Does AI know your business?
Enter your brand and domain. A large language model writes three questions your buyers might ask, answers them from what it absorbed during training, and we show you whether your name came up. It is a check of trained knowledge, not a live query of any AI search product. No account needed.
What this checks
One question, asked properly: is your business part of what a large language model already knows? That matters because a model answers a great many buyer questions from memory, and a brand it has never absorbed cannot be recalled no matter how the question is phrased.
- Three buyer questions — the model writes them itself for your category, using the industry you give it as a hint. They are the kind of question someone asks before they know which company they want.
- A verdict on each — the model answers privately from trained knowledge, and what comes back to you is whether your brand appeared, with the explicitly told it has no web access and must not pretend otherwise.
- Whether you were mentioned — the model reports this for each answer, and quotes the phrase where your name appears so you can see the context it put you in.
- A mention rate — the share of the three answers that named you, shown as a percentage.
What this is not
This is the part most tools in this category leave out, so here it is plainly. This check does not query ChatGPT. It does not query Perplexity, Gemini, Copilot or Google AI Overviews. Nothing is sent to any of them and no result here tells you what any of them would say today.
It does not fetch your website either. It is not live, it is not monitoring, and it keeps no history, so there is nothing to watch move week to week. And three questions put to one model on one occasion is far too small a sample to carry a statistical reading — a score that shifts from 33 to 67 between two runs has told you nothing.
If what you want is live rank tracking across several AI answer engines, that is a different category of product and you should buy one of those instead; the large SEO suites all offer a version. We are answering the question underneath it: does the model know you exist at all?
How to read your results
Zero mentions
The clearest signal the tool produces. Your brand is either absent from the model's trained knowledge or too thinly represented to surface for an ordinary category question. For most small and mid-sized businesses this is the normal result, and it is the one worth acting on.
One or two out of three
You are in there. Which questions you came up for is more informative than the number: a mention on a narrow, specific question and silence on the broad one usually means the model associates you with a niche rather than the category.
Three out of three
Well known within this model's knowledge, at least for the questions it chose to write. Run it again with a different industry hint before you believe it, since the model picks easier questions for some framings than others.
The quoted phrase
Read what the model said about you, not just that it said something. Being recalled as the wrong kind of company is a different problem from not being recalled at all, and it needs different work.
What to do about a low score
- Accept the timescale first. A model's trained knowledge is fixed when it is trained. Nothing you publish this month changes the answers it gives; it changes what a later model absorbs.
- Describe yourself plainly somewhere machine-readable. What you do, who for, where. Brochure language that never states the category is the most common reason a business is unrecognisable to a model.
- Get named on pages that are not yours. Trade press, directories, supplier lists, industry roundups, forums. Models learn brands from the web at large, not from your homepage.
- Answer the buyer questions directly. A page that answers the actual question, in the words it is asked, is far easier to absorb than one that circles it.
- Mark up who you are. Organisation and product schema state your name, category and location in a form that needs no interpretation. The schema markup checker shows what your pages currently declare.
The full picture sits in the guide to generative engine optimisation, which covers why models recall some brands and not others. If your pages are not yet saying clearly what they are for, start with the meta tag checker instead. The paid ClimbrIQ plan does this work continuously rather than listing it.
Frequently asked questions
- Does this check ChatGPT?
- No. We do not send anything to ChatGPT, Perplexity, Gemini, Copilot or Google AI Overviews, and nothing here reflects what those products would say today. What we do instead: we ask one large language model to write three questions a buyer in your category might ask, answer them from its own trained knowledge with no web access, and report whether your brand came up. It is a read on whether a model already knows your business, which is the thing you can actually influence over time.
- Do you look at my website?
- No. The domain you enter is passed to the model as text so it can tell brands with similar names apart. Nothing on your site is requested, read or crawled, so a result of zero mentions says nothing about your pages. If you want a check that genuinely reads the page you give it, the meta tag checker and the schema markup checker both fetch the URL.
- What does the score actually mean?
- It is the share of three answers in which your brand name appeared, so the only possible scores are 0, 33, 67 and 100. Three questions from one model on one occasion is a signal, not a measurement. Read the gap between zero and non-zero as meaningful and everything in between as noise.
- Why do I get different questions and a different score each run?
- The model writes fresh questions every time rather than working from a fixed list, and language models do not give identical answers to identical prompts. Two runs an hour apart can differ. No result history is kept and nothing you see here is stored against your brand, so this is a snapshot rather than a trend line. A hashed record of the request is logged, which is what enforces the hourly limit below.
- The answer mentioned us but got the details wrong. Why?
- Because the model is recalling rather than looking anything up. Trained knowledge is compressed, dated and occasionally confidently wrong, which is exactly why the mention itself is the useful part of the result and the surrounding sentences are not evidence of anything. Treat a mention as proof you are in there somewhere, not as a quote you could put on a slide.
- Is it free, and are there limits?
- The on-page result is free and needs no signup. There is a cap of 30 checks per hour from the same IP address, which is more than enough for a brand and its competitors. Entering your email gets you a longer written report going through each question, why your brand did or did not appear, and five specific actions to work on.