SEO, AEO and content teams a shared factual base. It does not mean their distribution systems are interchangeable. An organic mention, an AI-generated answer and a paid placement have different eligibility and measurement rules, even when they rely on the same product facts.

    That distinction matters for structured data. JSON-LD can organize explicit facts about an entity, product, service or page, but nothing in Google’s current AI advertising direction establishes schema markup as an ad-targeting control. Use valid structured data to describe visible content accurately. Do not promise that adding markup will make an ad appear in AI Mode or secure future Gemini inventory.

    Review the landing page against the decision map before expanding creative. The page should make the intended audience, supported use case, important limitations and next action easy to find. If the ad needs a verbal explanation to remain accurate after the click, the page is not ready for automated distribution.

    Make creative automation safe before switching it on

    Two marketers review automated ad layouts while approved assets pass through digital guardrails and rejected assets are set aside.

    Demand Gen’s consolidated Asset Optimization panel reduces the work required to activate automation. It does not remove the need to supervise the material being transformed. A weak original can produce more weak variations, and an outdated landing-page image can become campaign creative without anyone deliberately selecting it.

    Audit the system in this order:

    1. Record the current settings. Open Asset Optimization and document which video-resizing, video-shortening and image options are enabled. Keep that record with the campaign brief so a later performance change can be tied to a known configuration.
    2. Create an approved asset register. For each original image or video, record the owner, usage rights, supported claim, intended audience, required context and any expiration condition. An asset should not enter automation merely because it exists in a shared folder.
    3. Inspect landing-page imagery. Because Google can pull images from the destination page, remove obsolete promotions, unsupported product states and decorative images that would be misleading when detached from the surrounding copy. Make the strongest eligible image understandable on its own.
    4. Review shortened videos as new creative. Confirm that the automated cut preserves the core claim, necessary qualification, brand identity and call to action. Watch it without sound as well as with sound. A cut that removes the condition attached to a claim should not run.
    5. Review every generated shape you intend to use. Check whether resizing crops the product, speaker, demonstration, captions, qualification or call to action. Do not assume that a technically valid crop is a persuasive or compliant ad.
    6. Change settings deliberately. Where campaign volume allows it, avoid changing every automation control alongside the landing page and offer. A smaller set of simultaneous changes makes the result easier to interpret, even if it is not a perfect controlled experiment.

    The practical division of labor is simple. Your team owns truth, permissions, positioning and the quality of the originals. Google can own format adaptation and selection only within those boundaries. If the boundaries are not documented, leave the relevant automation off until they are.

    Measure the system you can buy, not the product you imagine

    AI-flavored placement does not change the need for a falsifiable campaign brief. Before launching or expanding automation, state which user decision the campaign addresses, which conversion represents success and which downstream signal distinguishes a valuable conversion from a merely completed form or click.

    Your working scorecard should preserve enough context to explain a result:

    • The campaign, audience and offer being evaluated.
    • The landing-page version used during the period.
    • The status of each asset-optimization control.
    • The original assets available to Google.
    • The primary conversion and a business-quality signal, such as a qualified opportunity, completed purchase or retained customer.
    • Any brand, policy or lead-quality guardrail that would make higher delivery unacceptable.

    Do not use click-through rate alone to declare an AI ad strategy successful. A new format can attract interaction while sending poorly matched users. Read the engagement metric beside conversion quality, acquisition cost and the business outcome your campaign was meant to produce.

    Keep a separate launch gate for future Gemini advertising. Before moving money, verify the inventory available to your account, eligible campaign types, placement and exclusion controls, creative-generation settings, reporting granularity, conversion attribution and the data used for personalization. If Google does not expose enough information to answer those questions, the responsible response is a limited test inside the controls that do exist, not a broad budget shift.

    Personalization deserves particular care. Google’s Personal Intelligence can draw on a user’s Gmail, Photos and Calendar, and Google has said that user data will not be sold or shared. Broader integration with Search remains a possibility rather than an advertiser capability you can plan around. Do not translate consumer-facing personalization into an unsupported claim that advertisers can access those personal signals.

    This measurement discipline also keeps organic AI visibility separate from paid reach. Track whether your brand is represented accurately in AI-generated answers, but do not count a citation, a paid impression and an assisted conversion as the same event. They can influence the same journey without proving the same thing.

