AI-powered shopping experiences are becoming a more consequential layer between ecommerce sites and potential buyers. Rather than presenting a conventional list of product links, platforms can compare options, summarize reviews and pricing, explain recommendations, and direct shoppers to merchants. That shift is often described as agentic commerce. It does not mean AI agents have replaced ecommerce search or that every product category is affected equally. It does mean product discovery increasingly happens in conversational and AI-generated interfaces, including Google AI Mode, Gemini, and ChatGPT. For merchants, visibility now depends not only on traditional organic rankings and shopping feeds, but also on whether AI systems can retrieve, interpret and confidently present product information. Google's agentic AI shopping announcement provides a concrete view of this direction. The company describes shopping features in AI Mode and the Gemini app, alongside agentic checkout capabilities. Google says its Shopping Graph contains more than 50 billion product listings, with 2 billion updated hourly, and that AI Mode can organize information such as price, reviews and inventory for shoppers. What agentic commerce changes in product discovery The important change is the format of the decision journey. A shopper can ask for a product that meets several constraints, such as a budget, use case, size, availability or feature set. An AI interface may turn that request into a shortlist and provide a rationale for its selections. OpenAI describes a similar shopping experience in ChatGPT. Its shopping research documentation explains that ChatGPT can surface product options, accompanying rationales and links to merchants. This is a different discovery environment from a standard results page, even where it still sends users to retailer or brand websites. For ecommerce teams, the practical implication is straightforward: a product page has to work as a source of accurate, accessible evidence. An AI-generated recommendation may draw on product data, merchant information, reviews, inventory signals and page copy. If those inputs are incomplete, contradictory or hard to interpret, the product may be harder for an AI system to compare with alternatives. The current evidence does not establish a universal formula for being recommended. Google and OpenAI describe capabilities and experiences, while independent work on how shopping agents evaluate products is still emerging. Platforms can also vary in how they retrieve information, weigh signals and formulate a response. Product information becomes a retrieval and interpretation problem Strong product data has long supported ecommerce SEO and Google Shopping. In AI shopping contexts, it also helps systems assemble a coherent answer to a shopper's request. Start with the information a buyer would need to compare an item fairly: A clear product name, brand and model or variant. Accurate price, currency, availability and delivery-relevant details where applicable. Product attributes such as dimensions, materials, compatibility, capacity, colour, intended use and key specifications. Genuine review and rating information where it is available and eligible to display. Clear policies and supporting details that remove uncertainty for a buyer. Use Product and Offer structured data where it accurately represents the page, and keep it aligned with the visible page content and merchant data. Structured data is not a guarantee of inclusion in an AI response. Its value is that it gives platforms a more consistent way to understand important product facts. Data freshness matters as much as completeness. A recommendation that cites an old price or unavailable variant creates a poor shopper experience and can undermine confidence in the source. Teams should therefore treat feeds, structured data, inventory systems and product pages as connected publishing surfaces rather than separate technical tasks. Write product pages for people and machine interpretation Machine-readable data alone rarely answers every buying question. Product pages also need plain, specific language that explains who the item is for, what it does, where it fits and what limitations may matter. That does not call for writing copy for a model instead of a customer. It calls for reducing ambiguity. Avoid relying solely on broad marketing terms when a product's practical features can be stated directly. For example, a page can clearly describe compatibility, setup requirements, included components or use-case boundaries alongside its commercial messaging. Helpful page content makes it easier for both shoppers and systems to distinguish between closely related products. It can also give an AI interface evidence to explain why one option fits a stated need better than another. The Search Engine Journal webinar that prompted this discussion frames the issue around product-page language, AI readability and underlying data. Its webinar landing page is promotional material, not a published performance study. Still, its focus reflects a genuine operational question for retailers: whether their product information is usable beyond a traditional search result. Measure visibility beyond rank tracking Traditional rank tracking remains useful, but it cannot fully describe AI-mediated discovery. A product may be mentioned in an AI-generated shortlist, cited as a merchant option, shown in a product comparison or omitted despite ranking well for a related keyword. A practical measurement programme should separate several signals: Signal What it can indicate Important limitation AI answer presence Whether a platform generates a shopping-oriented response for a target query Presence does not show which merchants or products were selected Brand or product mentions Whether the brand, product or category is included in the response A mention does not necessarily produce a visit or sale Merchant links and citations Whether a response points shoppers to a specific retailer or product page Link treatment and availability vary by platform and query Downstream site data Changes in referral traffic, landing-page engagement and conversion patterns Attribution may be incomplete and does not prove an AI response caused the result Use a controlled set of representative queries, including category, use-case, comparison and constraint-based prompts. Record the platform, date, location and query wording, because results can change with the interface, the market and the phrasing of the request. Compare product-level exposure across Google AI surfaces, ChatGPT and other relevant tools instead of assuming that a result on one platform will transfer to another. Most importantly, do not confuse AI visibility with business impact. A cited product may receive no measurable traffic, while an uncited product could still benefit indirectly from improved data quality and conventional search performance. Track the full path where possible, but be careful about causal claims. If your product catalogue is difficult to compare, has inconsistent attributes or lacks clear buyer-facing detail, create a GEO Search Leads account to begin assessing your website's GEO readiness and identify optimization opportunities. A structured review can help prioritize the pages and information gaps that matter most as AI-generated shopping discovery evolves. Frequently Asked Questions What is agentic commerce? Agentic commerce describes shopping experiences in which AI systems help shoppers research, compare and sometimes progress toward purchasing products. In practice, this can include AI-generated product shortlists, recommendation rationales, pricing and inventory information, and merchant links. Does Product structured data guarantee inclusion in AI shopping recommendations? No. Structured data can help platforms interpret product facts when it accurately matches the page, but it does not guarantee that a product will be retrieved, cited or recommended in an AI response. Which AI shopping platforms should ecommerce teams monitor? The relevant platforms depend on the audience and market, but Google AI Mode, Gemini and ChatGPT are important examples in the current AI shopping landscape. Results and product treatment can differ by platform and query type. Can AI shopping visibility be measured like a keyword ranking? Not exactly. Teams can monitor product mentions, merchant links, citations and answer presence for a controlled query set, then compare those observations with referral and conversion data. These signals should not be treated as direct equivalents to a conventional ranking. Conclusion Agentic commerce is making product information a more central part of AI search visibility. The near-term priority is not chasing unsupported promises of recommendation placement. It is maintaining accurate, structured and clearly explained product pages, then measuring how products appear across evolving AI shopping surfaces.