A 60-word answer block placed under your H1 is the most reliable way to earn a Google AI Overview citation. Open with one sentence that answers the question and repeats its key nouns, add two or three sentences of specific context, name the entity, and use a plain verb. Schema makes the page eligible. The block is what gets chosen.
That is the entire play. Most teams overcomplicate it.
What an AI Overview actually picks
Google AI Overviews read the live index and refresh fast, often within hours to a day of a change. That makes them the quickest AI surface to move, and the most underrated. ChatGPT and Perplexity lag behind by days or weeks because their retrieval catches up on a slower cycle. If you want a fast read on whether a content change landed, AI Overviews are where you watch first. The three speeds of AI citations explain why.
An AI Overview pulls a short, self-sufficient answer from near the top of a page. Depth lower down still matters for trust and for ranking, but the citation itself is decided by the block. Get the first 60 words right and the rest of the article supports the case rather than carrying it.
Being fetched is not the same as being cited. Engines retrieve far more pages than they quote. Ahrefs' 2026 analysis found roughly half of the URLs ChatGPT retrieves never get named in the answer at all. The block is your bid to be the half that gets the credit.
Almost every AI Overview appears on an informational question. Ahrefs put it at 99.9%. Transactional, navigational, and "best tool" queries trigger them far less often (shopping queries only about 3% of the time). So this is a play for your explainer and guide pages, the ones answering "what is", "how do I", and "why does". Your pricing and comparison pages fight a different battle.
The 60-word block, structured
Five parts, in order:
- Sentence one is the answer. Restate the question's nouns and answer it flat. "Stablecoin payroll is..." beats "There are a few things to consider...". If a reader gets only the first line, they should already have the answer.
- Sentences two and three are specific context. A real number, a named mechanism, a concrete who-it's-for. Vague support reads as filler and gets skipped.
- Name the entity early. The product, the concept, the company, inside the first five words. Models anchor citations to entities, so bury the entity and you bury your odds.
- Use plain verbs. "is", "works by", "covers", "lets you", "costs". Skip the puffed-up ones like "empowers" or "unlocks". Inflated verbs add length without adding answer.
- Place it immediately under the H1. Nothing between the heading and the block. No throat-clearing, no warm-up paragraph before the answer.
Keep the whole thing to 40 to 60 words. Long enough to be complete, short enough for an engine to lift it whole without editing. The general principle behind the length lives in the 40 to 60 word rule for AI extraction; here it is aimed specifically at the AI Overview slot.
One more move that compounds the effect: phrase your H1 and your section headings as the exact question a buyer would type, then answer each heading in the block directly beneath it. The closer your heading matches the prompt, the cleaner the match the engine has to make.
Anatomy of a block that gets cited
Here is a finished block for the question "What is answer engine optimization?":
"Answer engine optimization (AEO) is the practice of structuring content so AI engines quote it directly in their answers. It works by leading each page with a short, self-contained answer the engine can lift, keeping the page parseable, and earning citations from sources the model already trusts. B2B SaaS teams use it to stay visible when buyers ask an AI instead of searching Google."
Now the four moves, labeled:
- "Answer engine optimization (AEO)" puts the entity in the first two words, so the citation has something to anchor to.
- "is the practice of..." answers the question flat and repeats its nouns. A reader who stops here still has the answer.
- The middle sentence gives three concrete mechanisms (lead with the answer, stay parseable, earn trusted citations) instead of three vague adjectives.
- The last sentence says who it is for and when it applies, which is what an engine needs to decide the answer fits the prompt.
Sixty-one words. Every one of them carries either the answer or the proof. That is the bar.
Why front-loading wins: the data
This is not a hunch, and the evidence runs one way.
Ask Google's own AI Overview how to optimize for AI Overviews and it tells on itself. It says to write a concise 40-to-60-word summary at the very top of the page that directly answers the query, because the system frequently pulls the first few sentences of an article. That is the answer engine describing this exact recipe back to you.
CXL studied where AI Overview citations actually come from on a page. Roughly 55% landed in the top 30% of the source. Kevin Indig's citation research points the same direction: bury the answer under setup and your odds of getting retrieved fall off a cliff. Position beats length, and position beats markup.
So the block is not a style preference. It is where the citation gets decided.
Schema is eligibility, not the lever
Here is the part that argues with the room. Ask Google's AI Overview how to optimize and it names structured data: FAQPage, QAPage, HowTo. Google's official Search Central guidance says the opposite, that there is no special schema you need to add to show up in AI features. When the machine's generated answer is more bullish on markup than the company's own documentation, that gap is the tell. Most guides repeat the schema advice and stop, which leaves readers tuning JSON and wondering why nothing moves.
Schema still earns its place. It keeps your page eligible and machine-parseable, and we cover the types that matter in schema markup for AEO. But eligibility is not selection. Adding markup, on its own, does not move AI citations. Ahrefs tested it across their 2026 study set and found no citation lift: AI Mode and ChatGPT moved up a couple of points, indistinguishable from zero, while AI Overviews actually dropped about 4.6%, a small but real decline. CXL's citation analysis pointed the same way: what predicted a citation was where the answer sat on the page rather than how much markup wrapped it. Our own AI-citation benchmark found the identical shape. Google even retired FAQ rich results in May 2026, so the markup many pages still carry buys them less than they think.
