What is GEO? A guide to Generative Engine Optimization

How do AI answer engines choose which sites to cite?

Most AI assistants answer current questions by running a web search, fetching a few top pages, and writing a summary that cites them. A page is cited when it is reachable by the bot, ranks for the underlying search, and contains a clear, specific passage that answers the question.

1. The user asks a question; the assistant rewrites it into one or more search queries.
2. It retrieves results from a search index (its own, Bing, Google or a partner).
3. It fetches the pages, splits them into passages and picks the most relevant ones.
4. It writes the answer and links to the pages it drew from.

Two things follow. First, the passage matters more than the page: a crisp definition or statistic in one paragraph is easier to use than the same fact spread across five. Second, if a bot cannot fetch the HTML — because of robots.txt, a firewall or content that only appears after JavaScript runs — you cannot be cited.

What does research say actually improves AI visibility?

The 2023 paper “GEO: Generative Engine Optimization” by researchers from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi tested nine content changes and found that adding citations, quotations and statistics improved visibility in generative answers by up to 40%.

• Cite sources — link to the primary evidence behind your claims.
• Add quotations from recognised experts or official bodies.
• Add statistics — specific numbers with a date and a source.
• Write fluently and clearly; technical terms where your audience expects them.
• Keyword stuffing performed poorly in the study — it does not help with AI engines.

Which AI crawlers should I allow?

To be cited, allow the retrieval bots that fetch pages for answers — OAI-SearchBot and ChatGPT-User (OpenAI), PerplexityBot, Claude-SearchBot and Claude-User (Anthropic), plus Googlebot and Bingbot. Training-only bots such as GPTBot, ClaudeBot and Google-Extended are a separate choice.

How do I make AI systems understand who my business is?

Describe your organisation the same way everywhere: identical name, address, phone and description on your site, Google Business Profile, LinkedIn, directories and press mentions, and connect them with Organization schema using the sameAs property.

• An About page that states what you do, where, since when, and who leads it.
• Organization schema with logo, address, contact point and sameAs links to official profiles.
• Named authors with a short bio and credentials on every guide.
• Consistent spelling of your brand — AI systems merge or split entities based on these signals.

Do I need an llms.txt file?

No, but it does not hurt. llms.txt is a proposed standard from 2024: a Markdown file at /llms.txt that summarises your site for language models. Google states it does not need such files for AI Overviews, and other major providers have not confirmed they use it.

This site generates /llms.txt and /llms-full.txt automatically from the same content as the pages, so it stays accurate with zero extra effort. Treat it as a low-cost extra, not a ranking lever.

How do I measure GEO results?

Track referral visits from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai in your analytics, watch Search Console (Google includes AI Overview and AI Mode traffic in its Web totals), and regularly ask the assistants your customers’ questions to see who gets cited.

• Create an analytics segment for AI referrers and watch it month over month.
• Keep a list of 20–30 buyer questions and test them in each assistant once a month.
• Record which pages and competitors are cited and update your weakest answers first.