By Carson Cobb. Research from November 4 to 17, 2026.
Google AI Mode passed 1 billion monthly users in 2026, according to Google (Search Engine Journal). It looks like Google, but it doesn’t pick like Google. Here’s what decides its shopping recommendations, from my tests of 48 shopping questions and a side by side comparison of 19 questions in Google and AI Mode.
1. Product listings, not articles
In 11 of 19 shopping questions, AI Mode built its answer around product listings, each with a price, a star rating, a review count and a seller. Google says shopping results like these rely on product data in Merchant Center and its Shopping Graph. That makes your product feed as important as your rankings. (Google rankings vs AI Mode)
2. Not the first page of Google
Only 40% of the sites AI Mode cited ranked on Google’s first page for the same question, 36 of 91. Google’s number one result was cited for 7 of 19 questions. Reddit, the top Google result for 11 of 19, was cited by AI Mode 4 times.
3. Brand stores, including small ones
Brand stores were 16% of Google’s first page but 37% of the sites AI Mode cited. 27 of 60 product listings were sold by the brand’s own store, and several had fewer than 100 reviews.
4. Certifications and specific claims
Across 48 questions, AI Mode cited certifications in 25 answers, against 12 for ChatGPT, and organic claims in 14. It labeled picks for a specific need, like “best for heavy pullers,” in 51% of its labels. (Why AI recommends products, best for labels)
5. A short list that differs from ChatGPT’s
AI Mode named a median of 5 brands per answer. For a typical question, only 2 brands were named by AI Mode, ChatGPT and Gemini alike. Being in one engine’s answer says little about the others. (How many brands)
6. Facts from your data, mostly right
AI Mode got the price right for 19 of 24 products I checked, but ratings for only 12 of 22, and it struggled most where a review app added a second copy of the product data. (What AI gets wrong)
What this means for a brand
| What decides it | What to work on |
|---|---|
| Product listings | A complete Merchant Center feed: GTINs, product type, variants, shipping and returns |
| Not Google rank | Track AI Mode separately from rankings |
| Brand stores | Product pages that answer the question directly |
| Certifications | Name certifications and test results plainly |
| Facts | One Product block, ratings in structured data, matching feed and page |
In my scan of 617 Shopify brand stores, only 23% put a GTIN in product structured data and 11% put return or shipping terms there (agentic commerce readiness). That’s the gap AI Mode’s product listings expose.
Finding where AI Mode names you and why is part of my AI Search and Conversion Audit. The AI Visibility Check is a quick first look.
How this was done
- AI Mode tests: 48 shopping questions in November 2026 across six product areas, plus 19 questions compared side by side with Google search on November 14, 2026, from a US session.
- Fact checks: AI Mode’s stated prices and ratings compared with 24 product pages.
- Limits: this measures what AI Mode does, not Google’s internal rules, which can change.
No store or brand is named in this report.