Treat schema as the floor. The lever is a clean, liftable answer in the first 60 words. If you have one hour this week, do not spend it hand-tuning JSON. Spend it rewriting the top of your five highest-intent informational pages.
Five B2B SaaS blocks: weak vs strong
Each example shows the version most pages ship, then the rewrite. Steal the shape for your own pages.
1. "What is stablecoin payroll?"
Weak: "Paying a global team has never been more flexible. There are lots of exciting new options worth knowing about before you decide."
Strong: "Stablecoin payroll pays employees and contractors in dollar-pegged stablecoins, on their own or alongside local currency. It settles in minutes instead of days and strips out most cross-border transfer fees. Recipients can hold the balance or convert to local money themselves. Teams use it to pay distributed contractors fast where bank rails are slow or costly."
Why it wins: the answer is in sentence one, the mechanism is concrete (settlement speed, fees), and the use case is named. Toku's AI-cited pipeline is a real example of a page in this space getting named in AI answers.
2. "How long do AI citations take?"
Weak: "AI citation timelines vary depending on a number of factors, and there is no one-size-fits-all answer."
Strong: "AI citations move at three speeds. Google AI Overviews can cite a new, well-structured page within hours to a day because they read the live index. ChatGPT and Perplexity usually take days to weeks as retrieval catches up. Brand-level share of answer, how often you get named across a whole category, takes months. Fast first citation, slow durable share."
Why it wins: it refuses the non-answer and gives a structure the reader can act on.
3. "How much does an AEO agency cost?"
Weak: "Pricing depends on your needs. Contact us for a custom quote."
Strong: "AEO agency pricing for B2B SaaS runs roughly $5,000 to $18,000 per month in 2026, depending on scope. Lower tiers cover content and on-page answer structuring. Higher tiers add original research, off-page authority work, and measurement. Most engagements are monthly retainers rather than project fees, because citation share compounds over months."
Why it wins: a real range beats "it depends". One caveat that proves the rule: pricing questions rarely trigger an AI Overview, but the same block still gets pulled into ChatGPT answers, so it earns its keep on a different surface. See the full AEO agency pricing breakdown.
4. "Why isn't my SEO traffic converting?"
Weak: "There are many reasons your traffic might not be converting, and the honest answer is that it really depends on your situation."
Strong: "SEO traffic usually fails to convert for one of three reasons: the keywords pull readers with no buying intent, the page answers the query but never names a next step, or an AI Overview already gave the visitor the answer so only low-intent clicks reach you. Fix intent first, then the offer, then measure assisted conversions instead of last click."
Why it wins: it names the three causes up front, so the engine has a clean list to lift. The full version is in why SEO traffic isn't converting to pipeline.
5. "What is share of answer?"
Weak: "Share of answer is an important new concept in AI search that every modern marketer should understand."
Strong: "Share of answer is the percentage of AI responses in your category that name your brand. It is the AI-search version of share of voice. You measure it by running the buyer questions in your space through ChatGPT, Perplexity, and Google, then counting how often you appear. Below 5% means you are close to invisible in the answer layer."
Why it wins: it defines the term, gives the method, and sets a benchmark. More on share of answer and how to audit your category in 90 minutes.
Three mistakes that keep you out of AI Overviews
- Burying the answer under an intro. If the first thing under your H1 is a warm-up paragraph, the engine reads warm-up and moves on. Answer first, context second.
- Writing for length instead of liftability. A 3,000-word page with no clean 60-word block loses to a tighter page that hands the engine a finished sentence. Length is for trust; the block is for the citation.
- Optimizing schema and skipping the copy. Markup is the floor. If the answer at the top is vague, no amount of JSON-LD rescues it.
How to tell if it worked
Do not judge in a day. AI Overviews reshuffle constantly. Ahrefs found the visible content changes about 70% between checks two days apart, while the underlying meaning stays roughly 95% stable. The words, sources, and order churn; the conclusion barely moves. So a single snapshot is just noise.
Check the same set of questions across a week. Watch for your domain appearing in the AI Overview source list itself. Rank is the lesser signal here; inclusion in the cited sources is the win you are after.
Then watch the slower number: are you getting named more often across the whole question cluster, not just on one query. That is share of answer, and it is the metric that compounds. A 90-minute share-of-answer audit gives you the baseline to measure against.
The reusable scaffold
Copy this, fill the blanks, paste it under your H1:
[Entity] [is / works by / lets you] [one-sentence answer that repeats the question's nouns].
[Sentence two: the specific mechanism, number, or method.]
[Sentence three: who it is for, or when it applies.]Rules: 40 to 60 words total. Answer in sentence one. Entity named in the first five words. No throat-clearing, no inflated verbs, no hedging. Read it back and cut any word that does not add to the answer.
Start with five pages
Most teams treat AI search like a black box and wait for something to happen. It is not a black box. The cheapest win on the board right now is rewriting the first 60 words of your best informational pages so an engine can lift them whole.
Pick five. Rewrite the top of each into a 40-to-60-word block. Check the questions across a week. Then do the next five. If you want the full method end to end, read the complete guide to answer engine optimization. The block is the work.
