Cited Lens
All guides

How to write a page AI answers can quote: a before-and-after example

· 5 min read

When ChatGPT, Perplexity or Google's AI features answer a question, they do not summarise your whole page. They pull a few passages that look like answers and stitch them together. A passage that can be lifted on its own, states a fact and names its subject has a better chance than one that needs the surrounding page to make sense. The good news is that this is an editing problem, and you can fix it page by page.

What the research suggests

The GEO paper from Princeton, Georgia Tech and the Allen Institute for AI tested nine ways of rewriting a page and measured how much of a generated answer drew on it. In its main results table the tactics that added substance scored highest: adding relevant quotations (27.2), adding statistics (25.2) and citing sources (24.6), against 19.3 for the unchanged page. Keyword stuffing scored 17.7, below the baseline. The paper measured an academic pipeline, so treat the numbers as direction, not a promise. Our evidence-ranked guide goes through all nine.

Before: a typical product paragraph

Acme Invoice is a fictional product used for illustration. Every number below is a placeholder, so do not quote any of it.

"Acme Invoice is a best-in-class, next-generation invoicing solution that empowers freelancers to take control of their finances. With powerful features and an intuitive experience, our platform is trusted by thousands of happy customers worldwide."

Ask what a model could quote from that. It does not say who the product is for, what it costs, what it does or what it does not do. Every sentence could describe a thousand products, so an engine has no reason to prefer it.

After: the same page, rewritten to be quoted

Heading: How much does invoicing software cost for freelancers?

"Acme Invoice is invoicing software for freelancers who bill clients in more than one currency. It costs $12 a month, sends a reminder automatically when an invoice is 3 days overdue, and exports to CSV and QuickBooks. It does not handle payroll or inventory."

Then the supporting detail: a short comparison table with the same columns for each competitor, one sentence with a real, sourced statistic about the problem you solve (name the study and the year), and a quote from a named customer or expert with their role. The first paragraph can stand alone as an answer, and everything below it gives the engine something specific to corroborate.

Eight rules behind the rewrite

  1. Answer first. Put the direct answer in the first one or two sentences under the heading, then explain. Do not build up to it.
  2. Write headings as the questions people ask. "How much does invoicing software cost for freelancers?" matches a query. "Our pricing philosophy" matches nothing.
  3. Make each passage self-contained. Name the subject instead of writing "it" or "as mentioned above", because a lifted passage loses its surroundings.
  4. Use specific numbers with a source and a date. The research rewards real statistics and citations. A figure with no source reads as marketing to a reader and to a model.
  5. Quote real people with their names and roles. Do not invent a quote or a customer.
  6. State limits plainly. "It does not handle payroll" helps an engine answer "does Acme Invoice do payroll" correctly, and it builds trust with human readers.
  7. Use tables for comparisons. Same columns for every option make the facts easy to extract and compare.
  8. Show a real publication date and update it when you change the facts, so readers and engines can judge how current the page is.

Mark up what you wrote

Structured data states plainly what the page is, who published it and when. Google says no special schema is required to appear in AI Overviews or AI Mode, so treat markup as clarity, not a trick. Only mark up what is visible on the page, and never add FAQ markup for questions the page does not answer.

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "How much does invoicing software cost for freelancers?",
  "datePublished": "2026-11-04",
  "dateModified": "2026-11-04",
  "author": { "@type": "Organization", "name": "Acme Invoice", "url": "https://example.com" }
}

Mistakes that undo the work

  • Keyword stuffing. It scored below the baseline in the GEO paper's results.
  • Generic filler produced in bulk. If a paragraph could appear on a competitor's site unchanged, an engine has no reason to choose yours.
  • Fabricated statistics or quotes. They are easy to check and costly when someone does.
  • Burying the answer under a long introduction. A passage that begins with 200 words of context is harder to lift.
  • FAQ sections of questions nobody asks. Write the questions your sales or support inbox actually receives.

Test the rewrite

Pick the questions the page should answer, record whether you are cited today, make the changes, then re-check after the engines have had time to recrawl. Repeat each question at least three times, because answers vary between runs, and change only one page at a time so you know what caused the difference. The method is in our guide to checking AI citations, and you can see what reaches your pages in Google Analytics 4.

Frequently asked questions

How long should the direct answer be?

Short enough to lift whole: one or two sentences that name the subject and state the fact. Put detail after it. We know of no published study fixing an ideal word count, so be wary of guides that give one.

Should I add FAQ schema to every page?

Only where the page genuinely contains those questions and answers in visible text. Google says no special structured data is needed to appear in its AI features, so markup is for clarity, not for gaming.

Does adding statistics really help?

In the GEO paper's main results, adding statistics scored 25.2 against 19.3 for the unchanged page. That is one academic benchmark, not a guarantee for your topic. Use real figures with named sources.

Can I use AI to write these pages?

You can use it to draft, but generic output is the failure mode described above. Add what only you have: real numbers, named experts, product limits and first-hand detail.

Sources

Check your own site

Run the free audit to see which of these signals your site already has.

Run the free audit