---
title: "The 40-Word Answer AI Will Quote Word for Word"
seoTitle: "The 40-Word Answer AI Will Quote Word for Word"
description: "AI engines lift short, self-contained answers straight off a page. How to write the 40-word paragraph that gets quoted, and the structure to wrap around it."
datePublished: "2026-07-03T12:14:00Z"
dateModified: "2026-07-03T12:14:00Z"
category: ai
imageAlt: "Iron Goo blog featured image on writing the short self-contained answer an AI platform lifts off a page and quotes verbatim."
tags: [aeo, ai-search, citations, extractable-content, featured-snippets]
faq: true
---

I once wrote the same answer twice. A buyer question, asked the way buyers ask it: how long does a new commercial water heater take to install. The first version was careful and correct. "While every site is different and timelines can vary considerably depending on a range of factors, most standard residential and light-commercial installations, assuming no major complications, will generally tend to fall somewhere in the region of a single working day." All true. Nobody quoted it. I rewrote the same fact flat, in one breath, complete on its own: "A standard commercial water heater install takes about one working day. Older buildings with non-standard connections can add a second day for the extra plumbing." That second version got lifted into an AI answer, word for word, the next time the page was crawled. The fact never changed. The shape did. A quotable answer is that second sentence: one short, self-contained passage an AI platform can cut off the page, read with nothing around it, and use whole.

That gap is the whole craft, and it is smaller than it sounds. Not the topic, not the research, not the length of the page. The single sentence that resolves the buyer's question, written so it survives being pulled out of context. The first version above was buried inside its own hedging. The second stands alone. AI platforms quote the one that stands alone.

::::comparison{title="The same fact, written two ways"}
:::side{label="The hedged version, skipped"}
"While every site is different and timelines vary depending on a range of factors, most standard installations, assuming no major complications, will generally tend to fall in the region of a single working day." True. It depends on the sentence before it, hedges at both ends, and never says the number flat. There is nothing clean to lift.
:::
:::side{label="The flat version, quoted"}
"A standard commercial water heater install takes about one working day. Older buildings with non-standard connections can add a second day for the extra plumbing." Same fact. It answers the question in the first six words, stands on its own, and reads complete with nothing around it. An AI platform can quote it as-is.
:::
::::

## What does an answer AI will quote look like?

It is one short, self-contained passage, around forty words, that fully answers the question when read with nothing around it. Stated flat, no hedging or build-up, and placed directly under the question worded the way the reader actually asks it. That is the whole shape.

Read the paragraph you just read again. It is the thing it describes: roughly forty words, no setup, no "it depends", it answers its own heading in the first line and could be lifted out of this page and dropped into an AI reply without a single edit. That is not a coincidence. It is the shape the rest of this post teaches, demonstrated once so you can see it before you build one.

## Why short, and why around forty words

Length is the first thing people get wrong, in both directions. The instinct after being told to "write content AI will quote" is either to write more, on the theory that thoroughness gets rewarded, or to compress to a fragment that sounds quotable but does not actually resolve the question. Both miss.

Around forty words is the working shape, not a law. It is long enough to state a real answer with the one qualification that matters, and short enough that an AI platform can reproduce it without truncating. Go much longer and the passage gets cut mid-thought when the model pulls it, so the lift arrives incomplete. Go much shorter and you have a fragment that sounds confident but leaves the buyer's actual question half-answered, which is worse than not being quoted, because a half-answer that gets lifted makes you look wrong.

:::callout{type="key" title="Forty words is a shape, not a target to hit"}
Do not count to forty and stop. The number is the rough size of a passage that resolves one focused question and stays short enough to reproduce. A complete fifty-word answer beats a clipped thirty-word one every time. Write the answer whole, then check that it landed near this shape.
:::

The honest version of the rule: write the shortest passage that answers the question completely, and it will usually land somewhere around forty words because that is how much room one clean answer takes. The number is a description of what a finished answer tends to weigh, not a hoop.

## Self-contained means it survives being cut off the page

This is the property that decides everything, and it is the one careful writers fight hardest, because being self-contained looks, on the page, like being abrupt.

A self-contained answer is true and complete when read with nothing around it. No "as mentioned above". No "in that case". No pronoun pointing back at a noun three paragraphs up. The hedged water-heater sentence fails this badly: "assuming no major complications" only means something if you have read the setup, and "generally tend to fall in the region of" is the writer protecting themselves, not informing the buyer. Strip both and the fact is unharmed: it still takes about a day. The hedges were never the answer. They were the writer's nervousness wrapped around the answer.

Self-contained is also exactly what an AI platform needs, because it does not lift your page. It lifts a passage and drops it into a reply built from several sources. If your answer leans on the sentence before it, that crutch is gone the moment it is quoted, and a passage that depended on context now reads as broken or incomplete in someone else's answer. The passages that get reused are the ones that already work alone. This is the same instinct behind why [each self-contained sub-answer is what an AI platform pulls into its synthesized reply](/blog/query-fan-out): the model is assembling parts, and it reaches for the parts that stand on their own.

