Answer
Lead each answer with the direct response in its first sentence, keep it to roughly 40 to 90 words, and phrase the question the way people actually ask it [1]. Answer engines lift self-contained passages, so concrete numbers, plain language, and one question per heading make an answer easy for a model to quote and attribute [2].
The recipe
Answer engines retrieve short, self-contained passages rather than whole pages, so the first sentence of a section carries the most weight. An answer that opens with the direct response survives being pulled out of context, while one that opens with hedging or a paragraph of setup tends to get skipped. Front-loading the main point is a long-standing web-comprehension principle documented well before AI search, and it maps directly onto how a model decides which passage to lift.
Full-sentence, conversational questions tend to line up with real prompts better than keyword fragments. Drawing questions from Google's People also ask results, from sales and support conversations, and from autocomplete captures the phrasing users bring to ChatGPT, Perplexity, and similar engines. Keeping one question per heading, with the answer directly beneath it and nothing wedged in between, lets a model isolate a clean question-and-answer pair to quote.
Growth agency Lean Labs argues in How to write FAQ content answer engines will quote that answers should target 40 to 90 words, with the direct answer in the first sentence, roughly 30 to 50 words for definitional questions and 60 to 90 for cost or process questions. The post also recommends concrete figures over vague words (a stated price range instead of "affordable"), the question as an H2 or H3 heading in sentence case, and FAQPage schema in JSON-LD whose text matches the visible answer, since mismatches between markup and page content get the schema ignored.
Citations
4 total