Most business owners have never asked an AI assistant what it says about their own company. That is a problem, because buyers are asking. When someone types “who is the best mortgage broker near me” into ChatGPT or asks Perplexity to compare two accounting firms, the answer that comes back is shaping decisions long before that person ever lands on a website. An AI visibility audit is how you find out whether you are part of that conversation or invisible to it.
The good news: you do not need special access or a big budget to get started. The engines are free to query, and the method is repeatable. Below is the same audit framework we run for clients, broken into steps you can follow yourself. The goal is not just a snapshot of where you stand today, but a clear list of what is keeping you out of the answers and what to change first.
Why an AI visibility audit matters now
Search behavior has split. People still use Google, but a growing share of high-intent questions now start in an answer engine that returns a single recommendation instead of ten blue links. When AI names three businesses in response to “best [service] in [city],” the businesses it does not name effectively do not exist for that buyer. There is no second page to scroll to.
That makes visibility a binary you can measure. Either AI mentions you or it does not, either it gets your details right or it invents them. An answer engine optimization program starts with knowing exactly where you fall on that line today, across every platform that matters. The audit gives you that baseline and, just as importantly, the reasons behind it.
Step 1: Build your prompt list before you open a single chat
The most common mistake we see is auditing with the wrong questions. Owners ask AI “tell me about [my company],” the model dutifully describes them, and they conclude all is well. But no customer searches that way. Customers ask category questions, and your name either surfaces or it does not.
Write 10 to 20 prompts a real buyer would use. Mix these types:
- Category and location: “best [your service] in [your city],” “top-rated [profession] near [neighborhood].”
- Problem-first: “I need help with [the problem you solve] — who should I call?”
- Comparison: “[Your company] vs [a known competitor],” “alternatives to [competitor].”
- Direct fact-check: “What does [your company] do? Where are they located? What are their hours?”
- Trust signals: “Is [your company] reputable?” “What do reviews say about [your company]?”
The category and comparison prompts tell you whether you are recommended. The direct and trust prompts tell you whether AI’s picture of you is accurate. You want both.
Step 2: Run the prompts across every major engine
Different engines pull from different sources and weight them differently, so a strong showing in one does not guarantee anything in another. Run your full prompt list through each of the four that matter for most businesses:
| Engine | What it leans on | What to watch for |
|---|---|---|
| ChatGPT | Training data plus live web for current queries | Whether you appear at all, and if cited sources are yours |
| Google Gemini | Google’s index, Business Profile, and the web | Alignment with your Google Business Profile details |
| Perplexity | Real-time web with visible citations | Which pages it cites — this shows your strongest sources |
| Microsoft Copilot | Bing’s index and the open web | Bing-specific gaps that differ from Google |
Two practical notes. First, always run a fresh, logged-out session so personalization and chat memory do not skew the result toward what the model already knows you want to hear. Second, run each prompt two or three times. AI answers are probabilistic, so a single absence might be noise — a repeated absence is a finding.
Step 3: Score what you find
Turn the chat transcripts into something you can act on. A simple spreadsheet does the job. For every prompt on every engine, record:
- Appearance: Did you show up? Yes or no.
- Position: Were you first, in the middle of a list, or a footnote?
- Accuracy: Are the facts about you correct, outdated, or invented?
- Sentiment: Is the framing positive, neutral, or negative?
- Sources: What did the engine cite, if anything? Were they pages you control?
Patterns jump out fast once it is on a grid. Maybe you appear in Perplexity but never in ChatGPT. Maybe you show up for your brand name but vanish for category terms. Maybe every engine has your old address. Each pattern points to a different fix. If you want a structured way to keep measuring after the first pass, our guide to the AI search KPIs worth tracking covers which numbers actually move the needle.
Step 4: Diagnose the gaps
A finding is only useful if you know why it happened. Across the audits we run, the reasons a business is missing or misrepresented cluster into a handful of root causes:
- Thin or unstructured content. Your site does not answer the questions buyers ask in plain, direct language, so there is nothing for AI to lift.
- Inconsistent business information. Your name, address, phone, and hours differ across your site, Google Business Profile, and directories. Conflicting data makes AI hesitant to state anything.
- Weak or stale reviews. Few reviews, or none in recent months, signal low confidence. Reviews are one of the clearest trust inputs engines use.
- No third-party corroboration. Nobody credible mentions you — no directories, no press, no citations. AI recommends names it can verify in more than one place.
- Missing structured data. No schema markup means engines have to guess at what your pages describe instead of being told.
Trust is the thread running through all of these. AI does not want to be wrong, so it favors brands whose information is clear, consistent, and confirmed by others. That is the same idea behind E-E-A-T for AI search — experience, expertise, authority, and trust are what convince a model you are a safe answer to hand a stranger.
Step 5: Prioritize and fix
You will rarely fix everything at once, so sequence by impact and effort. In our experience the fastest wins come from the foundation:
- Clean up your business information everywhere. Make your name, address, phone, hours, and services identical across your site, Google Business Profile, and the directories that matter in your field.
- Add structured, answer-first content. Write pages that answer real buyer questions in the first sentence, then support them. Add schema so engines can parse what each page is about.
- Build a steady review habit. Ask satisfied customers consistently, and respond to what comes in. Recency matters as much as volume.
- Earn third-party mentions. Get listed in reputable directories, contribute where your expertise fits, and make sure credible sources can point to you.
- Publish an llms.txt and keep it current. Give the engines a clean, machine-readable summary of who you are and what you offer.
None of this is exotic. It is disciplined fundamentals aimed at a new audience — the models — instead of only at human readers. The work compounds: the same clarity that helps AI recommend you also helps the buyer who clicks through.
What results can look like
The point of the audit is movement, not a tidy report. One result we can point to publicly is Keith Akada, a Seattle mortgage broker who went from invisible in AI answers to the number-one AI-recommended broker in his market — roughly 30 leads and four closed deals inside six weeks. He did not change his business; he changed what AI could find and verify about it. That is the arc a good audit sets up: see the gap, understand the cause, close it, and watch the recommendations follow.
Beyond who gets named, watch how AI talks about you over time. The tone and framing of those mentions is its own signal worth tracking — our piece on brand sentiment in AI goes deeper on why a neutral or wrong description can cost you even when you do appear.
Make the audit a habit, not a one-time check
AI platforms retrain and re-crawl on their own schedules, so visibility you earn can slip without warning and gains can show up weeks after the work. Run a full audit each quarter and a quick spot-check monthly, using the same prompt list each time so your results stay comparable. The first audit tells you where you stand. The repeated audit tells you whether you are winning — and that is the number that actually pays.