Generative engine optimization (GEO): getting cited by AI

What is generative engine optimization?

A generative engine writes an answer instead of listing links. When someone asks ChatGPT or Perplexity for the best tool for a job, the reply may name a few brands and cite a few pages. GEO is about being one of them.

The term comes from a 2023 research paper, GEO: Generative Engine Optimization, by Aggarwal and colleagues, later accepted at KDD 2024. The authors tested ways of editing source pages and found some methods raised a source's visibility in generated answers by up to 40 percent.

There are two outcomes to aim for. A mention is when the answer names your brand. A citation is when the answer links to your page as a source. You can get one without the other, so track both.

How do AI engines find and choose sources?

Answers come from two places. The first is what a model learned during training. The second is live retrieval: the engine searches an index at question time and reads the pages it finds. Retrieval is where most GEO work pays off, because it can change in days rather than waiting for the next model.

Each company runs its own crawlers, often with separate bots for training, search indexing and user-triggered fetches. They are controlled through robots.txt.

• OpenAI: OAI-SearchBot surfaces sites in ChatGPT search results. GPTBot crawls content that may be used to train models. ChatGPT-User fetches a page when a user's request calls for it.
• Anthropic: ClaudeBot collects content that may be used for training. Claude-SearchBot indexes content to improve search results. Claude-User fetches pages when people ask Claude questions.
• Perplexity: PerplexityBot builds Perplexity's search index and, per Perplexity, is not used to train foundation models. Perplexity-User fetches pages for a user's question.
• Google: Gemini can ground answers in the Google Search index. The Google-Extended robots.txt token controls whether content is used for Gemini training and grounding, and Google says it does not affect Search inclusion or ranking.

Why server-rendered HTML matters for AI crawlers

Googlebot renders JavaScript. Many AI crawlers do not. A December 2024 study by Vercel of its network traffic found that none of the major AI crawlers it measured, including GPTBot, ClaudeBot and PerplexityBot, rendered JavaScript. They fetched script files but did not run them.

If your content, titles or schema only appear after client-side JavaScript runs, those crawlers may see an empty shell. Check by loading a page with JavaScript disabled, or by fetching the raw HTML with a command-line tool. What you see there is roughly what a non-rendering bot sees.

SEOAmazing applies approved changes with a pixel, which Google can see because it renders JavaScript. For AI crawlers, SEOAmazing's prerender service serves fully rendered HTML with the approved changes applied. On a customer site this needs bot traffic routed to the prerender with a small edge rule, so it is set up with our help, not switched on automatically. seoamazing.com runs this way.

GEO tactics that make pages easier to cite

The GEO paper found that adding quotations, adding statistics and citing sources gave the largest gains in its tests, while keyword stuffing did not help. That matches a simple idea: engines prefer passages that carry evidence and can be quoted whole.

1. Allow the bots you want. Check robots.txt for rules that block OAI-SearchBot, Claude-SearchBot or PerplexityBot. Decide separately whether to allow training bots like GPTBot and ClaudeBot.
2. Serve readable HTML. Put the main text, headings and structured data in the initial HTML response.
3. State who you are. Use one consistent name, a one-sentence description of what you do, and Organization schema. Repeat the same facts on your about page and profiles.
4. Write quotable facts. Short sentences with a specific number, date or definition are easy to lift into an answer.
5. Publish original data. Surveys, benchmarks and pricing studies give engines a reason to cite you rather than a summary of you.
6. Cite your own sources. Link to primary documents for claims, so your page reads as checked.
7. Keep pages fresh. Show an updated date and revise facts that change, such as prices and version numbers.
8. Earn third-party mentions. Reviews, comparisons, forums and press coverage that name your brand shape what models say about you.

How GEO differs from SEO and AEO

SEO earns ranked links in a search engine. AEO earns the extracted answer in features like featured snippets and People also ask. GEO earns a mention or citation inside an answer that an AI model writes from several sources.

The foundations overlap: crawlable pages, clear structure and trust. The differences are in the details. GEO cares more about how your brand is described across the web, because a model may name you without linking to you. It also cares more about raw HTML, since many AI crawlers skip JavaScript. And its results vary from one run to the next, so measurement needs repeated checks.

How to measure GEO results

Pick the questions buyers ask, run them across the engines that matter, and record whether your brand is mentioned and whether your domain is cited. Repeat on a schedule and watch trends, since a single answer proves little.

SEOAmazing's AI Visibility checks whether your brand or domain is mentioned or cited in answers from ChatGPT, Claude, Gemini and Perplexity for the questions you choose, on a schedule. The AI visibility tracking guide covers the method in detail.

Frequently asked questions