Free to get started · no cardWhen AI is asked, does it say your name?
More and more customers stop at the AI answer. They ask ChatGPT who to buy from and go with the names it gives. So there is a simple question worth knowing the answer to: is your business one of those names?
Tell the checker what you do. It writes the questions your customers would really type, asks each one several times over, and counts how often you get named — and who gets named in your place.
Step 1 of 4
Tell us about your business
Two things: the name to look for in the answers, and what you actually do. The description is only used to write the questions — the AI answering them is never told about you, or the whole thing would be rigged.
A few sentences in plain words — the words a customer would use, not your website copy.
Your description, your questions and your results stay yours. We never show them to anyone else.
How it works
Four steps to a number you can act on
Describe your business
A few sentences: what you sell, and who buys it. Add your website and any other name you go by, so every way you might be mentioned gets counted.
The AI writes the questions
Not questions about you — questions about your market, in the words a customer would use. "Best AI coding course for working professionals", not "is your company good". You can edit every one before anything is spent.
Each question is asked several times
An AI does not answer the same way twice. Asking once tells you almost nothing; asking ten times tells you whether you are a regular answer, a sometimes answer, or not an answer at all.
Every mention is counted
Your name, your website, your other names. You get the count, the share, and the exact sentence that mentioned you — so you can check the number rather than trust it.
What comes back
A score, a league table, and the proof
The point is not the score. The point is knowing which questions you lose, to whom, and in the AI's own words.
Your visibility score
How many of the answers named you, out of how many were asked. One number, on the same scale every time, so you can run this again next month and see whether the work you did moved it.
Who gets named instead
Add your competitors and they are counted from the same answers, at no extra cost. A score of 20% means one thing if the leader is on 25% and something very different if they are on 90%.
Question by question, with proof
Which questions you show up for and which you never do. Each one comes with the sentence the AI actually wrote, so nothing here is a number you have to take on faith.
Being straight with you
What this measures, and what it does not
Plenty of tools in this space imply they are watching ChatGPT over your customer's shoulder. None of them are, and neither are we. Here is exactly what the numbers in front of you are.
What the AI already knows
Ask in this mode and the AI answers from what it has read over the years, with no searching. When a model has seen enough about your business across the internet, it names you from memory. That is the strongest kind of AI visibility there is, and the slowest to build.
What the AI finds when it looks
Ask in this mode and the AI uses its own web search — the same tool ChatGPT uses when it looks something up. It runs the searches, reads the pages and shows you which ones it used. This is where a page you published last month can win you a mention.
What we still cannot show you
This is not somebody’s real ChatGPT session. It is questions we wrote, asked through the same AI models and the same search tool, which is as close as anyone outside OpenAI can get. What it cannot tell you is whether your actual buyers are asking these exact questions — so read the questions, and change the ones that do not sound like your market.
Both scores matter, and they are not the same job. A page you publish today can be found by a web search tomorrow, so that number moves in weeks. Getting into what the model itself knows takes being written about, listed and compared in enough places that it learns you belong in the answer — slower, but it is the visibility that keeps working when nobody clicks a link at all. Running both is how you find out which one is holding you back.
What a pack buys
How many checks each pack gives you
Tokens are the only thing you buy, and they work on every tool here. This is what they turn into on this one — pick a model down the side, read across to the pack.
Questions
Before you run a check
What is answer engine optimization (AEO)?
Search engine optimization is the work of getting your page to rank on Google. Answer engine optimization — AEO, sometimes called GEO — is the same idea for AI assistants: the work of getting named in the answer when someone asks ChatGPT, Gemini or Perplexity who to buy from. The difference matters, because an AI answer usually names two or three businesses and the customer never scrolls a list of ten. You are either in the answer or you are not in the conversation.
How does this checker actually work?
Three steps. You describe your business. An AI writes the questions your customers would ask when they are looking for what you sell — never naming you, because a question with your name in it would score a mention every time and measure nothing. Then each question is sent to an AI model several times over, and every answer is checked for your name, your website and any other name you gave us. What comes back is the count, the share, and the sentences.
Why ask the same question more than once?
