Free GEO audit checklist: 36 checks to see if AI can read and use your website

How to use this checklist

Work through the six tables in order. Each row has the check, the rule we use to call it a pass, and why it matters. For checks that look at many pages, run them on your homepage plus your 10–25 most important pages (services, products, pricing, contact, top articles).

Mark each check pass, warning or fail. Then fix the fails in the access-related areas first (navigability and agent collaboration): if a crawler can't get in, nothing else on the page counts.

A few honest caveats before you start:

• Passing every check doesn't guarantee you'll be cited or recommended by any AI assistant. Nobody can promise that.
• Some checks matter more for AI agents and assistants than for Google. Google says foundational SEO practices, such as a clear technical structure and unique, valuable content, are the foundation for visibility in its generative AI features, so good SEO still does most of the work there.
• Two checks (form labels and button text) only apply if your site has forms. Without forms, they're replaced by one check: does the site offer any action at all?

1. Content semantics (7 checks)

Can a machine tell what each page is about and how it's organised?

Google's title link guidelines ask for a title on every page that is "descriptive and concise", and warn against vague titles like "Home".

2. Structured data (7 checks)

Does the site describe itself in a machine-readable format?

Google says it uses structured data "to understand the content of the page" and to gather information about things like companies (intro to structured data). But in its guide to optimizing for generative AI features it also says structured data "isn't required for generative AI search, and there's no special schema.org markup you need to add". So treat this area as clarity, not a magic switch. Validate your blocks with the Schema Markup Validator.

3. Navigability (7 checks)

Can a crawler find and fetch your pages?

Two details catch people out. Google says robots.txt "is not a mechanism for keeping a web page out of Google" (robots.txt introduction); use noindex for that. And if a page is blocked in robots.txt, a crawler never sees its noindex tag (block indexing with noindex).

4. Actionability (5 checks)

Could an AI agent acting for a customer contact you, book or buy?

The W3C's forms tutorial says to "provide labels to identify all form controls", which helps screen readers and AI agents alike.

5. AI promptability (5 checks)

Is the content written and delivered so an AI can quote it?

Check 27 is the one most sites fail without knowing. Vercel and MERJ's analysis found none of the major AI crawlers they studied rendered JavaScript, naming OpenAI's GPTBot, OAI-SearchBot and ChatGPT-User, Anthropic's ClaudeBot and PerplexityBot, among others. Google's Gemini was the exception, because it uses Googlebot's infrastructure. To test it yourself, open a page with JavaScript turned off, or view the page source and search for a sentence from the middle of the page.

6. Agent collaboration (5 checks)

Does the site make its AI policy explicit and actually let the bots in?

Be realistic about llms.txt. It's a proposal from llmstxt.org for "a /llms.txt markdown file" that gives LLMs "LLM-friendly content". Google says it doesn't use these files for Search, and that creating one "will neither harm nor help" your Google visibility. We still include it because it's cheap and some AI tools and agents do read it. For the difference between bots that train models and bots that fetch answers, see how to allow AI search bots but block AI training.

Check 36 matters more than it looks. robots.txt is a request, but firewalls and CDNs enforce rules before a bot ever reaches your page. Our guide on how to check which AI crawlers can access your website walks through testing it.

What to fix first

If you only have an afternoon, do these in this order:

1. Let the right bots in (checks 16, 35, 36). Nothing else matters if crawlers get a 403.
2. Make the text visible without JavaScript (check 27). This is usually a developer or platform setting, so ask early.
3. Fix titles, H1s and meta descriptions (checks 1, 5, 6). Quick to do in any CMS.
4. Add Organization and WebSite JSON-LD (checks 10, 11), with your real name, logo, contact details and profiles.
5. Publish a sitemap, canonicals and an llms.txt (checks 17, 18, 32).
6. Add a short FAQ on key pages (check 29), answering the questions customers actually ask.

How long does a manual audit take?

For a small site, about an hour with a browser and a text editor. The slow parts are the per-page checks (titles, headings, word counts, alt text) and the bot test in check 36, which needs a command-line tool such as curl to send requests with different user-agents.

What this checklist doesn't cover

• Whether AI assistants already mention you. Check that by asking them and noting the sources they cite. Our checklist for getting cited by ChatGPT covers that side: search rankings, quotable answers and mentions on other sites.
• Content quality and accuracy. A technically perfect page with thin or outdated facts still won't be recommended.
• Off-site signals such as reviews, directories and press.

Run it automatically

You can do all of this by hand with the tables above. If you'd rather not, the free Web2Agents AI check scores your homepage across the same six areas, with results on screen in real time. The €19 AI-Ready Kit runs all 36 checks on up to 25 pages, shows the evidence for each, ranks the fixes, and includes ready-to-upload robots.txt, llms.txt and schema files. More guides are in our agent-ready websites section.