    Key takeaways

    • AI Mode is Google’s current environment for learning how labeled, relevance-dependent ads fit into AI-generated search experiences.
    • Ads in the Gemini app remain possible, but no dedicated buying format, launch date or workflow has been established.
    • Demand Gen’s asset controls reveal the immediate operational priority: provide strong originals and let automation adapt them under supervision.
    • Organize campaigns around a user’s decision, context, constraint, claim, proof and destination rather than treating conversational advertising as a longer keyword list.
    • Audit landing-page images because they may become ad assets, and review shortened or resized videos as new creative rather than harmless copies.
    • Keep structured data, organic AI visibility and paid placement conceptually separate. Shared facts help all three, but none guarantees the others.

    At your next campaign review, open the Demand Gen Asset Optimization panel, record its settings and inspect every page and asset it can draw from. Then build one decision map for your highest-value offer. When Google exposes more conversational inventory, you will have approved inputs and a measurement plan ready, without having paid for a strategy built on speculation.

    References

  • Transform Automated Workflows with Gamma Integration

    Transform Automated Workflows with Gamma Integration

    I’m thrilled to share that Profound Agents can now seamlessly create presentations, documents, and webpages within Gamma as part of my automated workflows. No more hassle of exporting data and rebuilding it elsewhere. My Agent takes the outputs from upstream nodes and crafts them into ready-to-share assets in Gamma, streamlining the entire process.

    Inspired by this post on Try Profound Blog.

    crushpress.ai community screenshot
  • Perplexity’s Amazon Bot Block: What Commerce Teams Should Do

    Perplexity’s Amazon Bot Block: What Commerce Teams Should Do

    If your AI commerce plan assumes an assistant can find a product, sign in and complete the purchase, the Perplexity-Amazon dispute exposes a flaw in that model: discovery, account access and transaction authority are separate permissions.

    A preliminary injunction now prevents Perplexity’s Comet agent from entering Amazon’s password-protected areas and requires Perplexity to delete the Amazon data it collected. That does not end AI shopping, but it gives SEO, ecommerce and agent teams a practical warning: being visible to an AI system does not give that system permission to act inside a platform.

    The injunction targets authenticated access, not all AI shopping

    The scope matters. U.S. District Judge Maxine Chesney issued a preliminary injunction concerning Comet’s access to password-protected parts of Amazon, including areas used by Prime members. It is not a final judgment declaring every AI shopping agent unlawful, nor does it establish that public product pages cannot be found, interpreted or recommended by AI systems.

    The central distinction is between two kinds of authorization. A customer may authorize an assistant to use the customer’s account, but the platform may still withhold authorization from the assistant itself. The judge cited strong evidence that users granted Comet access while Amazon did not. For anyone building an agent, user consent is therefore necessary but may not be sufficient.

    Amazon has accused Perplexity of computer fraud and unauthorized access, including allegedly allowing Comet to make purchases without identifying itself properly as a bot. Those are Amazon’s allegations, not settled findings on every claim. At issuance, the injunction was suspended for one week so Perplexity could appeal.

    The deletion requirement deserves as much attention as the access restriction. An agent team may need to identify and remove data by platform, account, user and collection method. If you cannot isolate data at that level, a dispute over one integration can turn into a much larger data-governance problem.

    Discovery, recommendation and purchase are separate systems

    Three connected but separate spaces represent product discovery, recommendation, and a locked checkout process.

    AI commerce is often discussed as one continuous journey, but three layers determine whether it works. Each has a different owner, failure mode and remedy.

    LayerQuestion it answersTypical responsibilityCommon failure
    VisibilityCan an AI system find and understand the product?SEO, content, structured data and public-site engineeringThe product is absent, misunderstood or cited inaccurately
    RecommendationDoes the product fit the user’s request well enough to be selected?Product information, positioning, availability and the agent’s decision logicThe product is understood but not chosen
    ExecutionCan the agent enter an account, modify a cart or complete a purchase?Authentication, platform policy, security, legal review and approved integrationsThe journey stops at sign-in, checkout or another protected action

    Schema markup can improve machine understanding at the visibility layer. Clear product details can strengthen the recommendation layer. Neither one grants an agent access to an authenticated account. Treating them as substitutes for platform permission creates a false sense of readiness.

