What should an AI search visibility audit include?

"AI visibility audit" is a new enough phrase that it can mean almost anything, from a thorough investigation to a screenshot of one ChatGPT answer and a sales call. Before you pay for one, it helps to know what a complete audit covers, what the finished work should look like, and which shortcuts make the result worthless. This guide sets that out, so you can judge any audit, including one you run yourself.

What questions should an audit answer?

Strip away the jargon and a useful audit answers three questions.

1. Are you named? When buyers ask AI tools the questions that lead to a purchase, do you appear, where, and described how?
2. Why not? Which of the reasons AI tools pass a business over applies to you: they cannot find you, cannot understand you, cannot verify you, or cannot quote you.
3. What should you fix first? In order of value, with an honest sense of effort and time.

An audit that only answers the first question is a test, not an audit. The value is in the diagnosis and the order of work.

Part one: what do the AI tools say?

This is the part most people picture, and the part most often done badly. Done properly it includes:

• The right questions. A set of real buying questions, agreed with you, covering category and location searches, problem-first questions, your business name, and direct comparisons with competitors. Ten is a sensible minimum.
• The right tools. At least ChatGPT with search on, Perplexity, and Google's AI Overviews, plus AI Mode, Copilot or Gemini where your buyers use them.
• Clean, repeated runs. Logged-out or temporary sessions so personal history does not colour the answer, and each question run more than once, because answers vary.
• Named competitors. Your two or three closest rivals tested on the same questions, so you can see who is being chosen instead and why.
• The sources. Every site cited under each answer, recorded. These show which directories, review platforms and publications the tools trust in your market.
• Accuracy. Every wrong fact about you, such as an old address, a dropped service or a wrong price, noted with where it probably came from.

Part two: can the tools read your site?

If the crawlers cannot get in, nothing else matters. The audit should confirm:

• robots.txt allows the AI search crawlers, such as OAI-SearchBot and PerplexityBot, as well as Googlebot and Bingbot. The guide on whether to block AI crawlers explains which is which.
• No hosting, CDN or firewall setting is silently blocking AI bots.
• Your important text is in the page itself, not loaded by scripts a crawler may not run, and not trapped inside images.
• Your pages are indexed in both Google and Bing, checked in Google Search Console and Bing Webmaster Tools.

Part three: are your facts clear and consistent?

AI tools recommend businesses they can describe with confidence. This part checks:

• Whether your homepage and service pages state plainly what you do, for whom, where, and since when.
• Whether your name, address, phone number and services match across your website, Google Business Profile, Bing Places, Apple Business Connect, review sites and industry directories.
• Whether structured data, the code that labels facts for machines, is present, accurate and matches the visible page.
• Whether you have pages that answer the questions buyers ask, in a form a tool can quote.

Part four: what does the rest of the web say?

AI tools look for confirmation beyond your own site. The audit should compare your third-party footprint with your competitors': reviews and where they sit, trade body and directory listings, press and partner mentions. The most useful output here is a gap list: the specific sources cited for competitors where you are missing or listed incorrectly. If a competitor keeps being named instead of you, this is usually where the reason is, as why ChatGPT recommends your competitor explains in more depth.

Part five: how do you rank in ordinary search?

AI answers are mostly built on search results, so an AI audit that ignores Google is incomplete. It should check where you rank for your most valuable searches, who ranks above you, and whether technical problems such as slow pages or poor mobile layout are holding you back. Many of the fixes overlap, which is why GEO and SEO are best treated as one piece of work.

An audit that tests AI answers without checking access, facts, third-party evidence and rankings can tell you that you are missing, but not why.

What should you get at the end?

The deliverable is where weak audits show themselves. A good one gives you:

• The raw evidence: the questions, the answers, the sources, and where you and your competitors appeared.
• A plain diagnosis: which of the four reasons applies, in your words, not jargon.
• A ranked list of fixes: most valuable first, each with who does it, rough effort, and when you might expect to see a change.
• A baseline: the same question set you can re-run monthly to see whether the fixes work, using the method in how to track brand mentions in ChatGPT.
• A conversation: time with the person who did the work, to ask questions.

Watch for these red flags:

• Any promise of guaranteed recommendations or rankings.
• A proprietary "AI visibility score" with no explanation of how it is calculated.
• Results based on one run of a handful of questions.
• A long report with no priorities, which leaves the hard part to you.
• Recommendations that all lead to one expensive retainer, or to llms.txt and schema alone.
• No competitor comparison, which removes the most useful context.

What does it cost, and what are your options?

You have three realistic routes.

Do a basic version yourself. The free ten-minute AI test gives you the four core questions to ask. Add the access and facts checks from this page and you will catch the most common problems for the cost of an afternoon.

Buy an AI check as part of an SEO audit. Market prices for SEO audits vary widely, often from a few hundred to a few thousand dollars or pounds, depending on site size and depth; how much an SEO audit costs breaks down what moves the price. Ask specifically whether AI answers, cited sources and competitor comparisons are included, because many SEO audits leave them out.

Buy a combined visibility audit. M/AFZAL's Visibility Audit is one example: $700 fixed, delivered in 7 working days. It covers Google rankings for your 10 most valuable searches, what ChatGPT, Perplexity and Google's AI Overviews say about you, a review of how your website converts, a review of your current ads, and two competitors, with a 45-minute walkthrough and a ranked list of fixes that you keep whether or not you hire anyone afterwards.

Whichever route you take, ask any auditor four questions before paying: which questions will you test, on which tools, how many times each, and what exactly will I receive. Clear answers to those tell you more about the audit than any sales page. The AI search visibility guide covers the fixes most audits end up recommending, if you want to know what you are likely to hear.

The test of an audit is whether, a week after reading it, you know exactly what to do on Monday morning.