Most retail pages that show up in a Google AI Overview or a ChatGPT answer already ranked on page one before the AI answer existed. Ahrefs analyzed 146 million search results and 1.9 million AI Overview citations in a study published January 20, 2026.
It found that 76% of cited URLs also ranked in the traditional top 10, with a median position of second. That's the first fact worth knowing about retail AI Overview citations: there's no separate track to climb.
What Actually Earns Retail AI Overview Citations
Rank alone doesn't explain everything. The same Ahrefs research, drawn from 76.7 million AI Overview citations, found that branded web mentions correlate with citation at 0.664 and branded anchor text at 0.527. Both outrank most of the technical signals SEO teams usually chase.
YouTube mentions scored highest of all, at 0.740, which is one reason YouTube is the single most-cited domain across AI Overviews generally.
None of that fits into a product page. It comes from other sites naming your store, reviewing it, or linking to it. Earning those mentions is closer to digital PR and content work than to a technical checklist.
Schema Markup Won't Save a Page AI Already Ignores
In a study published May 11, 2026, Ahrefs tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026, matched against 4,000 control pages. Citations in Google AI Overviews fell 4.6% relative to the control group after the schema went live.
Movement in Google's AI Mode and in ChatGPT stayed inside 2.4%, close enough to zero that the researchers called it statistically indistinguishable. Adding schema, they concluded, produced no major uplift in citations on any platform.
Google says much the same from its own side. Its guidance for generative AI search states plainly that structured data isn't required and that there's no special schema.org markup to add for AI Overviews.
That doesn't make markup pointless. The pages in the Ahrefs study already had over 100 AI Overview citations each before the test started, so schema might still help a page AI can't parse at all get picked up in the first place.
For a retail catalogue built on a template that hides prices and variants inside JavaScript, getting the underlying product data structured correctly is still worth doing. It just won't outrank a page nobody links to.
The Categories Losing the Most Ground
Photo by Brett Jordan on Pexels
Adobe's analysis of transactions across more than a trillion visits to U.S. retail sites, reported by Digital Commerce 360 in August 2026, found AI-referred traffic up 62% year over year in July 2026, converting far better than non-AI traffic for an eleventh straight month running.
Adobe also scored how much of each category's product content a large language model could actually read.
| Category | Share of product content that's machine-readable |
|---|---|
| Apparel | 76% |
| Electronics | 70% |
| Cosmetics | 68% |
| Sporting goods | 67% |
| Furniture and home | 64% |
| General merchandise | 63% |
| Grocery | 59% |
Grocery and general merchandise sit furthest behind. A large share of their catalogue content isn't something an AI system can reliably parse and quote.
What to Actually Do About It
Fix the order of operations. Rank has to be solid first, since AI Overviews draw from the same index as organic search. Then work on being named and linked from other sites, since that correlates with citation harder than anything on the page itself.
Schema and clean product markup come after that, as the floor that keeps a page readable rather than the lever that lifts it.
We walked a similar business through this for local search in our restaurant AI search case study. Visibility followed the sites already being mentioned elsewhere, not the ones with the most markup.
If your product pages are the kind JavaScript renders in after load, that's worth testing first. It's the kind of audit our AI and automation team runs before recommending anything else.
