You can run a solid first pass of an AI visibility audit yourself in an afternoon. You will not match a full 39-criteria review, but you will catch the loud problems: a blocked crawler, a price saved as an image, a brand that four AI tools describe three different ways. Below is the checklist we use, mapped to the six areas we score in our own paid audit, written so a business owner can work through it without a developer.

Copy it into a doc, tick the boxes, and note where you fail. At the end you will know whether you can fix things yourself or whether it is time to hand it over.

What is an AI SEO audit?

An AI SEO audit checks whether AI tools like ChatGPT, Gemini, Perplexity, and Claude can read your site, understand your business, and recommend it in their answers. It looks at technical access, content structure, business signals, and how you actually show up when someone asks the AI a real question. It overlaps with classic SEO but measures different things.

Think of it in six layers, from the machine reaching your pages to the reputation that makes it cite you. Work through them in order. Each one assumes the one before it is passing.

1. Can AI bots actually reach your site? (crawlability)

Start here, because nothing downstream matters if the crawlers cannot get in. AI tools read the web with named bots, and many sites block them by accident. If GPTBot or ClaudeBot hits a wall, you are invisible before the content even gets a vote.

  • Open yourdomain.com/robots.txt in a browser and read it. Look for Disallow rules that target GPTBot, ClaudeBot, PerplexityBot, Google-Extended, or *. If your key pages are disallowed for those agents, that is your first fix.
  • Confirm your important pages return a normal 200 status and are not hidden behind a login, a cookie wall, or an “accept” gate.
  • Check that your core content loads in plain HTML, not only after JavaScript runs. View the page source (not the rendered page) and search for a sentence you can see on screen. If it is missing from the source, some crawlers will miss it too.
  • Make sure you have a working XML sitemap at yourdomain.com/sitemap.xml and that it lists your real pages.

Red flag: you find a blanket Disallow: / or an AI-bot rule someone added “for safety.” Green light: every page you want cited is reachable, in HTML, without a gate.

2. Can an AI lift a clean answer from your page? (extractability and structure)

AI tools quote pages that answer a question in one clean passage. If your key facts are buried in a wall of text, trapped in an image, or written as vague marketing, the model cannot extract them, so it reaches for a competitor who wrote it plainly.

  • For each important page, find the one question it answers and check that the answer appears in the first two sentences under a clear heading, in 40 to 60 words, before you expand.
  • Make sure prices, hours, service areas, and specs are real text, not baked into a graphic or a PDF. If your pricing lives inside an image, an AI cannot read it.
  • Use question-shaped headings that match how customers actually ask (“How much does X cost?”), then answer directly underneath.
  • Cut the ghost phrases: “comprehensive solutions,” “world-class quality,” “tailored approach.” They are invisible to humans and to AI. Replace them with specifics.

Red flag: your best information is a beautiful hero image with no words a machine can read. Green light: a stranger, or a model, could answer your customer’s question by quoting one paragraph. Our guide to optimizing your site for AI search goes deeper on the passage format.

3. Do your schema and entities line up? (structured data)

Schema markup is how you tell machines, in a language they trust, exactly what your business is, what you sell, and what each page means. It will not force a recommendation, but it removes ambiguity, and ambiguity is what makes AI describe you inconsistently.

  • Run your homepage and a service page through a schema validator (Google’s Rich Results Test or schema.org’s validator) and confirm you have valid Organization or LocalBusiness markup with your name, URL, and contact details.
  • Check that the entity details in your schema match what is visible on the page. Conflicting facts between markup and copy erode trust.
  • For articles and FAQs, add Article and FAQPage markup so your question-and-answer content is machine-labelled.
  • Make sure one canonical version of your business name, address, and phone number is used everywhere, in the schema and in the visible text.

Red flag: no structured data at all, or schema that names a different phone number than your contact page. Green light: valid markup that says the same thing your visible content says.

4. What do the four AI tools actually say about you? (observed visibility)

This is the section people skip and the one that matters most, because it measures reality instead of theory. You are going to ask the AI tools the questions your customers ask and write down what comes back. Do it in an incognito window to reduce personalization.

  • Write 8 to 10 real customer questions. Include “no-name” ones (“recommend a good [your service] in [your city]”) and direct ones (“what does [your business] do?”).
  • Ask all four tools the same set: ChatGPT, Gemini, Perplexity, and Claude. The same question returns different businesses in each, so one tool is not enough.
  • Log three things per answer: are you mentioned, are you recommended, and which sources the tool cites.
  • Run the important prompts more than once. Answers drift, so a single result is noise; a pattern across repeats is signal.

