GEO SEO: A Practical Guide to Generative Engine Optimisation

In short: GEO (generative engine optimisation) helps AI search tools such as ChatGPT, Perplexity and Google's AI Overviews understand your website, use it as a source and cite it. GEO does not replace search engine optimisation; it builds on it. If Google cannot find and understand a page, most AI answers will miss it too.

This guide explains how AI answers are produced, what changes compared with traditional SEO and which seven steps prepare your website. Where we rely on research or on what the platforms themselves say, the source is linked. Last reviewed: October 2026.

What is GEO?

GEO stands for generative engine optimisation. The term was coined in the paper "GEO: Generative Engine Optimization" by Aggarwal and colleagues, published at the KDD conference in 2024. It describes optimising content for systems that write their own answers instead of only listing links.

The main goal is therefore not a position in a results list. You want your content to feed into the answer and your page or brand to appear as a cited source.

Several terms are used for roughly the same thing: GEO SEO, AI SEO, LLM optimisation or AEO (answer engine optimisation). GEO has nothing to do with geography, by the way. For visibility in a particular area, see local SEO.

How AI search tools choose their sources

A simple distinction helps in practice:

• Answers from training data: the language model answers from what it learned during training. New content on your website only reaches it in later model versions, if at all.
• Answers with web search: ChatGPT with search, Perplexity and Google's AI Overviews retrieve current pages for the question, summarise them and link to sources. The technical term is retrieval-augmented generation (RAG).

The second route is the one you can influence. To be considered, your page has to be accessible to the relevant search crawler, match the question and provide an answer that is easy to quote.

The platforms do not disclose exactly how they weigh sources. What you can rely on are their statements about crawlers and basic requirements, plus the results of controlled studies. This guide is based on both.

For how this works in ChatGPT specifically, and what OpenAI itself says, read our article on ChatGPT SEO.

GEO and SEO: what stays the same and what is new

In its documentation on AI features, Google states that there are no additional requirements and no special optimisations needed to appear in AI Overviews or AI Mode. The usual SEO fundamentals apply.

These stay the same:

• pages that are technically accessible and indexable,
• content that genuinely answers a search query,
• a clear structure with meaningful headings,
• trust built through expertise, evidence and mentions on other websites.

These become new or more important:

• Quotability: short, unambiguous statements that make sense without the rest of the page.
• Evidence in the text: figures, sources and quotations. In the GEO study, exactly these changes raised visibility in the tested answers by 30 to 41 per cent.
• More than one crawler: besides Googlebot, crawlers from OpenAI and Perplexity decide whether your page appears in their search.
• Different measurement: a mention in an AI answer does not always lead to a click. Track visibility and enquiries separately.

What the GEO study shows

The researchers used their own test setup to examine which changes to a text increase its share of generated answers. The main results:

• Adding quotations: up to 41 per cent more visibility, measured by the source's share of the answer text.
• Adding statistics and citing sources: around 30 to 40 per cent each.
• Improving fluency and readability: 15 to 30 per cent.
• Keyword stuffing: did not help and sometimes performed worse than the unchanged text.

Particularly relevant for smaller websites: sources that ranked lower in traditional search benefited far more. With the "cite sources" method, visibility for the source in fifth place rose by 115 per cent, while the top-ranked source lost share.

For context: these results come from the researchers' test setup and are not a promise for ChatGPT or Google. They do show clearly which direction your content should take.

GEO in seven steps

1. Describe your business and offer unambiguously

AI systems need to understand who you are, what you offer and for whom. Describe services, scope, audience and service area clearly on every relevant page. Keep your company name, offer and contact details consistent across your website, profiles and directories. Contradictions make it harder to attribute information to you.

2. Check access for search crawlers

Check your robots.txt and the actual server responses to see which crawlers are allowed. The key differences:

• Google: AI Overviews and AI Mode follow the normal rules for Googlebot. You control how content appears with familiar directives such as noindex or nosnippet.
• OpenAI: OAI-SearchBot powers search in ChatGPT, GPTBot collects content for model training. According to OpenAI, sites that block OAI-SearchBot will not be shown in ChatGPT search answers. Changes to robots.txt take effect after about 24 hours.
• Perplexity: PerplexityBot indexes pages for Perplexity search and respects robots.txt. Perplexity-User fetches pages for specific user questions and generally ignores robots.txt.

So you can opt out of training by blocking GPTBot and still stay visible in ChatGPT search by allowing OAI-SearchBot. Also make sure your main content is in the HTML the server delivers. Not every crawler runs JavaScript.

3. Answer questions directly

AI answers start with questions. Collect your customers' questions from conversations, enquiries and Search Console, and answer each one in its own section. Put the core statement in the first one or two sentences and the detail after it. That way the section can be quoted on its own.

4. Add evidence, figures and sources

Back up statements with concrete figures, examples from projects and links to primary sources. Only state what you can support: results need a time period, a measurement source and a link to the project. In the GEO study, this kind of evidence increased visibility the most.

5. Build mentions on other websites

AI search does not rely only on your own website. Trade articles, industry directories, comparison lists, review platforms and interviews show that others know you as a provider. Plan these mentions as deliberately as backlinks and focus on sites that fit your topic.

6. Keep structured data and business details consistent

Structured data (Schema.org markup) describes your organisation, services, people or FAQs in a machine-readable way. Google explicitly says it is not a requirement for AI Overviews. It still helps to make facts unambiguous, as long as it matches the visible content.

7. Keep content current and link it internally

Outdated prices, services or dates make a page an unreliable source. Review important pages regularly and show when they were last updated. Link guides to the relevant service pages and back, so that people and crawlers can see how they connect.

Do I need an llms.txt file?

An llms.txt is a text file in the root of a website that gives language models a compact overview of its most important content. Jeremy Howard proposed it in September 2024. It is a community proposal, not an official standard.

Google does not list any AI-specific file as a requirement for its AI features. An llms.txt is quick to create and does no harm, but it does not replace any of the seven steps. If you publish one, make sure the information in it is accurate and kept up to date.

How to measure AI visibility

Rankings can be counted; mentions in AI answers can only be observed. Three sources complement each other:

• Manual tests: ask your customers' most important questions in ChatGPT, Perplexity and Google on a regular schedule. Record the date, tool, question, answer and cited sources. Answers vary, so repeated tests matter more than any single result.
• Google Search Console: clicks and impressions from AI Overviews are included in the Performance report under "Web", but are not reported separately.
• Web analytics: referrals from chatgpt.com, perplexity.ai and similar services show whether mentions turn into visits.

What matters in the end is qualified enquiries, not the number of mentions. Track both separately so you can see which content actually leads to conversations.

For a step-by-step method with a question set and log template, read our guide How to measure AI visibility.

How long does GEO take and what does it cost?

At OpenAI, changes to robots.txt take effect after about a day. Depending on the tool, new content takes days to weeks to be found and used. Mentions on other websites and trust build up over months. No one can honestly guarantee that you will be cited.

Costs depend on how much content and technical work is needed. At Con Radius, ongoing SEO and GEO support starts at EUR 350 per month, with the exact scope agreed in advance. What we take on as a GEO agency is described on the service page.