ResearchNovember 5, 2026 · 5 min read
We checked 20 brands in ChatGPT and Claude. Here's who gets named
100 buyer questions, 200 AI answers, 20 brands from Wikipedia to a coffee roaster. What separated the brands AI recommends from the ones it skips.
Key takeaways
- Scores ran from 95 (Wikipedia) to 7 (Blue Bottle Coffee); a made-up site scored 0, as it should.
- Being famous isn't enough: X scored 20 because buyers' questions about social media led to other platforms.
- Niche tools lose to the category default: Tally was named in 1 of 10 answers, while Typeform, Google Forms and Jotform were named again and again.
- The answers lean on the same kinds of sources: comparison posts, review sites and community threads. That's where to work.
In early November 2026 we ran a live test of the free AI visibility checker on 20 websites. Each check reads the site's homepage, writes 5 questions a buyer in that category would ask, and puts every question to ChatGPT and Claude with web search on. That's 10 answers per brand and 200 answers in total.
We picked the 20 sites to be different from each other on purpose: developer tools, consumer apps, a shoe brand, a coffee roaster, a design agency, a German HR platform, two of the biggest sites on the internet, and one website that doesn't exist. The point was to see how the checker behaves on messy, real inputs. The results also say a lot about how AI assistants pick what to recommend.
The scores
The score is the position-weighted share of answers that name the brand, from 0 to 100 (being named first in every answer is 100). How the score works explains the weighting. Here are all 20, highest first.
| Website | Category the checker used | Score | Named in |
|---|---|---|---|
| wikipedia.org | free online encyclopedia | 95 | 10/10 |
| excalidraw.com | collaborative whiteboard tool | 86 | 9/10 |
| duolingo.com | free language learning app | 78.5 | 8/10 |
| stripe.com | payment API for marketplaces | 78.5 | 8/10 |
| linear.app | project management software | 70 | 9/10 |
| grammarly.com | AI writing assistant | 60 | 6/10 |
| allbirds.com | comfortable sustainable shoes | 59.5 | 7/10 |
| ouraring.com | smart ring for sleep tracking | 53.5 | 6/10 |
| notion.so | AI workspace for teams | 51 | 6/10 |
| plausible.io | privacy-friendly web analytics | 48.5 | 5/10 |
| blog.hubspot.com | marketing blog | 36 | 5/10 |
| screenshotone.com | screenshot API | 35.5 | 4/10 |
| cal.com | scheduling software | 33 | 4/10 |
| posthog.com | product analytics platform | 27 | 4/10 |
| personio.de | software product (generic guess) | 25 | 3/10 |
| x.com | social media platform | 20 | 2/10 |
| pentagram.com | design consultancy | 10 | 1/10 |
| tally.so | free online form builder | 8.5 | 1/10 |
| bluebottlecoffee.com | specialty coffee online | 7 | 1/10 |
| a made-up domain | CRM for freelancers | 0 | 0/10 |
The made-up domain scoring 0 is the sanity check we wanted: when a brand doesn't exist, no engine invents it.
Finding 1: owning the category name wins
The top of the table is brands that are close to synonyms for what they do. Ask for "a free online encyclopedia" and Wikipedia is the answer. Ask for "a collaborative whiteboard" and Excalidraw comes up in 9 of 10 answers. Duolingo is the default answer to "best free app to learn a language".
When a brand is the category in people's heads, it's also the category in the text the models learned from and in the pages web search returns. That's hard to copy, but the lesson for smaller brands is useful: pick a narrow category you can actually be the default for, and use those exact words on your site and everywhere you're listed.
Finding 2: being famous isn't the same as being recommended
X is one of the most visited sites in the world, and it scored 20. The checker asked buyer-style questions about "social media platforms", and the answers named Reddit, TikTok and YouTube more often. Wikipedia scored 95 on the same day.
The difference is the question. AI visibility isn't a measure of fame; it's whether you come up for the specific questions your buyers ask. A brand can be household-name famous and still lose "best platform for X" questions to rivals that fit the question better.
Finding 3: niche tools lose to the category default
Tally is a well-liked form builder with a generous free plan. It was named in 1 of 10 answers. Typeform, Google Forms and Jotform were named again and again instead. Cal.com, a strong open-source scheduler, was named in 4 of 10, behind Microsoft Bookings, Chili Piper and HubSpot Meetings in the answers that skipped it.
This pattern showed up across the smaller brands. When a buyer asks a broad question ("best free online form builder"), the answer leans on the brands that the most sources agree on. Smaller brands appear when the question is specific enough that they're the best fit, like "form builder that feels like a doc" for Tally. That's a strong argument for tracking the specific, long questions your best customers actually ask, not just the broad head term.
Finding 4: physical products and services are harder
Blue Bottle Coffee scored 7 and the design agency Pentagram scored 10. For "specialty coffee online", the answers named subscription services like Trade Coffee and Atlas Coffee Club and other roasters. For a design consultancy, answers named platforms and award bodies more than individual studios.
When the buyer's question is about a type of service, AI answers tend to name marketplaces, directories and lists rather than individual providers. For businesses like these, the realistic goal is often to be well represented on those lists and directories, because that's what gets cited.
Finding 5: the category you're judged in matters
personio.de was the one site where the checker couldn't tell the category from the homepage and fell back to "software product". The questions came out broad, and Personio was named in 3 of 10 answers, behind Gusto, Rippling and BambooHR. With "HR software for European SMBs" the questions would have been far closer to how Personio's buyers search. That's why the checker shows its guess and lets you edit it before running.
What the answers cited
Across the 200 answers, the sources the engines cited fell into a few groups again and again:
- Comparison and "best of" articles on blogs and publications.
- Review and listing sites, where products are compared side by side.
- Community threads, especially Reddit discussions where people recommend tools to each other.
- The vendors' own pages, mostly pricing, comparison and documentation pages.
If you want to show up more, those are the places to work on. Which websites AI engines cite goes through each type and what you can do about it.
How to run the same test on your brand
The free checker runs exactly this test on any website in about a minute, with no signup. You'll get your score, your rank in each answer, the rivals named instead and the sources each engine cited. Because one run is a snapshot, the useful next step is to track the same questions every day and watch the trend. That's what Citemark's paid plans do.