Because the answer changes. An AI model does not reply identically twice, so a single answer is one roll of the dice. If you are named in three answers out of ten, you are on the edge of what the model recommends — which is genuinely different from being named every time, and different again from never being named. Asking once could show you any of those three and you would not know which. The repeats are what turn an anecdote into a number.
Does this show me what ChatGPT would say right now?
You choose, because there are two answers to that. Ask from the AI’s own knowledge and it answers straight from what it has read over the years — no searching. That is the visibility that is hardest to buy and hardest to lose. Ask with web search and the AI goes and searches, reads what it finds and answers from that — the same thing ChatGPT does when it looks something up, using the same search tool. You can also run both and get a score for each.
Which one should I pick?
Run both the first time. On their own, either number is easy to read the wrong way: if the AI never names you from memory, that looks like a disaster until you find out live search names you constantly. And if you only ever win on search, that looks safe until a ranking change takes it away. Side by side the two say which of those is true for you — and they need different work. Being remembered is earned by other people writing about you. Being found is earned by your own pages ranking and being easy to quote.
Is the web search the same one ChatGPT uses?
Yes. We do not run our own search and feed it in — we hand the question to OpenAI with its web search switched on, and the AI does the searching itself, the same way it does inside ChatGPT. It writes its own search queries, reads the pages, and cites them. The report shows you both: the pages it used, and what it actually typed into the search box. And if it answers without searching, we throw that answer away rather than count it as a search result — you are not charged for it.
Why does the web search cost so much more?
Because OpenAI charges for it. Every search has a fee, and the AI is also charged for reading the pages it finds — which is far more text than a plain question. One searched question works out around nine times a plain one. We pass that through rather than marking it up, and you see the exact number before you start anything. If it looks steep, run fewer repeats: two or three still tells you a lot.
Why can I only use some AI models for the web search?
OpenAI only allows its search tool on some of its models, and the cheapest one is not among them. So when you pick a searching check, the models that cannot search are greyed out and we move you to one that can. Switch back to the AI’s own knowledge and every model is available again.
Which AI models can I test with?
You pick one before you start, and the report says which one answered. Right now these are OpenAI models — the same family behind ChatGPT — with the cheapest one as the default so you never pay more by accident. A stronger model reads and reasons more carefully, and costs more tokens.
Is the AI visibility checker free?
It is free to get started, which is not the same as free. Every question asked is a paid AI call, so the checker runs on tokens — and you get a batch of tokens free when you make an account, with no card needed. That covers a couple of real checks. After that you buy a pack of tokens once; there is no monthly fee, and the same tokens work on every tool on the site.
How much does one check cost?
It depends on the size, and you are shown the number before you commit to anything. The cost is simply the number of questions multiplied by the number of repeats, plus one call to write the questions. Five questions asked three times is sixteen calls. What you see quoted is the most it can cost — you are charged only for what the AI really uses, and the rest goes back to your balance.
My score is low. What do I do about it?
Low is the normal starting point, so it is not a verdict. Look at the question-by-question table first: the questions where a competitor is named and you are not are the ones to work on, and they tell you what the market is actually asking for. From there the work is the unglamorous kind — being written about somewhere other than your own website, being listed where your category gets compared, and having pages that answer those exact questions plainly. Then run the check again in a month and see whether the number moved.
Why does it need my competitors?
It does not — you can leave that empty. But it is the most useful field on the form, and it costs nothing: the same answers are already being read, so counting a few more names is free. A score on its own is hard to judge. A score next to the businesses you lose deals to is a to-do list.
What happens to what I type in?
It stays yours. Your description, your questions and your results are never shown to other users and never sold, and you can delete any check at any time.
Your tokens work on every tool on the site — see /ai-tools.
What we actually do
This tool is the small version. We build the whole thing.
The visibility checker does one job, for free, for anyone. Our actual work is building AI systems that run inside a business — the automations, the internal tools your team uses every day, and the parts that talk to the software you already have.
Not sure what is worth automating?
Write one paragraph about your business. We read it, ask a few questions, and show you which work a computer can do for you — and which work it should not touch. Free, and there is nothing to install.
About five minutes
Results by email
No sales calls
Already know what you want built?
Tell us in one line. We will say whether we can build it, what it would take, and we say no if it is not a fit.
Same people who built the tool you just used.