Red flag, and a common one: the AI describes you perfectly when you name yourself, but never brings you up unprompted. In our pilot for a dental practice, that exact gap showed up in 8 of 10 no-name questions. For a step-by-step version of this test, see how to check if AI recommends your business.

5. How do you stack up against competitors? (benchmark)

A score means nothing on its own. If you appear in 3 of 10 answers, is that good? The only way to know is to run the same questions for two or three competitors and compare. This turns a vague worry into a scoreboard.

  • Pick two or three businesses that would win the customer if you did not.
  • Run your no-name questions again and count how often each competitor is named and recommended, next to your own count.
  • Note what the AI says about them. Often the winners have something concrete you lack: reviews with locations, clear pricing, a strong “about” page, consistent listings.
  • Look at the sources the AI cites for them. Those pages are a map of where the signals it trusts actually live.

Red flag: competitors appear in answers where you are absent, and you can see why from their sources. Green light: you match or beat them on the questions that drive revenue.

6. Does your brand look credible off your own site? (authority signals)

AI weighs mentions and reviews roughly the way search weighs links. If your business is thinly represented across the web, or its details conflict from one source to the next, the model has little reason to trust and repeat you. Your own site is only half the story.

  • Search your business name and check that your Google Business Profile, social profiles, and directory listings all show the same name, address, and phone number. Fix every mismatch.
  • Look at your reviews: volume, recency, and whether they mention your location and services in real words. Sparse or stale reviews weaken you.
  • Check whether credible third parties mention you at all: press, partners, local directories, industry lists. Zero external mentions is a ceiling on how confidently AI can cite you.
  • Confirm your “about” and author information names a real person with real credentials. AI, like search, rewards clear authorship.

Red flag: three listings, three different phone numbers, and no reviews from the last year. Green light: consistent details everywhere and a steady trickle of recent, specific reviews.

When to hand it to a pro

Do it yourself when the failures are obvious and mechanical: a blocked crawler, a missing sitemap, prices trapped in images, inconsistent listings. Those you can fix in an afternoon, and you should, before paying anyone.

Hand it to a professional when three things are true at once. First, you have run the checklist and you still cannot tell why competitors win the answers. Second, the fixes need judgment, not just checkboxes, deciding which of twenty possible changes to make first, with limited time. Third, AI visibility is worth real money to you because customers in your market already ask AI for recommendations. At that point a repeatable, expert-read audit saves you weeks of guessing, and a done-for-you review will hold your site to all 39 criteria instead of the dozen you can eyeball.

That is exactly what our AI visibility audit does: the same six areas above, measured the same way every time, read by a person, with at least 5 prioritized fixes or a 100% refund. If you want to know what that costs before deciding, we publish it plainly in how much an AI visibility audit costs, and if you are still weighing it, is an AI visibility audit worth it? walks through when it is and is not. Want a broader starting point first? The free Marketing Audit scores your whole marketing in about three minutes.

FAQ

Can I do an AI SEO audit myself for free? Yes, for the loud problems. The six-part checklist above catches blocked crawlers, unreadable content, inconsistent listings, and weak visibility using only a browser and the free AI tools. What you cannot easily do alone is prioritize twenty possible fixes under time pressure or measure change reliably over time. That is where a structured, repeatable audit earns its price.

How long does a DIY AI visibility check take? Plan for an afternoon. The technical checks (robots.txt, schema, sitemap) take under an hour. The visibility test is the slow part: writing 8 to 10 real questions, asking all four AI tools, running the important ones more than once, and doing the same for two competitors. Rushing that section is the easiest way to draw the wrong conclusion.

Which AI tools should I test my business in? Test all four your customers actually use: ChatGPT, Gemini, Perplexity, and Claude. Results differ meaningfully between them, so checking only ChatGPT gives a false read. Use an incognito window to reduce personalization, and log the sources each tool cites, not just whether your name appears.

Is an AI SEO audit the same as a regular SEO audit? No. A classic SEO audit asks whether you rank in Google’s list of blue links. An AI SEO audit asks whether AI tools read you, understand you, and name you inside their generated answers. They overlap on technical basics like crawlability and structure, but the visibility measurement and the off-site signals that matter are different.