How to get your website quoted in ChatGPT, Perplexity and Google AI answers

What the four rules in the table do

Each row is one test you can run on a paragraph in under a minute. None of them asks if your writing is good. They ask if a machine can lift your sentence and drop it into its answer without fixing it.

The first rule does the most work. Put the answer first. Not the context. Not "it depends." The bare answer, then the nuance underneath. An AI scanning your paragraph reads the first sentence, and if it is a warm-up, it often moves on.

The second rule catches the most people. On your site, every paragraph leans on the one before. "As mentioned above" means nothing once it is lifted, because there is no "above" in the AI's answer. A self-contained sentence names its own subject.

The third rule is about hedging. "In many cases, depending on the business, it can sometimes make sense" holds no claim at all. State the claim. Then put your honest caveats in the next sentence, where they inform instead of dissolve.

The fourth rule is about the heading. Nobody types "Optimizing Financial Administration" into ChatGPT. They type "do I need a bookkeeper if I'm a freelancer?" When your heading uses those words, the AI finds the question and the answer sitting together.

Why this matters more than it did a year ago

The answer is replacing the list. When Google shows an AI summary, people click a normal link about 8% of the time, against 15% with no summary. They click a link inside the summary about 1% of the time (Pew Research Center).

So the sentence the AI lifts now does the job your homepage used to do. If nothing on your page is liftable, you don't get a thin version of the benefit. You get none of it.

And ranking won't save an unliftable page. The share of AI Overview citations taken from Google's top ten fell from about 76% to about 38% (Ahrefs).

The best page on her site, and nothing to quote

Priya, the bookkeeper in the book, had a long, careful page that answered "do I need a bookkeeper as a freelancer?" better than anyone she competed with. The AI never quoted it.

She read it the way a machine would. It was warm and true. Her real answer, the fifty-invoice line, was in there. But it was hidden inside "once you reach a certain volume of invoices." Fog, not a number.

So she rewrote it. The heading became the customer's question. The first sentence became the answer, with the number in it. The caveats moved to their own sentences behind it. She added no new facts.

Then she did the same to two more answers, including the one about price, where careful people hedge the most. She made herself lead with a real range.

A few weeks later a friend forwarded her a Perplexity reply. In the middle sat her own sentence: "Software handles the recording, but a bookkeeper catches the mistakes software makes silently." With a citation back to her page.

Two fixes that look right and fail

Keyword stuffing is the first. It comes from old habits, and it makes a sentence that names the topic and answers nothing.

The second is the wall of text. One long essay, no question headings, no place where an answer sits alone. The answers may be in there. The machine won't do the surgery to get them out.

I would start small. Take your three most important answers, the ones a customer asks to find someone like you, and run them through the four rules. Don't rewrite the whole page.

Put a real number or a named source in the answer

Many guides on this question say the same thing: answers with a checkable fact get quoted more. There is some research behind it.

The 2023 paper that named this field, GEO: Generative Engine Optimization, tested ways of rewriting web content and measured how visible it became in AI answers. Three methods did best: citing sources, adding quotations and adding statistics. The authors report a relative gain of 30 to 40% on their main visibility measure. Making text easier to read and more fluent also helped, by 15 to 30%. Keyword stuffing gave little to no gain.

That was a lab test on research systems, not a promise about ChatGPT. But it points the same way as the book. Priya's answers got lifted after she put real numbers in them: about fifty invoices a month, and $150 to $400 a month for a bookkeeper. A number gives the AI something firm to repeat.

Two rules keep this honest:

• Only use numbers you can stand behind. Your own prices, your own counts, or a figure from a source you link.
• Name the source in the sentence. "Pew Research Center found..." travels with the sentence when it is lifted. "Studies show..." does not.

How to check if you are being quoted

You won't get a report for this, so you check by hand.

1. Write down the three questions your rewritten answers cover, in your customer's words.
2. Ask each one in ChatGPT, Perplexity and Google's AI.
3. Look for your sentence, your name and a source link back to your page.
4. Check again from time to time. Answers change, and the book says to update your list of questions every few months.

In the book, Priya found out the way many owners do: a friend forwarded her a Perplexity answer with her own sentence in the middle of it.

The citability chapter, rule by rule

This is the problem Found by AI was written for.

Chapter 4, Can I Quote You Cleanly?, is the full version of the table above. It walks each rule with a before and after, then shows Priya rewriting three real answers and filling the four-line Citability Checklist. It also covers why the AI picks the plainest sentence over the prettiest one.

Chapter 5, Are You There When the Question Is Asked?, makes sure your quotable answers sit under the questions people really ask. Its Prompt-Match Map shows you how to harvest those questions from the AIs, your calls and your inbox.

The free Found by AI workbook has the checklist as a printable page. For how the quote fits into the bigger picture, read GEO vs SEO.