The qualification that genuinely matters is allowed to stay, as long as it stays inside the self-contained passage. "Older buildings with non-standard connections can add a second day" is not a hedge. It is the second half of the true answer, stated flat, and it makes the passage more useful to lift, not less. The test is simple: a hedge protects the writer; a qualification informs the buyer. Keep the second, cut the first.

## Where the answer goes, and why placement is part of being quotable

A perfect liftable sentence in the ninth paragraph is an unquotable page. Placement is not a separate concern from the answer; it is part of the answer's shape.

The answer goes directly under the question, worded the way the reader actually asks it, near the top of the page. When an AI platform reads your page looking for something to quote, it is scanning for a passage that resolves the query, fast. If the heading says the question in the buyer's own words and the next line answers it flat, the model finds the answer immediately and the match is obvious. If the answer is correct but sitting forty lines down after a preamble about why the topic matters, the model has no clean signal that this is the passage, and it quotes a competitor who put theirs up front. Same answer, different position, different outcome.

Wording the heading the way the buyer asks matters as much as position. "Installation timelines for commercial water heating systems" is how a company talks. "How long does a water heater install take" is how a buyer asks, and it is the second one that matches the question the buyer typed into the AI platform. Put the buyer's question as the heading, the flat answer directly beneath it, and you have done the placement half of the job.

## How to turn a buried answer you already have into a liftable one

Here is the part you can do this afternoon, on a page you already published. You almost certainly already have the answer. Most pages do. It is just buried, hedged, or sitting in the wrong place. Turning it into a liftable passage is four moves, none of which require new research.

- **Find the sentence that actually answers the question.** It is usually already on the page, somewhere in the middle, after the context. The fact itself, the number, the yes or no, the it-takes-a-day.
- **Strip the hedges and the lead-up.** Cut "generally", "in most cases", "while every situation differs", and the sentence that sets it up. Keep any qualification that genuinely informs the buyer; cut every one that only protects you.
- **Make it true standing alone.** Rewrite it so it has no pronoun or phrase that points at the surrounding text. Read it cold. If it needs the paragraph above to make sense, it is not done.
- **Put it directly under the question.** Word the heading the way the buyer asks, drop the flat answer on the next line, near the top of the page.

Then run one check before you move on. This is the whole test, and it is the only one that matters at the sentence level.

:::callout{type="tip" title="The lift test"}
Cut the passage off the page. Read it with absolutely nothing around it, the way an AI platform will quote it: no heading, no paragraph before, no page. Does it answer the buyer's question completely and correctly, on its own? If yes, it is liftable. If you reach for the missing context, it is not done yet.
:::

The lift test is deliberately literal. Do not imagine it. Actually select the sentence, paste it somewhere empty, and read it as a stranger would with no idea what page it came from. Most answers fail the first time, and the failure is almost always the same: a pronoun pointing back, a hedge that only resolves against the setup, or a number that needs the sentence above it to mean anything. Fix that, and the passage that was invisible becomes the one that gets quoted.

:::stat-grid
::stat{value="~40 words" label="the working shape of a complete answer"}
::stat{value="Line 1" label="under the question, not line 40"}
::stat{value="Stands alone" label="passes the lift test cold"}
:::

Treat that grid as the shape to aim for, not numbers to chase. The forty is a rough size, the placement is the habit, and the standing-alone is the one hard rule. None of it is measured from your specific page; it is the description of what a liftable answer tends to look like once it is finished.

:::quote{cite="A small-business owner, after the rewrite"}
The answer was already on the page. I just moved it to the top, cut the part where I was covering myself, and said it flat. That is the version the AI started quoting.
:::

One caution on mechanics. None of this is a claim about the model's internals. You do not get to read why a specific passage was chosen, and the behavior shifts between platforms and over time. What you can see is which version got quoted, and across enough rewrites the pattern holds: the flat, self-contained, well-placed answer gets lifted and the hedged, buried one does not. Keep the honesty there. You are shaping a sentence to be easy to use, not gaming a known formula.

Worth being clear about the boundary too. This is the craft of one passage: the sentence, its shape, its placement, the lift test. It is not the whole page. The page that liftable answer sits inside has its own work, a definition sentence, semantic structure a non-coder can build, several question-headings, a per-page check for whether the page is answer-ready, and that is [the full page rebuild the liftable answer sits inside](/guides/seo/content-that-earns-snippets-and-citations). Write the sentence first; build the page around it second.

So the move this week is one passage. Open a page that should be getting quoted and is not, find the buried answer, strip the hedges, say it flat under the buyer's own question, and run the lift test on it cold. When that one answer passes, go read the page-craft guide for how to structure the rest of the page around it, and if rewriting every buried answer across your site is more than you want to do by hand, that is the kind of work [the AI search service](/services/aio) takes off your plate.