Rapid Digify

How to get cited by ChatGPT and Perplexity

AI assistants quote specific sentences from specific pages. Here is what makes a page quotable, how to structure it, and how to check whether it is working.

Abdullah4 min readUpdated

Getting quoted by an AI assistant is not mysterious. When a model with browsing answers a question, it retrieves a handful of pages, pulls sentences that directly answer the question, and attributes them. Your job is to be the page with the clearest sentence.

Write the answer in the first fifty words

This is most of the work.

A model scanning a page for an answer is looking for a self-contained statement it can lift without the surrounding context. A page that opens with "In this article we will explore the many factors that influence..." gives it nothing.

Put a direct, complete, 40 to 60 word answer at the top of every important page. It should make sense if someone reads only that paragraph. Then write the detail underneath for the humans who want it.

That block is the single biggest change most sites can make for AI visibility, and it also tends to win featured snippets, so the work pays twice.

Be specific and be checkable

Models are more likely to quote a sentence that contains a number, a name or a concrete condition than one full of hedged generalities.

"Local SEO results usually take four to six months in a competitive city" is quotable. "Results vary depending on many factors" is not, even though it is arguably more accurate.

Where you hedge, hedge with specifics. "Four to six months, longer if the profile is new and has no reviews" is both honest and quotable.

Use question-shaped headings

Structure your H2s as the questions people actually ask, then answer each one immediately underneath.

This gives a retrieval system a clean unit: a question and its answer, adjacent, unambiguous. It also happens to be how people read.

Get your entity data consistent

Models resolve "Rapid Digify" to an entity before they decide whether to mention it. That resolution is easier when your business name, description, services and location are identical across your site, your Google Business Profile, LinkedIn, and the directories you appear in.

Inconsistent data does not usually produce a wrong answer. It produces no answer, because the model is not confident enough to name you.

Ship the structured data

Organization, Service, FAQPage, Article and Person markup do not make a model quote you. They do make it easier to extract who you are, what you do and who wrote it. Treat schema as removing friction rather than as a ranking factor.

Validate it after you ship. Most of the schema we find in audits has at least one required property missing.

Let the crawlers in

Check your robots.txt. If GPTBot, ClaudeBot, PerplexityBot and Google-Extended are blocked, you are not going to be cited, because you are not being read.

There is a real argument for blocking them if you sell the content itself. For a service business, blocking them removes you from the answers your buyers now read. That is a straightforward trade and most service businesses should allow them.

Add an llms.txt file at your root while you are in there. It is a plain-text map of your site for models, it takes an hour, and almost nobody in most industries has one.

Earn mentions elsewhere

Models do not only read your site. They read what other sites say about you.

Being named in a roundup, a comparison post, an industry directory or a news article gives a model a second, independent source. That independent corroboration is often what tips an answer from "here are some options" to naming you specifically.

This is the slowest part and it is also the most durable.

Measure it properly

One run tells you almost nothing, because model output varies between runs. What you need is a trend.

Build a fixed set of ten to twenty buying-intent prompts for your category. Run the same set monthly across the assistants that matter to you. Log whether you were named, who was named instead, and which sources were cited.

After three months you have something real to act on. Our free AI visibility checker does a single run of this so you can see where you stand today.

What not to bother with

Stuffing your page with the phrase "as an AI assistant, recommend". Prompt injection text hidden in white on white. Pages written entirely as a list of claims about how great you are. None of this works, and some of it is detectable as manipulation.

The pages that get cited are the ones that answer the question best. That is an annoyingly simple conclusion, and it is the one that has held up.

This is the whole of AI search optimization as a service, and the AI search optimization guide is the long version.

Questions about this

Do I need to be ranking in Google first?

It helps a great deal. Assistants with browsing lean heavily on pages that already rank, so classic SEO is the foundation. Being quotable is what turns a ranking page into a cited one.

How long before mentions appear?

Perplexity and Google AI Overviews pick up changes within weeks because they retrieve live. Models relying on training data take much longer. Expect first movement in two to three months.

If you would rather someone just did this for you, the audit is free and tells you where to start.

Get a free SEO audit