What is GEO? Generative engine optimisation explained

What is GEO?

GEO, short for generative engine optimisation, is the practice of making your content easy for AI answer tools to find, understand and cite. Those tools include Google's AI Overviews and AI Mode, and assistants such as ChatGPT and Perplexity. Instead of a list of ten links, the reader gets a written answer, sometimes with links to the pages it drew on.

You will also see AEO, answer engine optimisation. In practice the two terms describe the same thing, and neither is a standard with an official definition.

One warning about wording. "AI optimisation" is used for two different jobs. One is optimising your site for AI answers, which this article covers. The other is using AI to improve how a business runs, such as automating a process or building an assistant. That second job is AI development, and it has nothing to do with search.

How is GEO different from SEO?

For Google, it largely is not. Google's own guidance says its AI features rely on the same core ranking systems as normal search, and that you do not need special optimisation or extra technical requirements to appear in them.

What changes is how results are presented and measured, not what makes a page good.

The last row matters. Google publishes what it expects. Most other assistants publish far less, so claims about exactly how they pick sources are usually educated guesses.

How do AI answers choose their sources?

Google describes two techniques. The first, retrieval-augmented generation, means the system finds relevant pages in the search index and uses them to ground its answer. The second, query fan-out, means it runs several related searches across subtopics to gather more material.

Put simply, an AI answer is built from pages the system can already find. That is why the starting requirements are ordinary ones. For Google's AI features, a page needs to be indexed and eligible to appear in search with a snippet. Google states there are no additional technical requirements. Meeting them still does not guarantee you will be shown.

What actually helps?

Most of what helps is what already helps in search:

• Be findable. Pages must be crawlable and indexable, and important content should be in the HTML, not hidden behind scripts that only run later.
• Answer the question early. Put the answer in the first two sentences of a section, then explain. People and machines both quote clear answers.
• Say something original. Content that repeats what ten other pages say gives a system no reason to pick yours. A real point of view, your own data or first-hand experience does.
• Show who is behind it. Name the company, say what it does, and make contact details and authorship easy to find.
• Keep facts consistent. Your name, services and claims should agree across your site and your public profiles.
• Look after page experience. A fast, readable page on every device is still part of the picture.
• Keep structured data accurate. It is not required for AI features, but it is still worth having where it matches what is on the page.

If a page would be a poor answer for a person who found it in a normal search, it will not become a good source because you added the word "AI" to it.

What is a waste of time?

Google is direct about several tactics, and they are worth avoiding:

• llms.txt files and other special AI markup. Google says Search ignores them. Publishing one does no harm for other tools, but do not expect it to change anything in Google.
• Cutting content into tiny pieces for machines. Google says there is no requirement to do this.
• A separate page for every possible variation of a question. At scale this risks being treated as spam.
• Buying or faking mentions around the web. Inauthentic mentions are not a shortcut.
• Tools that claim to show Google's internal ranking or AI data. Google says no third-party tool has access to it.

If someone promises guaranteed placement in AI answers, treat that as a warning sign, for the same reason you would with guaranteed rankings.

How can you measure it?

There is no perfect measure, so combine three imperfect ones.

1. Search Console. Google says traffic from its AI features is counted within the Web search type in the Performance report, so you cannot split it out cleanly.
2. Your own analytics. Look for visits referred from AI assistants and watch whether they behave differently from search visitors.
3. A fixed list of real questions. Pick 10 to 20 questions customers actually ask, check them on a regular schedule, and record whether you are cited. Answers vary between runs, so look for a trend, not a single result.

Should you block AI crawlers?

That depends on what you want. Some bots collect content to train models. Others fetch pages in order to answer a user's question and link back. Blocking the first group does not remove you from Google Search or its AI features. Blocking the second can mean those assistants cannot cite you.

Check your robots.txt matches your decision. If you want to limit how much of a page Google can show, the existing controls such as nosnippet and max-snippet apply, but they also reduce how your content can appear in search.

Where should you start?

Start with a short list rather than a programme:

1. Check your important pages are crawlable and indexed.
2. Rewrite your key pages so each opens with a clear answer to the question it targets.
3. Add the company details and authorship a stranger would look for.
4. Decide your AI crawler policy and make robots.txt match it.
5. Set up the question list and check it monthly.

That is the core of our SEO and GEO services. If you run an online shop, the same steps apply to product and category pages, covered in ecommerce SEO.