Not crawl volume, not a blended visibility score. Which companies get named, on which engine, in what order, on the questions your buyers actually ask. 9.8 million brand mentions recorded across 1.30 million companies, and every figure on this page is read live from that database.
| Brand mentions recorded | 9.8M |
| Companies tracked | 1.30M |
| Mentions in the last 30 days | 1.7M |
| Engines, reported separately | 4 |
Read from the live counter, not typed into this page. If it says something different tomorrow, that is the point.
This is the finding Rankxa exists to surface. Put the same buying question to all four and the companies they name barely overlap, which makes any single engine reading, or any blended average of the four, close to meaningless.
| Engine pair | Prompts compared | Brand overlap |
|---|---|---|
| Claude vs ChatGPT | 52,756 | 10.6% |
| Gemini vs ChatGPT | 25,061 | 10.9% |
| Claude vs Gemini | 34,261 | 11.2% |
| Google AI Overviews vs ChatGPT | 1,273 | 1.0% |
| Google AI Overviews vs Claude | 1,301 | 0.8% |
| Google AI Overviews vs Gemini | 1,220 | 1.0% |
Rankxa production data, 30 days to 11 September 2026. Gemini and Google AI Overviews are both Google products and agree with each other no more than either agrees with a competitor model.
Every account runs the same sequence, and the first step is the one most tools skip entirely.
Reads your robots.txt the way each engine reads it and records a verdict per crawler. A site that disallows OAI-SearchBot cannot appear in ChatGPT search answers at all, so every other number is meaningless until that is clear.
The questions your buyers ask before they shortlist, not keyword volume. Built from your category, your competitors and the phrasing people actually use.
Every prompt against ChatGPT, Claude, Gemini and Google AI Overviews. Generative answers vary between identical asks, so a single run is never reported as a result.
Which companies each answer names, and where in the answer they appear. Position is kept as its own field rather than folded into a score that hides it.
Three kinds of AI bot visit your site and they mean completely different things. Tools that add them into one AI traffic number are selling you a figure that predicts nothing.
| Crawler | Fires when | Predicts citation |
|---|---|---|
| GPTBot, ClaudeBot | Bulk training crawl | No |
| OAI-SearchBot, Claude-SearchBot | Search index membership | Eligibility gate only |
| ChatGPT-User, Claude-User | A live human question pulls your page | The closest real signal |
Publishers already treat these differently. Among the top 10,000 sites, 17.6% block GPTBot but only 6.6% block OAI-SearchBot: they are refusing training while permitting answering.
Across 560,695 AI crawl requests, page count alone explains 73% of the variation in crawl attention. Across 75,000 brands, total page count correlates just 0.17 with actual AI visibility, while branded web mentions correlate 0.664. The metric that is easiest to buy is the one that works least. Rankxa reports the second one.
Four engines, four columns, never averaged.
Built for groups, franchises and agencies running several brands.
Who actually appears on your questions, not who you think competes.
The same data, in public, on commercial questions.
This is not run by a black box of tools. It is a team of Growth Architects who plan, build and optimize it, using AI where it accelerates the work and human judgment where it actually matters. Different specialists, one accountable team.
Meet the team →
Michael · Engineering Lead
Amal · Automation Architect
Bryn · Solutions Engineer
If we can't increase your revenue by the 6 month mark, our team works for free until we do.
“BusySeed created and managed our social media and wrote blogs that raised our SEO to a new level. We are getting new clients from organic search, and their lead generation campaigns made it easy to justify the investment by focusing on ROI.”
“BusySeed demonstrated tremendous ability in understanding our needs and taking us by the hand in the process of designing a digital strategy from scratch.”
“Great professionals. They manage our company's social media and designed our website. I highly recommend them for social media management and web design.”
Generative engine optimization, or GEO, is the work of getting your company named inside the answers AI assistants give, rather than ranked in a list of links. It has three parts: making sure the engines can retrieve your pages at answer time, publishing pages a model can lift a clean answer out of, and earning mentions of your company by name on independent sites. It is measured per engine, because the engines disagree with each other far more than most people expect.
A rank tracker asks where a page sits in an ordered list of ten blue links. Rankxa asks whether an answer names your company at all, how often across repeated asks, and how early in the answer you appear. Those are different measurements because the underlying systems are different. Ahrefs found only about 8% of ChatGPT citations rank in Google top ten for the same prompt, so a rank tracker cannot tell you what an assistant will say.
ChatGPT, Claude, Gemini and Google AI Overviews, each tracked and reported separately. It never blends them into one score. In our production data the three chat engines agree with each other about 11% of the time, and Google AI Overviews agrees with any of them around 1% of the time, so a blended figure describes a situation no company is actually in.
It means no engine is treated as the default and none is weighted above another. Some tools measure ChatGPT and infer the rest; Rankxa runs every prompt against all four and reports four separate results. That matters because the engines retrieve from different places. Google AI Overviews names about 3.3 brands per answer against 6.1 to 7.2 on the chat engines, so the same absence means different things on different surfaces.
Yes. Multi-domain tracking is built in, which matters for holding companies, franchise groups and agencies running several client brands at once. Each domain gets its own prompt set, its own competitor set and its own per engine reporting, and they can be compared side by side without mixing the data.
Every tracked prompt reruns weekly, and generative answers vary between identical asks so a single run is never reported as a result. Agencies can resell it under their own brand: see our generative SEO for agencies page. We built Rankxa for our own clients and still run it ourselves, so you are buying from the team that maintains it rather than reselling someone else's license.
Book a free 15 minute session and we will run your buying questions through all four engines before the call, so there is something real on the screen when we start.
