What retrievability means
Before an assistant can quote you, three things have to hold. A crawler must be allowed to fetch the page. The words must exist in what it fetches, rather than appearing only after a script runs in a browser. And the passage must be recognisable as an answer to something, so that when a question is compared against stored passages, yours comes back.
Each of those can fail on its own, which is why retrievability is worth naming separately from ranking. A page can rank respectably in ordinary search and still never be retrieved for an AI answer, because ranking rewards the page as a whole while retrieval competes passage by passage. The unit is smaller, and a good page made of vague sections has no strong passage to offer.
Why retrievability matters
It is the first gate. Authority, brand and backlinks decide very little if the text never reaches the system in the first place, and this is the failure mode nobody notices, because nothing breaks visibly — you simply stop appearing in answers you would have expected to appear in.
It is also unusually fixable. Most retrievability problems are structural rather than competitive: a blocking rule someone added years ago, content trapped inside a widget, an answer split across a heading and several paragraphs. These are cheap to correct compared with earning authority, and they are ordinary technical SEO work rather than anything new.
Where retrievability goes wrong
The blunt failures come first: robots rules that block the crawlers you actually want, pages behind logins or forms, key facts living only inside images, sliders or downloadable files. A price table published as a picture is invisible to everything that would quote it.
Then come the rendering failures. If the answer is inserted by JavaScript after load, some systems will see it and others will fetch an empty shell, and you will never be told which. Tabs and accordions that only load their content on click behave the same way.
The quietest failure is the writing itself. A passage that never names its subject, or that depends on the section above to make sense, may be fetched and then discarded as unusable. Being indexed is not the same as being retrieved, and a page can pass every technical check and still offer nothing liftable.
How to act on it
Test rather than assume. View the page’s source, or fetch it with a crawler, and confirm the sentence you care about is present in the HTML without a browser rendering it. Check your robots rules deliberately and decide which AI crawlers you want to allow — a decision worth making on purpose, since blocking them removes you from those answers entirely.
Then make each answer stand alone: its own heading phrased as the question, the answer in the first sentence, the subject named rather than implied, and the facts in text rather than in an image. Move anything important out of tabs, PDFs and pop-ups onto the page itself. Once a passage can be found, the next question is whether it can be lifted cleanly, which is extractability and a separate piece of work.