Extractability is how cleanly an AI engine can lift a single, self-contained answer out of a page. Retrieval and citation happen at the passage level, so a page an engine has already fetched still only gets quoted if it contains a passage the engine can pull whole — a statistic line, a definition, a table row, a direct answer — rather than a fact tangled in flowing prose. It is the difference between a page being available and a page being quotable.
Extractability sits between two neighbours. Retrievability is whether an engine can find and fetch the page at all; extractability is whether, once fetched, a clean unit can be lifted from it. The unit itself is the content chunk — a self-contained section answering one question. The practical markers of an extractable passage are consistent: answer-first, one fact per sentence, roughly 15 to 45 words, led by the direct answer, and set in real headings, lists, and tables rather than a wall of text.
Extractability is a writing choice you control on every page, unlike authority or corroboration, which build slowly off-site. In MaxAEO's 2026 analysis, statistic lines and table rows were lifted verbatim far more often than narrative prose — directional, single-vendor evidence for the durable rule that structured, self-contained passages get quoted and buried ones do not. Because no engine reports which passage it lifted, the way to know a page is extractable is to track its citation coverage per engine after you restructure it.