AI Shopping Runs on Your Product Feed, Not Your PDP

The data point that should reorder your priorities

If you run e-commerce SEO, you’ve probably spent the last two years optimizing product detail pages — titles, schema, reviews, unique descriptions. That work still matters. But there’s a new finding worth sitting with: 83% of the products ChatGPT surfaces in shopping queries match what’s in the top 40 of Google Shopping results. Not organic listings. Shopping results. Which means those recommendations are being pulled substantially from Google Merchant Center feed data, not from crawling and parsing your product pages.

This is a meaningful shift in where the ranking signal actually lives. For years, PDP optimization was the whole game — get the page right, and everything downstream (rankings, rich results, shopping ads) followed. AI shopping surfaces — ChatGPT, Gemini, Perplexity — are increasingly sourcing structured product data directly from the feed layer instead of re-deriving it from HTML. If your feed is thin, stale, or inconsistent, you’re invisible in a growing slice of purchase-intent queries, regardless of how good your PDP copy is.

What this means practically

The feed isn’t a secondary export anymore — it’s a primary ranking surface, sitting alongside the PDP rather than beneath it. That’s a real reprioritization for teams that have treated Merchant Center as a “set it up once, let it sync” utility.

Concretely, that means:

• Every product needs to actually be in Merchant Center, with complete, accurate attributes — brand, GTIN/MPN, high-resolution images, current price, availability, and category. Gaps here aren’t cosmetic; they’re the difference between showing up in an AI shopping answer and not existing for that query at all.
• Feed freshness is now a visibility factor. Price changes, stock-outs, and discontinued SKUs need to propagate fast. A feed that’s accurate on Tuesday and wrong by Friday is quietly costing you placement.
• Category structure inside the feed deserves a real audit — not because fixing it alone guarantees an AI recommendation, but because a feed with inconsistent or missing category mapping gives every downstream system (Google Shopping, and by extension the AI layers reading it) less to work with.

What this doesn’t mean

It doesn’t mean PDP work is wasted effort. The same research found that 88% of AI shopping offers still originate from product pages — the feed and the page are complementary inputs, not competing ones. Google Shopping surfaces the candidate set; the page (and its schema, reviews, and content depth) still does a lot of work in how that product gets described, priced-compared, and justified in an AI answer. Treat this as “feed work is now equally required,” not “feed work replaces page work.”

It also doesn’t mean feed optimization is a silver bullet. Getting your category taxonomy clean in Merchant Center is necessary hygiene, not a guarantee of an AI recommendation. Relevance, pricing competitiveness, reviews, and availability still decide who wins the slot once you’re in the candidate pool.

Add it as a KPI, not just a task

The practical follow-through here is measurement. Most e-commerce teams don’t currently check whether their products show up when you ask ChatGPT, Gemini, or Perplexity a shopping-intent question in their category. That’s worth adding as a recurring check — sample a handful of your core product queries monthly, note whether you appear, and treat drop-off as a signal to check the feed before assuming it’s a content problem.

For teams running large catalogs, this is also where feed hygiene issues compound fastest — missing GTINs on a few SKUs is a rounding error, but the same gap across ten thousand products is a structural visibility problem. If you haven’t audited your Merchant Center feed against this lens recently, that’s the place to start before touching another product page. If you want a second set of eyes on where AI-driven shopping visibility is leaking, our GEO audit covers exactly this gap between feed and page.