Writing for machines that quote you
An answer engine does not read your page. It lifts a sentence out of it. The sentence has to survive that on its own.

The short answer
- An answer engine typically extracts one passage from a page, often the opening direct-answer sentence, and presents it without the surrounding context. That sentence has to be complete and accurate by itself.
- A table with clear headers and units is easier to lift whole than prose. It is already structured data, not a paraphrase the engine has to construct.
- Naming a source with a date inside the claiming sentence, rather than in a footnote, keeps the claim verifiable even when only that one sentence gets quoted.
- Answering one specific question per page gives an engine a clean passage to match against a query. A broad topic page forces it to guess which part is the answer.
A page written to be read in full is not written to be quoted in part
Most writing assumes a reader who arrives at the top and moves down: context first, then the point, often with the direct answer a few paragraphs in after some setup. That works for a person reading start to finish. It works badly for an answer engine. The engine is not reading the page. It retrieves the passage that best matches a question and presents that passage, often one or two sentences, with no guarantee the reader ever sees anything else.
If the direct answer is buried under three paragraphs of setup, the passage most likely to be lifted is not the answer. It is whatever sentence scores as the best semantic match to the query. That might be a hedge, a caveat, or a half-formed thought never meant to stand alone. Writing for this pattern means putting the complete, standalone answer first. Not as a summary of what follows, but as the actual answer. If nothing else from the page is read or quoted, that sentence alone is still correct and useful.
What makes a passage easy to lift correctly
Three properties make a passage more likely to be lifted, and lifted correctly. First, it is self-contained. No "as mentioned above", no "this approach", no pronoun that only resolves by reading three sentences earlier. A sentence quoted alone still names what it is about. Second, a table beats the same claim in prose whenever the content is comparative or numeric. A table with headers and units is already machine-readable data, not a paraphrase an engine has to build and might get slightly wrong.
Third, and less obvious, a claim with its source named inside the sentence survives extraction with its credibility intact. "CAN-SPAM requires a working opt-out mechanism, honored within 10 business days" carries its own verification. A bare claim about a law, with the citation two paragraphs away, loses that verification the moment only the claim gets quoted.
This site is built on that structure, and it is checkable. Every answer and blog page here opens with a block of key takeaways: standalone sentences meant to be the whole answer for a reader who stops there, before any supporting argument. That is not a claim about how well it performs with any engine. It is a description of the page structure, which anyone can verify by reading one.
The fix: one question per page, answered first, sourced inline
The practical rule is to scope a page to one question and answer it in the first hundred words. Not as a teaser for the rest, but as the complete answer a reader could stop at. If a second, related question needs its own explanation long enough to earn a heading, it usually belongs on a different page. Otherwise it competes for the same passage-extraction slot.
Numbers and comparisons belong in a table with real headers and units, not folded into a sentence where the unit has to be inferred. Any claim about a law, a policy, or a competitor's pricing should name its source and a date in the same sentence. The sentence then stays checkable even when it travels alone, quoted by a person or a machine, with nothing else from the page attached.
Questions
- Is writing for answer engines different from writing for search rankings?
- Overlapping, but not identical. Search ranking rewards a page overall, including depth and backlinks. An answer engine lifts one passage, often the first direct-answer sentence, and quotes it out of context. That sentence has to stand alone and be accurate without the rest of the page.
- Do tables actually get quoted, or only prose?
- Tables get lifted whole more easily than prose. A table with clear headers and units is already structured data an engine can extract without paraphrasing. A table missing units, or using ambiguous headers, is much less useful to lift, even when the numbers are correct.
- Does naming a source in the sentence actually matter for citation?
- It matters for trust, whatever the retrieval mechanics. A sentence with a named, dated source is verifiable in a way a bare number is not. Pages that make unverifiable claims are the ones a careful engine, or a careful reader, is least likely to treat as reliable enough to repeat.