AI Visibility Audit
An audit tells you two things: whether AI engines name your company when your buyers ask for a recommendation, and which pages on the internet decide that answer. The result is only worth having if the same question was asked more than once. A single screenshot of an engine naming you proves nothing, because the next run may not. Ask any firm how many times they ran it.
Below is every check we run, in the order we run them. It is published in full because a method you cannot read is a method you cannot check. If you want to run part of it yourself, section four tells you how.
Have us run itWritten by Lourdes Paul Agilan, founder of GeoCited. Published .
What we check
Fourteen checks in three groups. The first group finds out where you stand. The second finds the pages that decide it. The third turns both into a list someone can actually work through.
Where you stand
The questions your buyers actually ask, each one carrying the place we found it. Reddit and industry forums, community Q and A, review site text, your own sales calls and support tickets, your published FAQs. If we cannot point at somewhere a person asked it, it does not go on the list.
Those questions sorted by intent, into buying, comparison, problem and research. You see the list before anything is run.
Every question run in a fresh session, with web search on and no chat history, because a logged in session with history will flatter you.
Every question run more than once, and reported as a ratio. Asked five times, competitor named five times, you named none. Not a screenshot.
Who was named instead of you, recorded question by question.
The same questions run across ChatGPT, Perplexity, Gemini and Claude, because they disagree. A page that decides the answer on one engine may not exist on another.
Which pages decide it
Every cited URL logged from every run, then tallied by site and by exact page across the whole set.
Those pages ranked by how often each one decides the answer, so you know which handful actually matter.
Why the winning pages win. We open them and look: published numbers instead of "contact us for case studies", a named author with a real bio, a directory profile, a forum answer nobody on your side wrote.
Your own site as the crawler reads it. Raw HTML, crawler access, structured data. Not as your browser renders it.
Your trust pages. What you do, who founded it, who the team is, what proof exists. These are the pages engines assemble a description of you from.
What to do about it
A list for your developer, in priority order, written so they can just do it.
A list of content and placements, each item naming the specific question it is meant to win.
For every item, which of the two supply lines it belongs to: something that goes on your site, or something that has to go on someone else's.
One rule governs both lists. If we cannot point at a specific question you are losing, it does not go on the list. No best practice items, no general hygiene, nothing included because it is usually a good idea.
What you get
A written report, and a call to walk through it.
The report names the questions, shows the ratio for each one, names who was named instead of you, and lists every page the engines cited. Then the two lists from checks twelve to fourteen, in priority order.
The prompt set stays with you. If you never work with us again you can hand it to anyone and they can run the same questions and get a comparable number. That is the point of publishing the method.
Three checks you can run yourself today
You do not need us for these, and doing them will tell you whether a full audit is worth your time.
Open a fresh browser session, signed out, with web search on. Ask the engine the question your best customer asked you last month, in their words rather than your category's words. Ask it five times, in five fresh sessions. Count how many times you were named.
Ask the same engine to name the best providers in your category, then click every citation it shows. Write down the domains. That short list, not your website, is where your next customer's answer comes from.
Search your own company name on its own. If the engine returns a different company, or nothing coherent, you have an entity problem rather than a content problem, and it is the cheaper of the two to fix.
What this audit will not tell you
It will not tell you how many people saw the answer. No engine publishes that, and anyone quoting you an impression figure for ChatGPT is modelling it.
It will not tell you what the answer will be next week. The sources move without notice, which is why the runs repeat rather than happen once.
It will not tell you that a fix is guaranteed to work. Some placements get cited and still do not get you named. When that happens the placement moves.
It is not a technical audit. Most sites we look at are technically fine and still invisible, because engines do not answer buying questions from vendor websites.
We ran it on ourselves first
The first company we audited was this one. The report scored GeoCited twenty two out of one hundred, and zero on the only number that counts, which is whether any engine named us. It is published with the scores intact.
Questions people ask
How do I check if ChatGPT recommends my business?
Ask it, in a fresh signed out session with web search on, using the words your buyer would use rather than your category's words. Then ask it four more times in four more fresh sessions. One run tells you nothing. Five runs tell you a ratio.
How do I get my business listed in ChatGPT?
There is no list. Nothing to submit to, no directory inside the product, no verification step. The engine answers from pages already on the internet, so the work is getting onto the pages it already reads in your category.
What does an AI visibility audit cost?
It depends on how many questions are in scope and how many engines you want covered. We quote it on a call rather than publishing a number, because a number without a scope is meaningless.
How long does it take?
The first run is quick. The useful part is the second and third run, because a single point in time cannot show you whether anything moved.
How often should it be re run?
On a schedule, not once. Sources drop out and appear without notice, and nobody tells you when a page that was deciding your category's answer stops being cited.
Is this the same as an SEO audit?
No. An SEO audit asks why you are not ranking. This asks whether you are being named, and by which sources. A site can rank well and still never be named, and that is the common case.
Can I run it myself?
Parts of it, yes, and section four above tells you which parts. What is hard to do alone is the tally: logging every cited URL across every run and every engine, then working out which handful of pages decide most of the answers.
Have us run it
Send three details and we will run the first pass and send the report back. There is no charge for it and no call required to receive it. If your category turns out not to be worth auditing yet, we will tell you that instead of sending a report padded out to look useful.
Would rather just email? Send the same three details to contact@geocited.io