Short answer
A proper audit measures four things: presence (how often you are named across a set of buyer prompts, and where you land inside the answer), accuracy (whether what the engine says about you is true), source set (which URLs get cited when you or a competitor is recommended), and readability (whether your site is server-rendered, carries clean structured data, and is not blocked at the CDN). Only that last item is on your website; most of what determines an AI recommendation lives on properties you do not own. Pricing splits three ways: monitoring software from a couple of tens of dollars to the high hundreds per month, a one-off boutique diagnostic in the mid to high four figures in USD delivered in one to three weeks, or the audit folded into month one of a retainer. Every project is scoped individually, so treat all of those as planning ranges rather than quotes.
An AI visibility audit is a structured check of how AI answer engines describe your business when a buyer asks them a question you should win. It answers three things: do you appear at all, is what the engines say about you accurate, and which sources are they pulling from. Then it turns that into a prioritized fix list. It is not a rank tracker with a new label, because AI answers do not have ranks. They have citations, and citations come from places most businesses have never audited.
We are a software and marketing studio that runs this baseline at the start of every marketing engagement, for clients in the US and in the Philippines. So we have a commercial interest here, and we will flag it where it matters. The 15-minute self-audit near the bottom is the honest version you can run today without hiring anyone.
What an AI visibility audit actually measures
Traditional SEO asks where you sit on a page of ten blue links. An AI visibility audit asks a different question: when the engine writes a paragraph of advice, are you in the paragraph?
Four things get measured.
Presence. Across a set of buyer prompts, how often are you named, and in what position within the answer. Being third in a list of five is a real result. Being absent is a real result too.
Accuracy. What the engine says about you when asked directly. Wrong service list, wrong location, a closed office still listed as open, a founder who left in 2023. Weak chatgpt brand visibility is often not an absence problem, it is a bad-source problem.
Source set. Which URLs the engine cites when it recommends you or a competitor. This is the most useful output of the whole exercise, because it converts a vague goal into a list of specific pages you need to exist on.
Readability. Whether your own site is machine-readable: server-rendered copy, clean structured data, an answer to the question in the first hundred words, crawler access that is not blocked at the CDN.
Notice that only the fourth item is on your website. Most of what determines an AI recommendation lives on properties you do not own.
The four checks inside a proper AI search audit
A credible ai search audit runs in a fixed sequence, and each step feeds the next.
1. Prompt set design. Twenty to forty prompts drawn from real buying behavior, not keyword tools: category questions ("best X for Y"), comparison questions, cost questions, and direct brand questions. Cost prompts matter most, because that is where transparent pages get cited and everyone else disappears.
2. Multi-engine sampling. The same prompts run across ChatGPT, Perplexity, Google AI Overviews, Claude, and Gemini, repeated over several days. Answers vary between runs, so a single screenshot proves nothing. Sampling gives you a rate, not an anecdote.
3. Source-gap analysis. For every prompt where a competitor won, list the sources cited. Sort them into buckets: your own site, directories, review platforms, press and roundups, community threads. The gaps between their buckets and yours are the work.
4. Technical and content review. Render check, schema check, answer-first structure, freshness signals, and whether your best information is trapped in a PDF, a JavaScript widget, or a gated form.
The output should be a scored baseline you can re-run in ninety days and a roadmap ordered by effort against impact. If a report ends at "you are not visible in ChatGPT", you bought a screenshot.
What an AI visibility audit should cost
Three tiers exist, and they are not really competing with each other.
Software. An ai visibility checker or geo audit dashboard runs monthly, from a couple of tens of dollars for a single-brand tracker up to the high hundreds for enterprise monitoring. These tools measure well and diagnose poorly. They tell you the score moved without telling you which of the eleven things you changed moved it.
A one-off audit. A standalone diagnostic from a boutique agency generally lands in the mid to high four figures in USD, delivered in one to three weeks. That is the tier most small and mid-sized businesses should buy first, because it produces a plan you can execute in-house.
Bundled into a retainer. Most agencies fold the audit into month one of an ongoing program, where retainers commonly start in the low four figures per month at the boutique and offshore end and move into five figures for larger scopes. Every project is scoped individually, so treat all of these as ranges for planning rather than quotes.
The honest warning: audits are cheap to fake. Ten prompts, five screenshots, a template deck. Ask any vendor for the prompt list and the raw source citations before you pay. If they will not share the method, the method is the product they are hiding.
Run a mini geo audit yourself in 15 minutes
This will not replace a full baseline, but it will tell you whether you need one.
- Write the three questions a customer asks before hiring you. Include one cost question.
- Run all three in ChatGPT and Perplexity, and in Google where an AI Overview appears. Record whether you are named and who is.
- Ask each engine: what do you know about [your business name]? Check every fact in the answer.
- In Perplexity, open the citations behind your top competitor's mention. Write down every domain. That list is your distribution to-do.
- Open your most important page and view the raw page source. If your headline and body copy are not in the HTML, no crawler is reading them.
If steps 3 and 5 come back clean and you still do not appear in steps 2 and 4, your problem is off-site corroboration, not your website. That is a distribution and reputation project, not a redesign.
When an audit is not worth buying
Skip it if your business does not have a searchable category, if you serve a market where buyers arrive entirely through referral, or if you already know your site is unreadable and unfixed. In the last case, fix the rendering and structured data first, then measure. An audit that documents problems you have already diagnosed is an expensive way to feel informed.
Run the 15-minute version this week. If the answers make you uncomfortable, we can run the full baseline and hand you the ninety-day roadmap that comes with it. Start a project →
Questions · 05
Frequently asked questions
How is this different from a rank tracker?
AI answers do not have ranks, they have citations. Traditional SEO asks where you sit on a page of ten blue links; this asks whether you appear in the paragraph of advice the engine writes. The citations come from places most businesses have never audited, which is why the useful output is a source list rather than a position report.
Can I check any of this myself before paying anyone?
A 15-minute version will tell you whether you need a full baseline. Write the three questions a customer asks before hiring you, including one cost question, and run them in ChatGPT and Perplexity, and in Google where an AI Overview appears. Then ask each engine what it knows about your business and check every fact, open the citations behind your top competitor's mention in Perplexity and write down every domain, and view the raw source of your most important page to confirm your headline and body copy are in the HTML.
How do I tell a real audit from a template deck?
Audits are cheap to fake: ten prompts, five screenshots, a template deck. Ask any vendor for the prompt list and the raw source citations before you pay, and if they will not share the method, the method is the product they are hiding. A credible run samples twenty to forty buyer prompts across ChatGPT, Perplexity, Google AI Overviews, Claude, and Gemini, repeated over several days, because answers vary between runs and a single screenshot proves nothing.
My site is clean but I still do not appear. What now?
If the engines describe you accurately and your page source is readable, but you are still absent from buyer prompts and from competitor citation lists, the problem is off-site corroboration rather than your website. That is a distribution and reputation project, not a redesign.
When is an AI visibility audit not worth buying?
Skip it if your business does not have a searchable category, if your buyers arrive entirely through referral, or if you already know your site is unreadable and unfixed. In that last case, fix the rendering and structured data first, then measure. An audit that documents problems you have already diagnosed is an expensive way to feel informed.
