Why do AI assistants quote my competitor's page instead of mine?
Often because their page states the answer as a self-contained claim near the top and yours makes the reader assemble it. Extraction favors sentences that survive being lifted out of context: a definition, a formula, a direct yes or no with its condition attached. Pages that open with background, bury the answer in the sixth paragraph, or say "as we explain below" give a summarizer nothing clean to take.
What a summarizer is actually looking for
A system assembling an answer needs a passage that is true on its own. It cannot safely quote a sentence whose meaning depends on the two paragraphs above it, because those paragraphs will not travel with it. So it prefers passages that carry their own subject, their own qualifier and their own conclusion.
This is a formatting property as much as a writing one. It has nothing to do with keyword density and it is not solved by adding a summary box at the bottom.
The lift test
Copy any single paragraph from your page and paste it somewhere with no surrounding context. Does it still answer a question? Run this on the first paragraph especially. If the first paragraph is scene-setting, the page fails the test at the point where it matters most.
- Answer first. The direct answer belongs in the opening, before the reasoning that supports it.
- Name the thing. Pronouns and "it" break extraction. Repeat the subject more often than feels elegant.
- Attach the qualifier. "Yes, if the system exposes a webhook" is quotable. "Yes, but see the caveats below" is not.
- Give a computable form. A formula or an explicit denominator is disproportionately likely to be quoted, because it is unambiguous.
Structures that block extraction
The answer living only inside an image or a chart with no prose equivalent. The answer split across a collapsed accordion. Marketing framing wrapped around the substance so the extractable sentence is about your company rather than about the question. And headings that say nothing — a heading that reads "Overview" tells a parser exactly as much as no heading at all.
What extractability does not fix
Being quotable does not make you findable, and it does not make you credible. Assistants still have to encounter the page and still weigh whether the source is worth citing. Structure is the part you control cheaply; the rest is the ordinary work of being genuinely useful and being linked. We treat the two as one program under AI search optimization, with the page structure of the answer hub doing the extraction half.
Topics: AI search · extraction · featured snippets · content structure
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.