Google's AI Mode linked only to Google's own pages in 46.5% of U.S. product answers, and 57% of all links in those answers led to Google Shopping product pages. That is the headline finding of a new Findrix study shared by Search Engine Land on October 29, 2026, and it changes how ecommerce brands should think about visibility in AI search. If your growth model depends on Google sending shoppers to your product pages, the front door has quietly moved inside Google's own walls.
The short answer for busy operators: product discovery in Google AI Mode is increasingly a closed loop. Google answers the question, shows its own Shopping results, and often links nowhere else. Fashion and beauty categories are hit hardest. Services and software still get outbound links, which means the impact depends heavily on what you sell.
What the Findrix study actually measured
Findrix, an AI visibility platform, tested 18,400 buyer questions across 35 U.S. markets between October 14 and 17, 2026. Each question was run once per platform on ChatGPT, Gemini, and Google AI Mode, producing 55,198 AI-generated answers. The study measured the links shown in those answers — not actual clicks, and not purchases.
Two numbers stand out from Search Engine Land's coverage of the study:
- 46.5% of U.S. product answers in AI Mode linked exclusively to Google's own pages.
- 57% of links in AI Mode product answers led to Google Shopping product pages.
Findrix co-founder Yosif Merman attributed the pattern to Google's product database, which contains prices, ratings, and inventory information. In other words, AI Mode can assemble a complete product answer from Google's own Shopping Graph without ever citing a retailer's website.
Which categories are most affected
The study's category breakdown is where this gets practical. Here is how often AI Mode linked only to Google pages, by product category:
| Product category | Answers linking only to Google pages |
|---|---|
| Swimwear | 60.4% |
| Women's everyday fashion | 57% |
| Sun care | 54.1% |
| Activewear | 53.8% |
| Sports nutrition | 7.6% |
| Local services (plumbers, lawyers, dentists) | 5.1% of links to Google pages |
| Software | 1.1% of links to Google pages |
Fashion and beauty dominate the top of that table, with one notable exception: sports nutrition, where just 7.6% of answers linked only to Google pages. The pattern suggests Google's product database is richest and most trusted in visual, commodity-driven categories where price and rating comparisons decide the purchase.
For local services and software, the story is nearly the opposite. Only 5.1% of AI Mode's links for local services went to Google pages, and for software questions the figure dropped to 1.1%. If you sell services or software, AI Mode still cites the open web — your content still earns its place in the answer.
Why Google can answer shopping questions without you
Classic search worked on a simple exchange: Google indexed your product pages, ranked them, and sent you traffic in exchange for content quality. AI Mode collapses that exchange when the answer can be built entirely from Google's own Shopping Graph — prices, ratings, availability, and product imagery that merchants already feed into Google Merchant Center.
That is the uncomfortable implication of Merman's point. The data retailers upload to sell through Google Shopping has become the raw material Google uses to answer product questions without linking out. Your product feed is no longer just an ad asset; it is now your primary representation inside AI answers. Referral analytics alone will not capture that visibility, because a shopper can discover, compare, and even act on your products without your website ever being visited.
ChatGPT and Gemini behave very differently
The same study shows this is a Google design choice, not a law of AI search. None of Gemini's measured links went to google.com. ChatGPT included roughly 25 links per answer, compared with about three for AI Mode — and in product answers, ChatGPT linked to Walmart in 45.1% of cases and Amazon in 6.1%.
Reddit citations also diverged sharply: ChatGPT cited Reddit in 35.6% of all U.S. answers, versus 5.1% for AI Mode and 1.5% for Gemini.
The practical translation: different AI engines route product demand to different destinations. Google routes it to Google. ChatGPT routes it heavily to Walmart, Amazon, and community discussions. A visibility strategy that only optimizes for one engine is leaving entire demand channels unmeasured.
What this means for your ecommerce strategy
Here is our read at Vivid, based on how we build organic visibility for product and service businesses:
- Treat your product feed as SEO. Complete Merchant Center data, fresh pricing, ratings, and inventory now determine whether you appear inside AI Mode's Google-sourced answers at all.
- Measure share of voice inside answers, not just referrals. If 57% of product links never leave Google, your analytics undercount your real visibility. Track appearance and citation in AI answers directly.
- Know your category's exposure. If you sell fashion or beauty, assume AI Mode is a walled garden and invest in the surfaces it favors. If you sell services or software, your informational content still earns outbound citations — double down on it.
- Diversify across engines. ChatGPT's Walmart-and-Amazon pattern means marketplace presence and Reddit visibility matter more there. One engine's closed loop is another engine's open door.
- Watch click data as it matures. Google has said click data for AI Mode is coming; until then, studies like Findrix are the best available signal. We covered the link redesign that made this measurement gap explicit in a recent piece on AI Mode's link update and the missing click data.
If you are rethinking where your products actually get discovered, this is exactly the kind of shift we map for clients — you can start a conversation with us here.
Read next
- Reddit says AI Mode "does not drive traffic" while Alphabet sheds $500B — the same-day reckoning
- Anthropic wants courts to pull the plug on risky AI. OpenAI wants "shutdown controls."
- GitLab Orbit is here: the context graph that makes AI agents more accurate and cheaper to run
- Earlier: Google's AI Mode link makeover: more clickable than ever, still no click data
Conclusion
The Findrix study, reported October 29, 2026, quantifies what many ecommerce teams suspected: Google AI Mode keeps product shoppers inside Google's ecosystem almost half the time, sending 57% of product-answer links to Google Shopping and linking exclusively to Google pages in 46.5% of U.S. product answers. Fashion and beauty brands face the most closed loop; services and software still receive genuine outbound citations, and ChatGPT's Walmart-heavy, Reddit-rich linking shows that alternative engines play by different rules. The strategic response is to treat your product feed as a first-class SEO asset, measure visibility inside answers rather than relying on referrals, and diversify across the engines where your buyers actually ask questions. The storefront window has moved; the smart move is to redecorate the new window, not stare at the old door.
