What is GEO? How to get your site cited by ChatGPT, Claude and Perplexity

More and more people ask ChatGPT, Claude, Perplexity or Google's AI answers instead of scrolling ten blue links. Those systems read the web, pick a few sources and write an answer, sometimes with citations. Generative Engine Optimization (GEO) is the practice of making your pages easy for them to find, read, trust and quote.

The term comes from a 2023 research paper, "GEO: Generative Engine Optimization", by researchers at Princeton and other institutions. It found that how content is written and structured changes how often generative engines cite it. The field moves fast, but the fundamentals are stable, and most of them are just good SEO done carefully.

How GEO differs from SEO

Traditional SEO is about ranking a URL for a query. GEO is about being used as a source inside a generated answer. The two overlap heavily, because most AI answer engines pull candidates from a search index first. But a few things matter more in GEO:

• Extractability. Can a model lift a clear, self-contained answer from your page? Answers buried in long intros or spread across tabs are harder to quote.
• Raw HTML access. Many AI crawlers do not run JavaScript. If your content only appears after client-side rendering, it may as well not exist for them.
• Crawler access. AI companies use their own user agents. If robots.txt or your CDN blocks them, you are invisible to that engine.
• Specificity. Concrete numbers, named steps, and definitions are more quotable than marketing language.

Know the crawlers

Each AI company runs more than one bot, and they do different jobs. Most separate training (collecting data to train future models) from search and retrieval (fetching pages to answer a question right now):

• OpenAI: GPTBot (training), OAI-SearchBot (ChatGPT search), ChatGPT-User (fetches a page when a user asks).
• Anthropic: ClaudeBot (training), Claude-SearchBot (search), Claude-User (user-initiated fetches).
• Perplexity: PerplexityBot (indexing), Perplexity-User (user-initiated fetches).
• Google: Googlebot crawls for Search, including AI Overviews. Google-Extended is a robots.txt token, not a separate crawler, that controls whether your content is used for Gemini models.
• Apple: Applebot, with Applebot-Extended controlling use in Apple's AI models.
• Common Crawl: CCBot, whose public dataset is used by many model builders.

If you want to show up in AI answers, the search and user bots matter most. Whether you also allow training bots is a business decision. Allowing them is how a small brand ends up "known" to future models.

A practical GEO checklist

1. Let them in

Check robots.txt and your CDN or firewall. Some hosts block AI crawlers by default with a toggle that robots.txt cannot override. Test with a request using the crawler's user agent and confirm you get a 200, not a 403 or a challenge page.

2. Serve real HTML

Use static generation or server rendering for every public page. View the page source. Your headline, product description and pricing should be there as text.

3. Answer the question in the first two sentences

Put the definition or the answer at the top of the section, then explain. "Web Scan is a site-health checker that runs 85+ checks on any URL" is easier to quote than three paragraphs of build-up.

4. Use headings that match real questions

Write ## How much does it cost? rather than ## Our philosophy. Question-shaped headings line up with how people prompt AI tools, and FAQ sections give models ready-made question and answer pairs.

5. Be specific and consistent

State facts the same way everywhere: product names, prices, limits, what is live and what is not. Inconsistent claims across your site, docs and social posts make it harder for a model to trust any one version.

6. Add structured data

JSON-LD for Organization, SoftwareApplication, FAQPage and BlogPosting gives machines an unambiguous summary of who you are and what you offer.

7. Publish an llms.txt

An /llms.txt file is a short Markdown briefing about your site for language models. It is a proposed convention, not a standard, and support varies, but it costs ten minutes. We wrote a separate guide to llms.txt.

8. Earn mentions elsewhere

Answer engines weigh what other sites say about you: reviews, forum threads, comparison posts, docs that link to you. GEO does not replace being worth talking about.

How to measure it

There is no Search Console for AI answers yet. Practical signals:

• Ask the engines directly what your product is, and note whether they cite you.
• Watch referral traffic from chatgpt.com, perplexity.ai, claude.ai and similar domains in your analytics.
• Check your server or CDN logs for the AI user agents above, to confirm they are actually crawling.

Check your site in one click

Web Scan includes GEO checks alongside SEO, security and performance. It looks at what an AI answer engine can actually read on your page, flags what is missing, and gives you a fix you can paste into Claude, Cursor or another coding agent. No account needed.