How is AI readiness scored? Every check, what each one looks for, and what the score cannot see

Five categories, every check inside them, and the things we deliberately leave out. A score you cannot interrogate is just a number someone chose, so this page exists to let you argue with ours.

Check your site for free

The five categories

In the order an AI engine runs into them. The first one is a gate: fail it and the other four never get looked at.

  1. Crawler access

    Everything else is irrelevant if the crawler never arrives.

    • Your robots.txt, read against twelve crawler names across six AI engines
    • A block on a crawler that only gathers training data, GPTBot or ClaudeBot, is reported and costs no points
    • Whether a sitemap exists and is reachable
    • Whether an llms.txt file is present
    • Whether the page answers at all, and with what status
  2. Structure

    A machine reads structure before it reads prose.

    • Structured data, and whether it is valid
    • Whether headings run in order rather than jumping levels
    • Whether the organisation can be told apart from the content
    • Title and meta description, present and the right length
  3. AI understanding

    Being readable is not the same as being understood.

    • An AI reads the page and tries to name the business
    • Whether it can name the services offered
    • Whether it can place the industry
    • Whether it can say where you are
  4. Speed

    A crawler that waits too long leaves with half the page.

    • Load performance, measured through PageSpeed Insights
    • Whether the main content needs JavaScript before it exists
  5. Security

    A small check, usually an easy fix, and a trust signal to every engine.

    • HTTPS, and whether it is served correctly

How the total is worked out

Four steps, in this order. The one on the sample report runs the whole way down.

  1. Each category is scored on its own

    Out of 100, from the checks listed above it. Crawler access and security are mostly pass or fail; understanding is what the page states plainly on its own.

  2. The five are weighted and added

    They do not count equally. A blocked crawler costs more than a slow image, because one of them ends the visit and the other delays it.

  3. The engine ceiling is applied

    No page scores 100 from this check alone. What a model knows about you from the rest of the web is outside what one page can show.

  4. The total lands in a band

    The band is the verdict, and it is what the report leads with. The five are below.

Four checkers, one scan

One audit behind all four. Each leads with the number that answers its own question.

  • AI readiness checker

    Can AI find, read and trust your site?

    • Crawler access
    • Structure
    • AI understanding
    • Speed
    • Security
  • AEO checker

    Can an answer engine quote you?

    • Crawler access
    • Structure
    • AI understanding
    • Schema completeness
    • Business entities
  • GEO audit tool

    Can a generative engine cite you?

    • Crawler access
    • AI understanding
    • Structure
  • ChatGPT visibility checker

    Can ChatGPT reach and understand you?

    • ChatGPT's own crawl access
    • Structure
    • AI understanding
    • Speed

Under that number every report is identical.

These are not checkers

Six more pages answer one factual question on the spot. No audit, no score and nothing asked for: these questions come up constantly and deserve a straight answer rather than a funnel.

llms.txt checker
Fetches /llms.txt, checks it is served as text rather than wrapped in a page template, and reads it against the proposed format.
AI crawler checker
Reads your robots.txt against every named AI agent we track and tells you which ones are allowed through.
Sitemap checker
Fetches /sitemap.xml, counts the URLs in it, follows a sitemap index, and checks your robots.txt declares it.
robots.txt tester
Parses the file into its groups and rules, and flags the mistakes that quietly turn a rule off.
Structured data checker
Reads every JSON-LD block on the page, reports what each one declares, and hands back stubs for the types that are missing.
llms.txt generator
Builds a draft from your own title, description and sitemap, with the gaps marked where the page did not answer.

What the verdict means

Five bands, highest at the top. Every report lands in one of them.

  1. 90 and above

    AI Ready

    Engines can reach it, parse it and say what you do.

  2. 80 to 89

    Visibility opportunities

    Strong overall readiness. Review the remaining findings; individual checks may still fail.

  3. 60 to 79

    Recommendation gaps

    Readable, but an engine cannot confidently say what the business is.

  4. 40 to 59

    Mostly invisible

    Something structural is in the way.

  5. Below 40

    AI invisible

    Usually a blocked crawler or a page that is empty without JavaScript.

What we deliberately do not score

A methodology page is only honest if it says what it leaves out.

Whether an AI engine actually names you
Not measurable from your own page. That depends mostly on what other websites say about you, which cannot be measured from your own page. Anyone selling you a number for it is estimating.
Backlinks and domain authority
Already covered by SEO tools. Real factors in how often you get quoted, and thoroughly covered by existing SEO tools. We do not duplicate them.
Content quality
Not a judgement we make. We check whether a machine can follow your page. We do not grade your writing.
Anything behind a login
A crawler never reaches it. We only read what a crawler could read.

Where the data comes from

Nothing is estimated and nothing is averaged against a benchmark we have not measured.

Crawler and structure
Read straight off your site.
Speed
Google PageSpeed Insights.
Understanding
We read your page text and record which business facts it makes clear.
If PageSpeed is down
We use the response time we measured, and drop the Core Web Vitals panel.

Questions about the score

How is an AI readiness score calculated?

Five categories, out of 100: crawler access, structure, AI understanding, speed and security. They do not count equally, and the order they are listed in is the order of how much each one costs you. Every check behind every category is on this page, along with what the score cannot see.

Why do two AI readiness tools score the same site differently?

Because the scores measure different things and almost nobody says which. Check counts run from five to forty-two across this category, and a tool that leans on schema will disagree with one that leans on crawler access. Compare what each tool checks, not the numbers it returns.

Should an AI readiness check look at one page or the whole site?

Both, for different reasons. Crawler access and security are site-wide by nature, so one check settles them. Structure and understanding are judged per page, because the page an engine lands on is the page it has to understand. The free check reads the one you give it.

What is the difference between an AEO checker and a GEO checker?

The vocabulary, mostly. AEO asks whether a passage on your page can be lifted as an answer. GEO asks whether your page is a source an engine reaches for. Both are read off the same signals, which is why one audit answers both and four separate scans would tell you the same thing four times.

Do I need separate tools for AEO, GEO and AI readiness?

Not at the start. One check that covers crawler access, structure, speed, security and whether an AI can describe your business will tell you more than three dashboards will. Separate tools start earning their keep when you are tracking what engines say about you over time.

What should an AI readiness check actually test?

Five things. Whether AI crawlers can reach the page. Whether its structure can be parsed. Whether a model reading it can say what the business does. Whether it loads before a crawler gives up. And whether it is served securely. Anything that scores you without testing those is guessing.

Are free AI visibility checkers accurate?

They are accurate about what they can see, which is your own page, and silent about what they cannot, which is what other sites say about you. Treat any free score as a readability result rather than a prediction of being recommended. Ours is published in full on this page.

Which check should I run first?

The AI readiness checker, unless you already know which half of the problem you have. It is the whole audit and the other three are narrower views of it.

Does a high score mean ChatGPT will recommend me?

No, and no score can promise that. A high score means an engine can reach your page, parse it and describe your business from it. Which business gets named in an answer also depends on what other sites say about you, which cannot be read off your own page.

See your own numbers.

The same five categories, scored on your page, with the fixes in order of what they are worth.

Run the free check