In short
"Write for humans, not search engines" was correct advice for its era, but content today has a third reader — the AI system deciding what to cite — and that reader needs structure, not just good prose, to extract and quote your work accurately.
The advice that shaped a decade
"Write for humans, not search engines" was the correction the content industry needed after a decade of keyword-stuffed, robotic copy written purely to game rankings. It was good advice. It's still true. But it was never the whole picture, and the gap it left is bigger now than it was when the advice first spread.
There's a third reader now
A human reads your article and forms an impression over the whole piece — tone, narrative, argument. A search crawler indexes it and extracts signals about relevance and quality. But increasingly, a third reader shows up first: an AI system deciding, in real time, whether a passage from your page is worth quoting as the answer to someone's question. That reader doesn't experience your whole article. It extracts a sentence or two and decides, on the spot, whether it stands alone.
Good prose and good structure are different skills
Writing that reads beautifully in full — building context, delaying the payoff, weaving in nuance — is often exactly the writing an AI system can't cleanly extract, because the value is distributed across the whole piece instead of concentrated in one quotable sentence. This isn't a knock on good writing. It's a recognition that "readable by a human end to end" and "extractable by a machine mid-paragraph" are different requirements, and most style guides only optimize for the first.
What changes in practice
It doesn't mean writing worse, or writing for robots instead of people. It means adding a layer the old advice didn't anticipate: headings phrased as real questions, key answers stated fully in one self-contained sentence before the elaboration, FAQ sections that hold up as standalone units. None of this makes the writing worse for a human reader — a clear, direct answer near the top of a section is good writing regardless of who or what reads it next.
The failure mode of ignoring this
A well-written, well-researched article that buries its actual answer three paragraphs deep, wrapped in throat-clearing and caveats, is invisible to a system deciding in milliseconds whether to quote it. Meanwhile a much thinner, worse-researched competitor that states its claim plainly in sentence one gets cited instead — not because it's better, but because it's shaped to be found.
The real update to the old advice
Write for humans first, still. But recognize that being genuinely useful to a human reader and being structured for machine extraction aren't in tension — they're both served by the same discipline: say the true thing clearly, early, and in a form that doesn't depend on everything around it to make sense. That discipline was always good writing. It's just no longer optional if you want to be found.
Related on Troiana: Content Farms Can’t Win the AI Era · Why Your Brand Needs a Point of View to Rank.
Reference: the authoritative guidance lives at Google’s guidance on AI features in Search.
Common questions
Does writing for AI extraction make content worse for human readers?
No — leading with a clear, complete answer and structuring content around real questions tends to improve readability for humans too. The two goals mostly align rather than conflict.
Is this the same as keyword stuffing for old-style SEO?
No, it's closer to the opposite — this is about clarity and self-contained answers, not repeating phrases mechanically. Keyword stuffing degrades readability; this approach improves it.
Should every paragraph be optimized this way?
No — reserve the discipline (clear answer first, self-contained framing) for the sections most likely to be searched or quoted directly, like key definitions and FAQ answers, and let narrative sections breathe normally.