In traditional search, a shopper lands on your product page and judges it. In AI shopping, the judging often happens first, using data you supplied to a feed, descriptions written by reviewers and opinions posted in forums. Your page may be visited afterwards, or never.
That shifts ecommerce AI SEO away from a single playbook. Different AI surfaces pull from different places, so the useful question is which source each one trusts for each kind of buying question.
Google Says Little, and Says It Plainly
Google’s guidance on its AI features states that there are no additional requirements to appear in AI Overviews or AI Mode, and no special schema is needed. The only shopping-specific line is a reminder, listed among general good practice, to check that your Merchant Center and Business Profile information is up to date. That page was last updated in December 2025.
A commercial agency’s reading of Google’s product data rules adds detail: required attributes include id, title, description, link, image link, availability and price, with brand required for most new products and GTIN strongly recommended where one exists. That comes from a secondary source, so confirm it in Google’s current specification before changing a feed. The agency also found no Google page making a feed a condition for AI Mode listings, which is an absence of evidence, not proof.
ChatGPT Shopping Leans on Feeds
The strongest recent evidence comes from Profound, an AI visibility vendor. Tracking customers’ shopping prompts in ChatGPT, it reported that feed-integrated product picks jumped from 8.26% to 61.54% on a single day in July 2026. By its estimate, products without a feed integration may not be considered for about 65% of shopping recommendations.
Two other numbers matter. The top ten merchants’ share of mentions rose from 22.5% to 41.8%, and the number of unique merchants referenced fell from 13,524 to 10,607. Smaller stores may be losing ground.
Treat this with care. The data covers only Profound’s customers, it is observational, and Profound sells tracking tools. It attributes the shift to the GPT-5.6 release on 9 July, but reporting on the study notes OpenAI has not confirmed that link. The 65% figure is an extrapolation. The sensible conclusion is narrower: feed quality probably matters more than it did, and you should find out whether your catalogue is reaching ChatGPT.
Google’s AI Overviews Cite Third Parties
A very different pattern appears when Google answers shopping questions. A small October 2026 study from Hasmord searched 60 buyer-style questions across 12 categories on one day. Of the 55 that produced an AI Overview, each cited a median of seven sites.
Brand sites were cited in all nine “how to choose” overviews but only eight of 23 “best X for Y” overviews. In 11 “X vs Y” comparisons, a compared brand’s own site was cited just once. About 64% of cited sites sat outside the top ten organic results for the exact question.
The caveats are real: 60 questions, one search each, US English, desktop only, and one person classifying sites. These are illustrations of a pattern, not market estimates. Still, they suggest that when a shopper asks which product is best, other people’s opinions get the citation.
What to Do This Month
Audit the feed. Compare titles, prices, availability and identifiers against your live pages. Mismatches create doubt, and the agency’s reading of Google’s rules is that a product missing a required attribute cannot serve in ads and free listings.
Write product pages that answer comparisons. State dimensions, materials, compatibility, warranty and who the product is wrong for. Reviewers and assistants both reward candour.
Publish one real buying guide per category. Base it on questions customers send to support, and name trade-offs plainly.
Make independent coverage easy. Offer review units, clear spec sheets and fair press contacts. Do not seed fake reviews; assistants and shoppers both penalise obvious manipulation.
Watch your own demand signals. Sarkar SEO suggests logging product enquiries that mention an assistant or a review video, since these show influence that analytics may not capture.
A Quick Diagnostic for Your Best Sellers
Pick your five best-selling products. For each, ask the assistants your buyers use three questions: one of the “how do I choose” kind, one “best for” and one “versus a named rival.” Note which source types appear each time: product cards that look feed-driven, review sites, video, forums or your own pages. The gaps show where to invest first. Because answers vary between runs, repeat each question a few times before acting on a single result.
Smaller merchants should read the concentration figures as a prompt, not a verdict. If large stores are gaining share, completeness of your data and distinctive independent coverage become the ways to stand out.
What Not to Conclude
Do not abandon your product pages. Shoppers still verify price, stock and returns before buying, and these studies measure citations, not sales. Do not assume today’s pattern will hold either. Profound’s data covers a few weeks, and shopping features are changing fast.
Sarkar SEO treats ecommerce AI visibility as a data-quality and reputation problem first, and a content problem second. If you want an audit of your feed, product pages and the independent sources shaping your category, Sarkar SEO can map the gaps and prioritise fixes that fit your catalogue size.
