AI shopping is not a future concern for ecommerce teams. Google AI Overviews and ChatGPT are already surfacing product recommendations in response to purchase-intent queries. If your product pages are not built to be retrieved by these systems, you are losing consideration at the very moment a buyer is making a decision.
Chris Sachs and Aris Vrakas distilled the problem into four practical moves: make product pages retrievable, use shopper language, connect feeds with pages, and test the results. That framework is deceptively simple. Each step requires deliberate work, and most ecommerce sites fail on at least two of them.
Retrievability Is Not the Same as Indexability
Many product pages are technically indexed but practically invisible to AI systems. Retrievability goes further than crawlability. It means the content on the page is structured in a way that lets an AI model extract and use it confidently. Thin descriptions, duplicate boilerplate, and content buried inside JavaScript renders all create friction for AI extraction - even when Googlebot can reach the page.
The question to ask is not "can Google index this?" but "can an AI model construct a useful, accurate answer about this product using only what is on this page?" If the answer is no, you have a retrievability problem. Product name, key specifications, use cases, compatibility, pricing signals and availability all need to be present in crawlable HTML, not locked behind dynamic rendering.
For teams running AEO programmes, this is where a structured content audit pays off. Go through your top-revenue product pages and test them with that single question. What can an AI model confidently say about this product based solely on the page content? The gaps tend to become obvious quickly.
Shopper Language Is a Specific Technical Requirement
The instruction to "use shopper language" sounds like basic copywriting advice. It is not. In the context of AI search, it is about semantic alignment between the words buyers use in queries and the words present on your product pages and in your feed attributes. AI models match intent to content. If your product descriptions are written in the language of your warehouse team or your brand guidelines, they may not match the natural language queries arriving through ChatGPT or Google AI Mode.
A buyer searching for "waterproof walking shoes for wide feet under £80" is using specific, layered intent. If your product pages say "hydrophobic trail footwear - E-width fitting" and nothing else, the semantic gap is wide enough that AI systems may not surface you even if you are the perfect match. The fix is to include the full range of ways buyers describe what they want - in product descriptions, in FAQs on the page, and in feed attributes.
The Feed-to-Page Gap Is Where Visibility Dies
One of the most consequential moves Sachs and Vrakas identify is connecting feeds with pages. This is a specific structural problem that affects a large number of ecommerce businesses. The product feed submitted to Google Merchant Center - with its carefully maintained titles, descriptions, and attributes - often does not match what is on the actual product page. The feed has been optimised for Shopping ads. The page has been optimised for conversion rate. Neither has been optimised for AI retrieval.
AI systems draw on both signals. A well-structured feed tells Google what a product is. A well-structured page tells an AI model how to describe it, recommend it, and answer follow-up questions about it. When these two sources contradict each other - different titles, missing specifications, inconsistent pricing language - AI systems lose confidence in the product data. The result is reduced visibility in AI-generated shopping recommendations.
The practical fix requires cross-team coordination that many ecommerce operations lack. The feed team, the content team, and the trading team all touch product data, often without a shared standard. Establishing a single source of truth for product information - and ensuring it is reflected consistently in the feed, the page, and any structured data markup - is the operational work that makes AI shopping visibility possible.
Testing Is Not Optional - It Is the Whole Point
The fourth move - test the results - matters because AI shopping surfaces are not transparent. You cannot check a ranking report to see where your products appear in ChatGPT responses or Google AI Overviews. You have to run queries. That means building a systematic testing programme: identify the high-intent queries your buyers use, run them regularly across AI platforms, and document which products appear, which competitors appear, and what language the AI uses to describe them.
This is manual work, but it is not unstructured. A query set built from your existing search term reports, customer reviews, and site search data gives you a representative picture of real purchase intent. Run those queries in Google AI Overviews, in ChatGPT with browsing enabled, and in Perplexity. Record the citations. When a competitor appears and you do not, the gap is usually traceable back to one of the first three moves: retrievability, shopper language, or feed-to-page alignment.
Attribution is the harder problem. AI-referred shopping traffic does not always arrive with clean referral parameters. Users who discover a product through a ChatGPT recommendation may then search directly or navigate to your site through a different route. Building UTM-based tracking for AI referral sources - and cross-referencing with your CRM to identify the revenue contribution - requires deliberate setup. Without it, you are optimising blind.
What This Means in Practice for Ecommerce Teams
The four moves outlined by Sachs and Vrakas are not a one-off project. They describe an ongoing operational posture. Retrievability degrades as sites change. Shopper language shifts as buyer vocabulary evolves. Feed data drifts as inventory changes. The testing programme keeps you honest about whether the work is having an effect.
The ecommerce teams that will win in AI shopping surfaces are not necessarily the ones with the biggest budgets or the most sophisticated technology. They are the ones that treat product page quality as a commercial priority, not a marketing afterthought. Product information that is accurate, complete, linguistically matched to buyer intent, and consistent across feeds and pages is the foundation. Everything else - schema markup, structured data, digital PR citations - builds on top of it.
If your product pages are not currently built to that standard, the AI shopping visibility problem you have today will compound as these surfaces grow. The time to close the gap is now, before the retrieval habits of AI systems harden around your competitors' data.