Free tool
Can an AI system actually use your page?
Search is being answered rather than listed. This checks whether one of your pages can be reached, read, quoted and attributed by the systems doing the answering — and tells you exactly what is in the way.
Free, no signup, no email. One page is checked, not the whole site.
What gets checked
Twenty-one checks in four groups. Each one carries a published weight, and the score is the share of that weight the page passes — nothing more.
Can an AI system reach it?
Whether your robots.txt lets AI crawlers in, and whether there is an llms.txt telling them what the site contains.
Can it parse the page and act on it?
Whether an agent can navigate the page, identify what is on it, and read your tables as data rather than pictures of data.
Can it answer a question using the page?
Whether there is a passage an AI answer could quote — a direct answer, a question and answer pair, a citable claim.
Would it trust and attribute it?
Whether the claims on the page are sourced, whether the organisation and author are identified, and whether anything says how current it is.
What this score is, and is not
- It is a share of a published checklist. Not a prediction of how much AI traffic you will get, and not a probability of being cited by anything.
- It reads one page. A strong home page and an unreadable product catalogue score the same if you submit the home page.
- It reads what your server returns. Content that only appears once JavaScript has run is invisible to this check — and largely invisible to AI crawlers too, which is worth knowing either way.
- A check that cannot apply counts as passed. A page with no tables gets full credit for the table check, which makes the score a little generous. We would rather say so than have you work it out.
- The group weightings are a judgment call. They are reasoned, published, and not derived from measured citation outcomes — because nobody has that data yet, including us.
Crawlix is built by Visara Labs. The paid product does this across every page of a site, fixes what it can, and measures afterwards whether the fix actually helped.