Can an AI answer
actually quote you?
The paid tools tell you whether you were mentioned, which needs thousands of costly model queries — so their free tiers return "N/A". This tests the thing you control: whether your page can be reached, read, quoted and credited at all.
Five stages, in order
Being cited is a funnel, and the stages are not equal. Fail the first and nothing after it can save you.
1. Reach
We send a real request as GPTBot, ClaudeBot, PerplexityBot and four others, and report what your server actually returns. Then we parse robots.txt properly, respecting group scoping rather than grepping for "Disallow".
2. Read
How much content exists before any JavaScript runs, what share of the file is readable text, and whether the page shows the signature of being assembled in the browser.
3. Extract
Whether the opening paragraph is a self-contained answer of quotable length, whether headings are shaped like real questions, and whether each question is followed straight away by its answer.
4. Attribute
Identity schema, a named human author, a machine-readable date and a canonical URL. An engine that cannot work out who published something has nobody to credit.
5. Trust
Concrete figures rather than adjectives, outbound citations to real sources, sentence length, and whether the page is padded with stock phrases that appear on a million others.
And the passage itself
We score every paragraph the way an engine would and show you the one most likely to be lifted. If it is not the sentence you would have chosen, that is the most useful thing on the page.
Three things the paid tools will not tell you
We use them too. This is where the honest line sits.