GEOly Research · Published 2026-11-02
How AI answers talk about brands: 35 million evaluations across 311,339 brands
Every time ChatGPT, Google AI Mode or Google AI Overviews recommends a product, it also judges it: good value, fragile, hard to set up. We extracted those judgments from weekly runs of 26,557 shopping and product prompts and measured what AI praises, what it criticizes, and who gets named at all.
Key findings
- 89.7%of AI remarks about named brands are positive; only 7.3% are negative.
- 18.2%of remarks about price are negative — price is both the most discussed and most criticized attribute.
- 65.1%of all AI brand mentions go to the top 1% of brands; the bottom half share 0.7%.
- 58.9%of topics get the same #1 brand from ChatGPT and Google AI Mode — the rest disagree.
- evaluations
- 35.0M
- brands
- 311,339
- topics
- 26,557
- AI engines
- 3
AI is overwhelmingly positive about the brands it names
When an AI answer names a brand and says something about it, the remark is positive 89.7% of the time, neutral 3.0% and negative 7.3%. Being named is mostly an endorsement — which makes not being named the real risk.
Google AI Mode is the most positive engine
ChatGPT carries most of the volume and is the most critical of the three engines. Google AI Mode is the most positive. All three stay above 89% positive.
| Engine | Positive | Neutral | Negative | Sampled evaluations |
|---|---|---|---|---|
| ChatGPT | 89.5% | 3.0% | 7.5% | 326,972 |
| Google AI Mode | 94.0% | 1.8% | 4.2% | 13,161 |
| Google AI Overviews | 92.2% | 4.4% | 3.4% | 5,131 |
Price is what AI talks about most — and criticizes most
Each evaluation is tied to the attribute it judges. Price is the single most discussed attribute (5.9% of all evaluations) and the most criticized: 18.2% of price remarks are negative, and price alone accounts for 14.8% of all negative evaluations. Reliability, portability, availability and maintenance are also criticized far above the 7.3% average. Style, versatility and finish are almost never criticized.
- Price5.9%
- Durability4.7%
- Comfort2.3%
- Value for money2.2%
- Ease of use2.0%
- Reliability1.9%
- Portability1.8%
- Style1.8%
- Quality1.7%
- Build quality1.6%
- Price18.2%
- Maintenance15.3%
- Availability15.0%
- Weight13.4%
- Portability12.5%
- Reliability12.4%
- Battery11.3%
- Customer support9.9%
- Durability9.6%
- Compatibility6.5%
New-tech and baby categories draw the most skepticism
Among 126 categories with at least 400 sampled evaluations, the average negative rate is 7.3%. Emerging hardware (humanoid robots, AR and smart glasses) and baby gear (monitors, strollers, breast pumps) are judged hardest. Beauty and apparel are judged most gently.
- Humanoid robots20.8%
- AR glasses16.7%
- Fitness trackers16.7%
- Smart home hubs16.3%
- Smart glasses15.6%
- Baby monitors14.4%
- Strollers14.1%
- Breast pumps13.7%
- Eyeliner1.0%
- Hair styling products1.1%
- Lip liner2.0%
- Bodysuits2.3%
- Shorts2.4%
- Concealer2.5%
- Pants2.6%
- Custom jewelry2.7%
The top 1% of brands get 65% of AI mentions
Across 40.96 million brand mentions, attention is extremely concentrated. The top 1% of brands (about 3,100) receive 65.1% of all mentions; the top 10% receive 93.8%; the bottom half of brands share 0.7%. Inside a typical category the pattern repeats: ChatGPT names a median of 245 brands per category, yet the top 3 take 36.5% of mentions and the top 10 take 65.0% (2,458 categories).
- Top 1% of brands65.1%
- Top 10% of brands93.8%
- Bottom 50% of brands0.7%
- Top 3 within a category (median)36.5%
Engines disagree on the #1 brand in 4 of 10 categories
We compared the five most-mentioned brands for 1,541 topics tracked on all three engines over the same five weeks. The engines share most of their shortlist, but the brand they put first often differs: ChatGPT and Google AI Mode agree on the #1 brand in 58.9% of topics, ChatGPT and AI Overviews in 53.8%. Even Google’s two surfaces disagree a third of the time.
| Engine pair | Same #1 brand | Shared top-5 brands | Topics |
|---|---|---|---|
| ChatGPT × Google AI Mode | 58.9% | 3.51 / 5 | 1,541 |
| ChatGPT × Google AI Overviews | 53.8% | 3.35 / 5 | 1,541 |
| Google AI Mode × Google AI Overviews | 66.8% | 3.91 / 5 | 1,541 |
What brands should do with this
1. Get named first, then worry about tone
Nine in ten AI remarks are positive, and the bottom half of brands share 0.7% of mentions. For most brands the gap is presence, not reputation. Start by finding the prompts where competitors are named and you are not.
2. Make your price easy to justify
Price drives 14.8% of all negative remarks. Publish clear pricing, what is included, and how you compare on value, so engines have a source to quote other than a complaint thread.
3. Back durability and reliability with evidence
Reliability (12.4% negative) and durability (9.6%) are judged from what others say. Warranty terms, test results and third-party reviews give engines something concrete to cite.
4. Track every engine separately
ChatGPT and Google AI Mode pick a different #1 brand in about 4 of 10 topics. Winning one engine says little about the others.
See what AI says about your brand
GEOly runs the same measurement for your brand and competitors across ChatGPT, Google AI Mode, Gemini, Google AI Overviews and Microsoft Copilot — mentions, position, citations and the attributes AI praises or criticizes.
Methodology
- Prompts: 26,557 shopping and product-research topics, 98% of them US English, run weekly between May 22 and October 10, 2026. Most volume falls between May 22 and September 4.
- Engines: ChatGPT, Google AI Mode and Google AI Overviews. GEOly also tracks Gemini and Microsoft Copilot; they are not part of this dataset. Perplexity, Claude and Grok are not tracked.
- Evaluations: each brand mention in an answer is split into attribute-level remarks (for example "battery: negative"), each with a polarity and an evidence quote. The full table holds 35.0 million evaluations.
- Sampling: sentiment, attribute and category figures come from a reproducible 1% block sample (345,264 evaluations). For a category with 400 sampled evaluations and a 10% negative rate, the 95% margin of error is about ±3 points. Concentration and cross-engine figures are exact counts over the full mention table.
- Limits: AI answers vary between runs; this is a description of what engines said in these weeks, not a ranking of product quality. Brand-level results are not published in this study.
Download the data
All chart data on this page is available as CSV. Free to reuse with attribution and a link to this page.
- Sentiment by engine (CSV)
- Attributes (CSV)
- Categories (CSV)
- Mention concentration (CSV)
- Engine agreement (CSV)
How to cite this study
GEOly, "How AI answers talk about brands: 35 million evaluations across 311,339 brands", published 2026-11-02, https://www.geoly.ai/research/ai-brand-perception