By Carson Cobb. Data collected November 4 to 11, 2026.
On Google, a shopping search shows ten results, ads and a carousel. In an AI answer, a shopper sees a short list. I counted exactly how short across 144 answers: 48 shopping questions asked in ChatGPT, Gemini and Google AI Mode.
The short version
- ChatGPT names a median of 4 brands per answer. Never more than 8.
- Gemini and Google AI Mode name a median of 5.
- A typical question has 8 brands in play across the three engines.
- Only 2 are named by all three. And 5 of 48 questions had no brand in common at all.
- The lists barely repeat. ChatGPT named 193 different brands, and only 8 appeared in more than one answer.
Brands named per answer
| Engine | Median | Average | Most in one answer | Answers with no brand | Different brands across 48 answers |
|---|---|---|---|---|---|
| ChatGPT | 4 | 4.2 | 8 | 3 | 193 |
| Gemini | 5 | 5.5 | 12 | 3 | 254 |
| Google AI Mode | 5 | 5.1 | 11 | 3 | 230 |
How much the engines agree
| For each question | Median |
|---|---|
| Brands named by any engine | 8 |
| Brands named by all three | 2 |
Of the 419 brand and question pairs in the data, 222 (53%) were named by only one engine, 101 by two and 96 (23%) by all three.
By product area
| Product area | Different brands named | ChatGPT median | Gemini median | AI Mode median | Named by all three, median |
|---|---|---|---|---|---|
| Supplements | 41 | 3 | 5 | 5 | 2 |
| Beauty | 51 | 3 | 4 | 4.5 | 1.5 |
| Food and drink | 96 | 5 | 8 | 6 | 2 |
| Pet | 71 | 4 | 6 | 5 | 2 |
| Home | 61 | 4.5 | 5.5 | 5 | 2 |
| Apparel | 73 | 4.5 | 6 | 5.5 | 2 |
What it means
Four slots is a much smaller shelf than a page of search results. And because the lists barely repeat from question to question, being named for one question says little about the next. A brand that shows up for “best gluten free granola” can be absent for “best granola for kids”.
That’s why I test many question variations, not one. I covered how much wording changes the answer in how question wording changes AI picks, and how stable the top pick is in how often AI recommendations change.
What to do with this
- Write down the 20 questions your buyers actually ask. By need, not by your product names.
- Ask all three engines. Agreement across them is rare, so one engine’s answer isn’t the picture.
- Track which questions you’re in, not a single score.
- Go after the questions you’re closest to. Where you’re named by one engine, find what the source behind it says, and earn the same coverage elsewhere.
Testing where AI 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, eight each in supplements, beauty, food and drink, pet, home and apparel, written the way shoppers ask.
- Engines: ChatGPT, Gemini and Google AI Mode, each in a fresh session, November 4 to 11, 2026.
- Counting: every brand named in each answer was coded by hand and counted once per answer.
- Limits: one run per question per engine. Answers vary between runs, so treat the counts as typical, not fixed.
No store or brand is named in this report.