We ranked 139 brand sites by traffic
Does ChatGPT only recommend the big names? We checked by putting each tracked brand's website on a public popularity ranking and counting how often the engines named it when a buyer asked for options.
The sample is every answer from ChatGPT, Claude, Perplexity and Google AI Overviews collected from October 4 to November 2, 2026, for a buyer question that doesn't name the tracked brand: 25,504 answers to 3,141 questions for 139 brand websites in 116 accounts, each with at least 20 answers. We dropped repeated identical answers.
For size we used the Tranco list generated on November 2, 2026, a research ranking of the top million websites that combines five traffic and popularity lists over the same 30 days.
The top 10,000 sites were named in 53% of answers. Sites outside the million, 3%.
Share of answers that named the brand, by the brand's Tranco rank (95% intervals):
- Top 10,000: 7 sites, 53.1% (50.3% to 56.0%).
- 10,001 to 100,000: 8 sites, 35.9% (32.9% to 39.1%).
- 100,001 to 1,000,000: 16 sites, 21.5% (20.1% to 23.1%).
- Not in the top million: 108 sites, 3.4% (3.2% to 3.7%). The median site in this group was named in 1.2% of its answers.
Every engine was harder on the small sites
For the 108 sites outside the top million, ChatGPT named them in 4.9% of answers, Perplexity 3.4%, Claude 3.0% and Google AI Overviews 2.2%. Even ChatGPT, the kindest of the four, named them less than a tenth as often as it named the top 10,000 sites (57.2%).
Ahrefs found the same pull toward size in May 2025: across 75,000 brands, mentions of the brand across the web had a 0.664 rank correlation with AI Overview mentions, against 0.218 for backlinks. But that study kept only sites with an Ahrefs Domain Rating above 40 and a top keyword searched at least 800 times a month, then dropped the 26% with no AI Overview mentions. Most small companies never got into it. In our sample they're 108 of 139 sites.
Small sites won a few questions, not the category
44 of the 108 small sites were never named at all. The other 64 were named 703 times, and 478 of those (68%) came from each site's own top three questions. Only 300 of their 2,439 questions named them even once.
So the small brands that got named didn't get a little of everything. They owned a corner. 31 of the 64 had at least one question, asked four or more times, where they were named in half the answers or more.
The questions they won were narrow
We read 15 randomly drawn questions where a small site was named in at least half the answers. Every one asked for a job plus a constraint: a city or country, a language, a buyer type, or two features together. None was a plain "best X".
An illustrative pair, not from our data: "best accounting software" gets the category leaders. "Accounting software for a two-person bakery that takes card payments in Kenya" leaves room for a company that does exactly that.
What we'd do if your site is small
- Find the questions that already name you. If you have two or three, they show you what the engines believe you're for.
- Write the next questions around them: the same job with a nearby city, buyer or feature. Check which ones you win before you write pages for them.
- Don't measure yourself against the category leader on "best X". At 3% you can't tell a real change from noise without a lot of asks.
- Get named where the engine looks for your corner: the niche roundups, forums and directories it cites when it answers those narrow questions.
How Proofsource handles it
Proofsource asks each buyer question on every engine on a schedule and records, for each answer, whether it named you. So you can see which questions you already win, and on which engine, instead of one blended rate.
Where this could be wrong
Website traffic isn't company size. A business that sells offline or through an app can be big with a small site. Tranco ranks whole domains, so a brand on a subdomain of a site builder counts as unranked.
This is a correlation. Bigger sites are also older and have more press and reviews, and we can't say which of those does the work.
We write each brand's questions from its own site, so a narrow question can describe a brand's exact pitch, which flatters the small sites' wins. The top groups are small (7, 8 and 16 sites) and include a few household names. It's one month, mostly English questions, and brands our users chose to track.
