Howseen data · 2026
359 Brands Tested in AI Search: 85.5% Were Never Named (2026 Data)
We pooled the first Howseen scan of 359 brands: 6,923 answers from ChatGPT, Gemini and Google's AI to 2,343 buyer questions. Most brands are invisible at the start, and the ones AI names usually have their own site cited.
3-day free trial, cancel in 1 click. Trial: 3 days, plans from €39 per month. Cancel anytime from your settings.
Summary
How often does AI name a brand on its buyer questions?
Not often. On their first scan, 307 of 359 brands (85.5%) were not named in a single answer to the questions their buyers ask, and only 2.9% of all answers named the brand.
Brands
359
first scan, Aug 26 to Oct 20, 2026
AI answers
6,923
to 2,343 buyer questions
Never named
85.5%
307 of 359 brands
Named in ChatGPT
11.3%
36 of 319 brands
Named in Gemini
8.5%
29 of 340 brands
- Most brands start invisible. 85.5% were never named. Among the 148 brands with at least 10 answers, it is still 79.1%.
- ChatGPT names slightly more brands than Gemini: 11.3% against 8.5% of brands named at least once.
- Brands named by AI usually have their own site cited. 89.5% of brands whose site was cited were named, against 5.6% of the others.
- AI answers lean on sources. 81.1% of answers cite at least one link.
By engine
Does ChatGPT name brands more often than Gemini?
Slightly. On the first scan, ChatGPT named 11.3% of brands at least once and Gemini 8.5%. Google's AI surfaces and Perplexity run only on paid plans, so their samples are small.
See the data
| Engine | Brands with answers | Named at least once | Share |
|---|---|---|---|
| ChatGPT | 319 | 36 | 11.3% |
| Gemini | 340 | 29 | 8.5% |
| All engines | 359 | 52 | 14.5% |
Google AI Overviews (50 brands), Google AI Mode (22) and Perplexity (22) are tracked on paid plans only, so we do not headline them here. Google warns that its two AI surfaces do not behave the same way:
"AI Mode and AI Overviews may use different models and techniques, so the set of responses and links they show will vary."
The pattern
Are brands whose site AI cites named more often?
Much more often. 34 of the 38 brands whose own site was cited in at least one answer were also named (89.5%), against 18 of 321 brands whose site was never cited (5.6%).
Only 10.6% of brands had their own site cited on the first scan. This is a correlation: an answer that names a brand often links to it, so the two move together. It still points to the practical lever. AI engines quote pages that answer the buyer's question directly, and research on generative engines found that the way a page answers changes how often it is cited (Aggarwal et al., GEO, KDD 2024).
Sources
Which sources do AI engines cite for these brands' questions?
Official and public sites lead. Among the 5,613 answers that cite a source, government sites appear in 14.3%, YouTube in 7.4% and Reddit in 3.3%.
See the data
| Type of site | Share of answers |
|---|---|
| Government sites | 14.3% |
| YouTube | 7.4% |
| Maps, directories, reviews | 4.9% |
| Social networks | 4.7% |
| 3.3% | |
| Wikipedia | 0.6% |
Government sites rank high because many of these brands are French trades and services, where buyers ask about rules, aid and permits. Maps, directories and review sites show that local listings feed AI answers too. For a source-by-source view per engine, see our study of 5,331 AI answers and who they cite.
Are you in the 85.5%?
Howseen runs your buyer questions in ChatGPT, Gemini and Google's AI, shows who is named instead of you and which pages AI cites. 3-day free trial, cancel in 1 click.
What to do
How do you get recommended by ChatGPT, based on this data?
Start where the data points: get a page of your own site cited on your buyers' questions, show up on the sources AI already trusts in your niche, then measure each engine separately.
- Answer each buyer question on your own site, one page per question, with the answer in the first lines. See how to write content AI cites.
- Be present where AI looks: listings, review sites, videos and forums of your niche. See how to find the sources AI cites.
- Make your site readable by AI crawlers, and check it with the free AI crawler checker.
- Measure the visits too: track AI traffic in GA4 with one regex. Keep in mind that many people read the answer without clicking (Pew Research Center, July 2025).
The full method is in the free Howseen Academy, and the background in how AI chooses brands.
Method
How we measured it, and the limits
We took the first completed scan of each brand on Howseen, kept the buyer questions that do not contain the brand name, and counted the answers that name the brand. Data read on October 21, 2026.
- Sample: 359 brands whose first scan ran between August 26 and October 20, 2026; 6,923 answers to 2,343 questions. Howseen's own test accounts are excluded. 9 brands with only branded questions are left out.
- Who: mostly French small and mid-sized businesses (325 of 368 brands set French as their language), so the results describe that market best.
- Engines: ChatGPT and Gemini answers are collected with web search on; Google AI Overviews, AI Mode and Perplexity only on paid plans.
- Named means the answer mentions the brand. Own site cited means a source link contains the brand's domain.
- Limits: one scan per brand, a median of 5 questions each, and answers change from one run to the next. Correlations do not prove cause. No brand is named or identifiable on this page.
FAQ
Questions, answered
Short answers to the questions we get most about this topic, with the same numbers and sources as the page above.
How often does ChatGPT name a brand on its buyer questions?
Rarely, at first. Of 319 brands with ChatGPT answers on their first Howseen scan, 36 (11.3%) were named at least once. Across all engines, 307 of 359 brands (85.5%) were never named.
Does having your own site cited help?
In our data the two go together: 34 of 38 brands whose own site was cited were also named (89.5%), against 18 of 321 brands whose site was never cited (5.6%). It is a correlation, not proof of cause.
Which sources do AI engines cite in these answers?
81.1% of answers cite at least one source. Among those, government sites appear in 14.3%, YouTube in 7.4%, maps, directories and review sites in 4.9%, social networks in 4.7% and Reddit in 3.3%.
Which brands are in the study?
Brands that ran a first scan on Howseen between August 26 and October 20, 2026, most of them French small and mid-sized businesses. No brand is named on this page: results are pooled, and Howseen's own test accounts are excluded.
Why only the first scan?
The first scan shows where a brand starts, before any GEO work. Only 4 brands had a later scan more than 14 days after the first, too few to measure progress, so we do not report a before and after.