Free tool

AI visibility checker

Paste one page. The checker reads that page and the site’s robots.txt, then scores how reachable and how extractable it is for ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews.

The assessment is written by a language model reading your page against a fixed rubric. That is a heuristic, not a measurement. It tells you whether the door is open and whether the content is in a shape an engine can lift. It does not tell you whether an engine names you, because it never asks one.

COST · FREESIGN-UP · NONELIMIT · 3 CHECKS PER HOURUPDATED · 2026-08-15

Enter one page URL

Any public page. The checker fetches that page and the site's robots.txt, and nothing else.

Access and structure get read

Crawler rules for five engines, ten content structure checks, and how clearly the page states what it is about.

You get a score and the fixes

A 0 to 100 readiness score against a stated rubric, engine by engine, with the fixes ordered by impact.

What it checksone page, one robots.txt

Three things, on the page you gave it.

Access first, because structure is irrelevant on a page a crawler is told to skip. Then structure, because a page that is read and cannot be lifted from is a page that gets summarised without attribution. Then entity clarity, because an engine that cannot say who published something tends not to name anyone.

Crawler access, five engines

What your robots.txt says to GPTBot, PerplexityBot, ClaudeBot, Google-Extended and Googlebot, and which of the five your file never mentions.

Content structure, ten checks

Whether a direct answer sits in the first 300 words, whether headings are extractable, whether definitions, lists, tables and FAQ answers stand on their own, whether structured data is present and dated, and whether the body copy is in the initial HTML.

Entity clarity

Whether the page states plainly what it is and who published it, in a form a model can attribute without guessing.

How it scores

The access, structure and entity rubric. 40, 40, 20.

One rubric, published here so you can check the arithmetic against your own result. The model is given these weights and asked to apply them, which means the score is only as good as the model’s reading of your page. The band is the useful part. The exact number is not.

Crawler access

40 POINTS

Eight points per engine where access is explicit, six where the engine is not mentioned and general access is open, zero where it is disallowed. Five engines, forty points.

Content structure

40 POINTS

Four points per check passed, two for partial, zero for failed, across the ten checks CS1 to CS10. Forty points.

Entity clarity

20 POINTS

Twenty points where the page states its entity plainly, ten where it is implied, zero where a reader would have to guess. Averaged across the five engines.

BANDS · 80 AND ABOVE, AI-VISIBLE · 60 TO 79, PARTIALLY VISIBLE · 40 TO 59, LIMITED VISIBILITY · BELOW 40, NOT VISIBLE

Stated in advance

What this checker cannot see.

A free tool that overstates itself costs more credibility than it earns. Four limits apply to every run, and they are here before you run it rather than discovered afterwards.

Whether any engine has ever named you.

The checker reads your page and your robots.txt. It never sends a prompt to ChatGPT, Perplexity, Claude, Gemini or Google, so it cannot report an answer, a share, or which competitor was named instead of you. That is a different instrument and a different job.

Whether a crawler obeyed the file.

robots.txt is a request, not a control. Compliance is each crawler's own claim, and a rule that grants access is not evidence that anything came, read, or used the page.

Anything the page renders after load.

One fetch, raw HTML. Content injected by JavaScript is invisible here. It may also be invisible to a given engine, but this check cannot tell you which of the two you are looking at.

Its own consistency.

A language model scoring a rubric is not deterministic. The same URL can come back a point or two apart, so a small gap between two pages carries no information. Treat bands, not decimals.

One more, worth saying plainly. This check arrives from one address with one user agent. A firewall that lets it through can still turn away GPTBot, and a firewall that blocks it may be letting every engine in. If the fetch fails, that is a fact about our request, not a verdict on your site.

Method

The disclosure for this tool.

instrument: A language model, run through our own tooling. Not Caul, and not a capture from any engine.
what was measured: Nothing is measured. One page and its robots.txt are fetched and read, and a model scores them against the access, structure and entity rubric published above.
how: A single HTTPS request for the page, a second for the robots.txt at the same host, then one model call with the extracted signals and the rubric. No engine is prompted at any point.
over what window: The moment you press the button. Each result is stamped with the date it was produced and nothing is stored between runs.
what this cannot tell you: A model-generated assessment is not a measurement, not a citation share, and not evidence that any engine has named you. It reads raw HTML only, treats robots.txt as though it were obeyed, and is not deterministic: the same URL can score a point or two apart. Read the band, not the number.

This tells you the door is open. The report tells you who walked through it.

The free AI visibility audit is the measured version: a fixed prompt set run against the engines we have compliant access to, benchmarked against competitors you name, delivered on a call so the findings get explained. It states its engine coverage, its run count and its noise floor, and you keep the report and the underlying data whatever you decide.

Get your free audit What is in the report

Free. The ~~$1,500~~ is what this sweep is priced at when it is sold, not a survey of anyone else’s rates.