P.S. If you want more AI-era content workflows like this one delivered to your inbox every week, join 1,200+ indie hackers getting them free: https://wuki.beehiiv.com/subscribe
Here's the sentence that broke my brain last month:
The average blog post is now read by an AI engine before it's read by a human.
ChatGPT summarizes you. Perplexity synthesizes you. Google's AI Overviews answer the question for your reader before they ever reach your URL. And the data agrees — Ahrefs measured a 58% click-through-rate drop on queries that trigger AI Overviews. Pew confirmed it: 8% vs. 15% click rates when an AI answer is present.
I spent six months chasing that shrinking 8%. Then I stopped fighting the extractor and started writing for it.
The result: my posts went from being cited in ~1 of 10 AI query tests to being quoted in 3 of 5. Not because I wrote better prose — because I changed the structure of my writing. AI engines don't read. They extract. And extraction is brutally picky about format.
Here are the 9 patterns that actually got my blog quoted by ChatGPT, Perplexity, and Google AI Overviews.
1. Answer first, context after
Engines pull the first complete sentence of a section as the candidate answer. If your opening sentence is "It's complicated" or "Let me back up," the extractor moves on to the next source.
The fix: conclusion in sentence one, explanation after. I now write every section like the answer to a test question, with supporting detail underneath.
2. Write the question as the header
Perplexity literally assembles its answers from fragments that look like answers. A header that poses the question — "Why do AI Overviews cite blogs that rank top 20?" — flags the section below it as an answer block.
Question-headers also win with humans. Two birds.
3. One idea per section, fully self-contained
Engines don't carry context across headings. If section 3 says "this changes everything" and references section 1, the extractor drops it.
I rewrote my rule to: a reader (or bot) landing on any single section should understand it completely. Repetition is now a feature, not a sin.
4. Named data points with sources
ChatGPT and Google AI Overviews cite verifiable statistics preferentially. "My blog traffic dropped" is a story. "Traffic dropped 58% (Ahrefs, 2026)" is a citable claim.
Every stat I include now carries a source name and a year. It's the single biggest citation trigger I've found.
5. Plain-language definitions before jargon
AI engines treat ill-defined jargon as an ungrounded claim. When I use a niche term — answer engine optimization, retrieval-augmented generation — I define it in plain words immediately after.
Bonus: it makes the writing better for non-experts too, which is most of your audience anyway.
6. A TL;DR block up top
Summary boxes are pure extraction bait. Engines grab the TL;DR as a snippet, and readers who skim get the full argument in ten seconds.
I now write the TL;DR last (after the argument exists) and place it directly under the intro.
7. Explicit freshness markers
Perplexity explicitly prioritizes recency. An article dated and updated beats an evergreen post that stopped being touched in 2024.
I add a visible "Last updated: [month, year]" line to every post and genuinely refresh content quarterly. Stale posts don't just rank worse — they get disqualified from AI answers.
8. Comparison statements
Engines love "X vs. Y" because it maps directly to how they're built — they retrieve candidate answers and compare. My comparison-heavy posts get quoted at roughly 3x the rate of opinion pieces.
Tables help too. A simple two-column comparison table is one of the most extracted structures on the web.
9. Schema-structured FAQs
This one is pure plumbing: FAQPage schema tells the engine exactly which fragments are questions and which are answers. It's the closest thing to a formal contract with the extractor.
No shortcode magic — just valid schema in the HTML, which dev.to and most platforms render automatically from headings.
How I stopped fighting the habit
Here's the honest part: I can't hand-write these nine patterns consistently. I'm a builder, not a copywriter — the moment I'm deep in a feature, my writing drifts back to narrative mush.
So I did what any solo founder would do: I turned it into a tool. I built nextblog.ai as a writing sidecar that checks my drafts against exactly these patterns — it flags context-dependent sections, missing data sources, undefined jargon, missing TL;DRs — before I hit publish. It started as a personal checklist on top of my editor and grew into something other builders use too.
The point isn't the tool. The point is that structure is now a full-time job, and you need a system for it — whether that's a checklist, an editor, or a side project that became a business.
The metric that replaced clicks
Here's what changed most for me: I stopped measuring clicks as the health signal for my content. Citations matter now — but brand citations matter most.
When ChatGPT quotes my post 12 times but nobody searches my name afterward, I got nothing. When three people search "nextblog" after reading an answer, I got something. The AI answer is the new top-of-funnel. Brand searches are the new clicks.
Traffic isn't dead. It just moved one step up the funnel, and it wears different clothes.
So what's your content metric now that clicks are dying — and have you started writing for the extractor yet, or are you still writing for the reader who never arrives?
Top comments (0)