Writing

Write So an AI Can Quote You

When an AI assistant answers a question about your industry, it does something specific: it pulls a claim from somewhere and restates it. Not your whole page. One sentence, sometimes two.

That has a consequence most businesses haven't adjusted for. The page that gets cited isn't the most beautifully written one. It's the one containing a sentence that survives being lifted out and set down somewhere else.

What "quotable" actually means here

A quotable sentence is self-contained. It makes sense without the paragraph around it, without the heading above it, and without knowing what came before.

Compare two versions of the same idea.

"As we discussed above, this is the primary factor, and it's why so many of these efforts don't work out the way people expect."

"Most AI pilots stall because nobody was made responsible for changing the workflow."

The first sentence is fine prose and useless in isolation. Lift it out and it means nothing: which factor, what efforts, expected by whom. The second one travels. It can be quoted in an answer to someone who has never seen your site and it still says something true and specific.

Nothing about the second version is a trick. It's just clearer. That's the whole pattern, and it's worth saying plainly because this territory attracts a lot of advice about gaming AI systems. Writing that gets cited is writing that states things directly, which is also writing that serves the human reader better. The two goals point the same way, which is convenient and also the reason this approach holds up as the systems change.

Four things that make a page usable

Answer first, context after. If a page is about a question, the answer belongs in the first two or three sentences, not at the end of a build-up. Readers scan for the answer anyway. AI systems extract the first clear one they find.

One idea per paragraph. A paragraph carrying three arguments is hard to quote from and hard to skim. The fix is the same for both audiences: break it up.

Real headings that say something. "Overview" and "Background" tell nobody anything. "Why the preview looks correct when the link is broken" tells a reader and a machine exactly what's underneath it. Headings are the map of your page, and vague headings are a map with no place names.

Specific over general. "Significantly faster" is unquotable because it commits to nothing. "Two hours down to twenty minutes" survives extraction. Specificity is what makes a claim worth repeating.

The format nobody uses well

Plain question-and-answer sections are the most underused format available for this, and it isn't close.

An FAQ section, written properly, is a stack of self-contained answers to real questions, each one already in the shape an AI system needs. Most FAQ pages waste this by answering questions nobody asks ("Why choose us?") in language that commits to nothing.

The version that works uses the questions your customers actually ask, in the words they actually use, with answers that could stand alone. If people call and ask how long installation takes, the question is "How long does installation take?" and the answer starts with a number. If they ask whether you work with their industry, say which industries, by name.

That's it. It's not clever. It just requires knowing what people ask you, which most businesses do and few write down.

The test

Read any page on your site and ask: if someone lifted one sentence from this and put it in front of a stranger, would it still be true, clear, and useful?

If no sentence passes, the page has nothing an AI can use, and probably nothing a hurried human can use either. That's the useful part of this whole exercise. The changes that make your content legible to machines are the same ones that make it legible to a person who arrived in a hurry and has thirty seconds.

None of it matters, though, if the page can't be reached in the first place. A page behind a broken link or one that never renders is invisible regardless of how well it's written, which is a failure worth checking for separately, and the same reason identical Copilot licenses produce brilliant answers at one company and confident nonsense at another. Reachable first, then quotable.


Checking what AI systems actually say about a business, and why, is a service I run directly. Get in touch if you'd rather know than guess.


Rosemarie Withee has spent thirteen years helping operations teams get real work out of their software, first Microsoft 365, now AI. She’s written six books for Wiley and builds AI products at Portal Integrators.