Fetchable, chosen, extractable
The three tests a page passes before an AI answer cites it: the platform can read it, picks it, and can lift a passage out of it.
Fetchable, chosen, extractable is a way of breaking down what has to go right before an AI answer cites one of your pages. First the platform has to be able to get the page. Then it has to pick your page over the others it found. Then it has to be able to take a usable passage out of it.
It's useful because each step fails for different reasons and needs a different fix. If you know which step you're failing, you don't waste time on the other two.
Fetchable
Can the platform's bot reach the page and read its text? Your robots.txt might block the bot. A firewall might turn it away. The page might return an error, or the words that matter might only appear after a script runs, so a bot reading the HTML never sees them.
Chosen
Out of all the pages the platform found for the question, does it pick yours? This depends on how well the page matches what was asked, and on whether it looks like a source worth trusting. A comparison on an independent review site may beat your own product page when someone asks which brand is best, even when both can be read.
Extractable
Once picked, can the platform lift out a passage that answers the question on its own? A direct answer near the top, under a heading that matches the question, is easy to use. The same fact buried halfway through a long paragraph is harder.
An example
For example, say Kestrel Roasters' brewing guide never comes up in answers. Kestrel checks in order. The bots are allowed, and the page loads, so it's fetchable. The platforms do read it for a few prompts, so it's sometimes chosen. But the guide opens with two paragraphs of company history, and the actual brewing ratios sit in an image. That's an extractability problem, and rewriting the opening is the fix.
In AnswerOrbit
Each step has a place to look. For fetchable, the Audit and Pages tabs in Website show which AI bots your robots.txt allows and what a bot reads on each page. For chosen, Sources shows which pages the answers used, and Cited, not named shows when your page was used for an answer that named someone else. For extractable, Page checks look for things like the prompt in a heading and an opening that answers.