By Carson Cobb. Research from November 4 to 11, 2026.
Gemini sits inside Google’s apps, Android phones and Chrome, so it reaches shoppers who never open ChatGPT. I asked it the same 48 shopping questions I asked ChatGPT and Google AI Mode, and compared how it picks.
The short version
- Gemini names more brands. A median of 5 per answer and up to 12, against 4 and 8 for ChatGPT.
- It quotes certifications twice as often as ChatGPT. 23 of 48 answers versus 12.
- It rarely shows prices. 12 of 48 answers, against 38 for ChatGPT.
- It rarely asks about the shopper. It suggests related topics instead.
- It gets ratings wrong more than prices. Prices right for 22 of 24 products, ratings for 14 of 23.
Gemini, ChatGPT and AI Mode side by side
| Measure | Gemini | ChatGPT | Google AI Mode |
|---|---|---|---|
| Brands per answer, median | 5 | 4 | 5 |
| Most brands in one answer | 12 | 8 | 11 |
| Different brands across 48 answers | 254 | 193 | 230 |
| Answers citing certifications | 23 | 12 | 25 |
| Answers citing organic claims | 16 | 7 | 14 |
| Answers showing prices | 12 | 38 | 30 |
| Correct price, specific product | 22 of 24 | 22 of 24 | 19 of 24 |
| Correct rating, specific product | 14 of 23 | 17 of 23 | 12 of 22 |
Sources: how many brands AI recommends, why AI recommends products, prices in AI shopping answers, what AI gets wrong.
What stands out
Gemini behaves like a researcher making a longer list. More brands per answer, more emphasis on certifications and claims, fewer prices. It also agrees with ChatGPT less than you’d think: for a typical question, only 2 brands were named by Gemini, ChatGPT and AI Mode alike (brand overlap).
The longer list is good news for smaller brands. A fifth or sixth slot is a real opening. The weak spot is ratings: Gemini got the rating wrong most often where a review app had added a second copy of the product data, right just 2 times out of 8.
What to do with this
- Name your certifications plainly on product pages, with the certifying body.
- Use one Product block with your rating in it, so Gemini doesn’t pick the wrong number.
- Test Gemini separately. Its list differs from ChatGPT’s.
- Make your price easy to find on the product page and in structured data, since Gemini rarely volunteers it.
Testing where Gemini names you is part of my AI Search and Conversion Audit. The AI Visibility Check is a quick first look.
How this was done
- Questions: 48 shopping questions across supplements, beauty, food and drink, pet, home and apparel, asked in Gemini, ChatGPT and Google AI Mode in fresh sessions, November 4 to 11, 2026.
- Fact checks: each AI’s stated price and rating compared with 24 product pages.
- Limits: one run per question per engine, and Gemini changes often.
No store or brand is named in this report.