What is Generative Engine Optimisation?
Generative search systems do not answer a question with a list of links but with a written response. To do so, they retrieve web pages, summarise their content and sometimes link to their sources. This is where GEO comes in: a company’s content, facts and brand should be prepared and anchored on the web in a way that lets AI systems find them, understand them correctly and include them in their answers.
The term comes from academic research. In November 2023, Pranjal Aggarwal and colleagues published the paper “GEO: Generative Engine Optimization”, which was presented at the KDD conference in 2024. GEO has since become the umbrella term for optimising for AI search. Related terms include Answer Engine Optimisation (AEO) and AI SEO.
What the GEO study found
The authors tested nine ways of rewriting web content, using their own benchmark of queries from many subject areas and the AI search engine Perplexity. According to the abstract, GEO can boost visibility in generative engine responses by up to 40 per cent.
- Three methods performed best: citing sources, adding quotations from credible sources and adding statistics.
- Classic keyword stuffing produced little or no improvement in the tests.
The results come from a test environment reflecting the state of 2023, including a system based on GPT-3.5. Today’s AI search products use different models and retrieval methods. The study therefore shows a direction, not a transferable formula for success.
GEO and SEO: differences and overlaps
Google does not see a contradiction. In its guide to optimising for generative AI in Google Search, Google states that, from the perspective of Google Search, optimising for generative AI search is still SEO. According to Search Central, a page must be indexed and eligible to be shown with a snippet to appear as a supporting link in AI Overviews or AI Mode. Sound search engine optimisation remains the foundation.
In practice, some priorities still shift:
- Goal: a mention and source link within an answer rather than a position in a results list.
- Queries: questions put to AI systems are often longer and conversational. Google may also break them down into sub-questions, see query fan-out.
- Sources: besides your own website, press articles, industry portals, forums and reviews can feed into an answer.
- Measurement: there is no stable position. Instead, you measure shares across many prompts, see AI visibility.
The most important GEO levers
- Technical accessibility: the providers’ crawlers must be able to fetch your content, for example Googlebot, OpenAI’s OAI-SearchBot or Anthropic’s Claude-SearchBot. Key content belongs in the HTML, not only in PDFs or scripts loaded later.
- Unambiguous facts: prices, locations, services and key figures are stated in one place, kept up to date and consistent across pages.
- Answer-ready content: your audience’s questions are answered directly, backed by evidence and first-hand experience. In its guide, Google recommends content with its own perspective rather than interchangeable standard copy.
- Structured data: schema.org markup helps with clarity, but Google says it is not a requirement for generative search.
- Mentions beyond your own website: trade press, associations and comparison sites shape how a brand is described. In its guide, Google says that seeking inauthentic mentions across the web isn’t as helpful as it might seem, because its systems focus on high-quality content and block spam.
- Monitoring: regular queries show whether measures are working and whether AI systems state incorrect information. See AI monitoring.
What GEO cannot do
For outsiders, AI systems are largely a black box, and their answers vary. A study by SparkToro and Gumshoe with almost 3,000 runs found that if you ask ChatGPT or Google’s AI the same question a hundred times, the chance that any two answers contain the same list of brands is below one in a hundred. Nobody can seriously promise a guaranteed mention.
Google’s guide also lists measures that are not necessary for generative Google Search: special AI text files such as llms.txt, breaking content into small chunks and rewriting content specifically for AI.
GEO in practice
The starting point is an inventory: what do customers ask, what do AI systems answer today, and where do errors come from? In one client project, we checked 450 AI answers about a large organisation, 150 prompts each in ChatGPT, Claude and Gemini, comparing every statement with a verbatim passage from the organisation’s website. Only 20 per cent were fully correct, 37 per cent correct but imprecise, 38 per cent partly wrong and 2 per cent wrong. The remaining 3 per cent or so did not deal with the organisation and were not assessed. Around a third of the errors could be traced back to the organisation’s own website: information available only in PDFs or in the online portal, or scattered across many pages, gaps in content and contradictory figures on different pages.
Causes like these can be fixed directly, before budget goes into reach beyond your own website. A GEO audit offers a structured starting point. For an overview of all our services around AI search, see our page on Generative Engine Optimisation.
Would you like to make your brand visible in AI answers and need support? Then get in touch or read more about our GEO services! Would you like to find out more or have your team trained on this topic? Then take a look at our GEO seminar.
