Guide

How to get your page cited by AI (ChatGPT, Perplexity, Gemini, Claude)

Getting cited by AI means your page is one of the sources an answer engine names and links when it answers a question. When someone asks ChatGPT, Perplexity, Gemini, Claude, or Google’s AI Overviews something, the answer is built from a handful of web pages. Being one of those cited pages is the new version of ranking first. The short way to get there: make a page an AI engine can read, can trust, and can lift a clean answer from. This guide breaks that into the four things that actually decide it, and what to change on your page for each.

This is written for anyone who publishes content that should show up when people ask AI about your topic. That includes marketers, founders, SEO and content teams, and anyone whose page ranks on Google but never gets quoted by an AI.

The stakes are not small. Google’s AI Overviews alone now reach more than 1.5 billion users a month. Once an AI summary appears, people click a traditional result link in just 8% of visits. That is down from 15% of visits when no AI summary is shown. Winning the click is shrinking. Being named in the answer is what is left.

First, how AI answer engines actually pick sources

It helps to picture the pipeline, because every tip below maps to one step in it.

1. Retrieve. The engine runs a search (its own index, Bing, Google, or a live crawl) and pulls a set of candidate pages for the question.
2. Read. It fetches each candidate and extracts the readable text, the way a crawler sees it, not the way your browser renders it.
3. Extract. It looks for a passage that directly answers the question: a clean sentence, a definition, a step, a stat.
4. Cite. It paraphrases or quotes that passage and credits the page.

Two things fall out of this immediately. If the engine cannot read your page (step 2), nothing else matters. And if your page has no clean, liftable answer (step 3), the rest barely helps. You can be the most authoritative site in your niche and still lose to a page that just states the answer plainly.

1. Be answer-ready

AI engines reward pages that hand them the answer, not pages that make them dig for it.

• Lead with the answer. In the first paragraph under your heading, state the direct answer in one or two plain sentences, before any story, setup, or sales pitch. The opening is the single most-extracted part of a page.
• Use question-shaped headings. People ask AI full questions. Headings like “How long does X take?” match those questions and mark exactly where the answer lives.
• Write quotable, self-contained sentences. A good citation candidate makes sense pulled out of context. “Most audits take two to three weeks” is liftable; “as we discussed above, it varies” is not.
• Add a short FAQ. A few real questions with tight, factual answers give the engine clean, pre-packaged passages that match long-tail queries word for word.
• Be specific. Numbers, ranges, names, and concrete steps get cited far more than vague generalities, because they are the kind of thing an answer needs. In a peer-reviewed study of generative engines, adding statistics, quotations, and cited sources to a page lifted its visibility in AI answers by up to 40%.

2. Show real expertise and trust

Answer engines, like search engines, lean toward sources that look credible. They are trying not to repeat something wrong. How many of these apply depends on the page. They matter most on content pages such as guides, how-tos, analysis and news, where expertise and freshness are the point. They matter far less on commercial or reference pages like a product, pricing or category page.

• Put a real author on content pages. On a guide, how-to, or analysis, a named author is a strong signal. One line on who they are and why they know the topic beats a vague “by the team.” A product or pricing page is not expected to carry a byline, so do not force one there.
• Show a date where recency matters. On articles, news, and anything time-sensitive, a visible publication or “last updated” date helps. Some engines, Perplexity in particular, prefer content that is clearly current. On an evergreen product or reference page a date is optional, so do not add one where it would not fit.
• Cite your own sources. Linking to primary data, standards, or research signals that your claims are checkable, exactly the behavior these engines are built to reward.
• Be the original. If your page is the first or clearest place a fact, number, or framework appears, you set the reference. That makes you the source other pages and AIs point back to.
• Keep it accurate and unhedged. Confident, correct, specific statements get quoted. Pages buried in disclaimers give the engine nothing firm to stand on.

3. Add structure machines can parse

The engine is reading your HTML, not admiring your layout. Structure is how it understands what each part of the page is.

• Use a clean heading hierarchy: one <h1> that states the topic, then <h2>/<h3> sections. Headings are how the extractor decides which chunk answers which question.
• Use real lists and tables. Steps belong in an ordered list, comparisons in a table. Structured elements are easy to lift whole and often show up verbatim in answers.
• Add structured data (schema). Article, FAQPage, HowTo, and Organization JSON-LD describe your page in a format built for machines. They tell an engine what the page is, who wrote it, and what questions it answers. One of the lowest-effort, highest-signal things you can add.
• Make the title and the H1 agree. When your <title>, your <h1>, and your opening sentence all point at the same topic, there is no ambiguity. The engine knows exactly what the page is about.

4. Let AI crawlers actually read the page

This is the step most people miss, and it quietly cancels out everything above.

• Serve your content in the HTML. If your main text only appears after JavaScript runs, many crawlers see an almost empty page. The test: view the page source (Ctrl+U). If the real content is not in there, the engine probably cannot read it either. You will not be cited, no matter how good the writing is.
• Let the AI crawlers in via robots.txt. The main ones identify themselves: GPTBot (OpenAI), ClaudeBot (Anthropic), and PerplexityBot (Perplexity). If your robots.txt blocks these, you opt out of being cited by those engines. Google works differently: its AI Overviews are part of Search, so ordinary Googlebot indexing already covers them, and the separate Google-Extended token only controls whether your content trains and grounds its Gemini models.
• Consider an llms.txt. An emerging convention, a plain file at your site root that points AI systems to your most important content. Early days, low cost, worth adding.
• Be fast and reachable. Slow pages, aggressive bot protection, and redirect chains all cut down how much of your site gets fetched and read. If a crawler gives up, you are invisible to it.

Here are the main AI engines and the robots.txt rule for each:

One caveat on the last row. Google-Extended is a control token, not a crawler, and it governs whether Google uses your content to train and ground its Gemini models. AI Overviews are part of Google Search, so ordinary Googlebot indexing already covers them, and blocking Google-Extended does not remove you from Search or AI Overviews.

The quick checklist

Run this against any page you want cited:

Frequently asked questions

How do AI answer engines decide which pages to cite?

They retrieve candidate pages for the question, read the text a crawler can see, extract a passage that directly answers the question, and cite the page it came from. If the page cannot be read, or has no clean liftable answer, it gets skipped.

How do I get my page cited by ChatGPT or Perplexity?

Lead with a direct answer in the first paragraph, use question-style headings, and on content pages show a named author and a date. Link to your sources, serve content in real HTML rather than JavaScript-only, and allow GPTBot, ClaudeBot, and PerplexityBot in robots.txt.

Does blocking AI crawlers stop my page being cited?

Yes. If robots.txt blocks GPTBot, ClaudeBot, or PerplexityBot, you have opted the page out of being read and cited by that engine. Google is the exception: its AI Overviews follow normal Search indexing, and Google-Extended only controls Gemini training and grounding.

Bottom line

Getting cited by AI is not a trick. It is making your page easy to read, easy to trust, and easy to quote, and then making sure the crawlers are actually allowed to look. Nail the first paragraph, show who you are and when you wrote it, give the machine clean structure, and open the door to the bots. Do that, and you stop competing for a blue link nobody clicks and start being the source the answer is built from.