    The same distinction applies to robots.txt and other crawl controls. A public crawling directive is not a purchasing authorization system. It does not answer whether an agent may use a signed-in session, accept terms, place an order or retain account data. Those questions need explicit product, security and legal decisions.

    Your SEO work still matters, but it cannot grant access

    The wrong reaction would be to stop optimizing products for AI discovery. The injunction concerns authenticated access, while much of the discovery and evaluation journey happens through public information. Your content can still help an assistant understand what a product is, who it suits and where the customer can continue safely.

    • Give each important product a stable public destination. Use a consistent canonical URL and make variant handling predictable so an agent does not have to reconcile several conflicting versions of the same offer.
    • Put decision-critical facts in accessible page text. Product names, identifiers, specifications, compatibility, options, limitations and fulfillment conditions should not exist only inside images or interface states that require interaction.
    • Keep structured data aligned with the visible page. Markup that contradicts the page can produce incorrect extraction and erode trust. Treat structured data as a machine-readable representation of the offer, not a place to publish claims the customer cannot verify.
    • Separate product availability from transaction capability. An agent may be able to report that an item appears available without being authorized to buy it. Use language and interfaces that do not blur those two states.
    • Provide a durable human handoff. If automated checkout is unavailable, preserve the selected product or variant in a public deep link and let the customer sign in, review the cart and confirm the purchase.
    • Publish an approved path for automation if you offer one. Document the permitted integration, identity requirements, data limits and prohibited actions. Do not force agent developers to infer transactional permission from crawlability.

    Measure these layers separately as well. AI referrals and product-page visibility tell you about discovery. Product selection or cart initiation tells you about consideration. Completed orders tell you about execution. Combining all three into a single “AI traffic” measure hides the exact permission gate where the journey fails.

    Audit the handoff before an agent reaches login

    A commerce specialist inspects digital permission tokens before an automated shopping device reaches a secured account gate.

    You do not need to wait for another court dispute to find the weak point in your own workflow. Trace one high-value purchasing journey from the first public result through order confirmation, then record the identity, permission and data rules at every transition.

    1. Mark every access boundary. Label which pages and actions are public, account-gated, membership-gated or restricted to an approved integration. Include cart changes, saved payment methods, order history and purchase confirmation.
    2. Name the permission owner. Record whether the customer, merchant, marketplace, payment provider or another party controls each action. If two parties must consent, capture both rather than treating the customer’s approval as universal authorization.
    3. Define agent identity. Decide how an automated system identifies itself and how your service distinguishes it from the human account holder. Do not rely on the fact that the agent is operating through a customer’s browser session.
    4. Minimize retained data. Keep only what the approved workflow needs, attach provenance to it and make deletion possible by source and account. The Amazon data-deletion requirement shows why broad, unlabelled data stores create operational exposure.
    5. Design a graceful stop. When the next action is not authorized, the agent should explain the boundary, preserve useful context and return control to the customer. It should not repeatedly retry, conceal its identity or route around the restriction.
    6. Test the fallback as a primary path. Confirm that the customer lands on the correct product and variant, can see what remains to be reviewed and can complete the protected steps without rebuilding the transaction.

    If your agent enters authenticated services or makes purchases, do not attempt to evade a platform block or disguise automated traffic. That can increase contractual, security and legal exposure. Have qualified counsel review the relevant terms, authorization model and data practices before launch; this dispute is too narrow and preliminary to serve as a universal legal rule for another platform or implementation.

    Key takeaways for AI commerce teams

    • A user’s permission to use an account does not necessarily provide the platform’s permission for an agent to access it.
    • The injunction is specific to Perplexity’s Comet agent and password-protected Amazon areas; it is not a general ban on AI product discovery or shopping assistance.
    • SEO, AEO and structured data improve visibility and understanding, but they do not authorize account access or transactions.
    • A useful agent-ready journey needs both a machine-readable discovery layer and an explicitly permitted execution path.
    • When full automation is unavailable, a precise human handoff is better than an agent that fails silently at login or checkout.
    • Data provenance and targeted deletion are core integration requirements, not cleanup tasks to invent after a dispute begins.

    Your next move should be concrete: diagram one purchasing journey, circle every point where the agent crosses from public information into protected action, and assign an owner to each permission. Keep optimizing the public layer for discovery, but do not describe the journey as agent-ready until the authenticated steps have an approved path or a tested human handoff.

    References