Guide · Search
Generative Engine Optimization: GEO, AEO and AI Search
Generative engine optimization, or GEO, means shaping web content so that AI answer tools, such as AI search summaries and chat assistants, are more likely to use and cite it. The term comes from a 2023 research paper. This guide explains what that paper found, how GEO relates to SEO and AEO, what Google says, and which agency promises to doubt.
What GEO means
A generative engine is a search tool that writes an answer instead of only listing links. It gathers sources, then uses a large language model to summarise them. Examples include AI answers inside search engines and chat assistants that browse the web.
GEO is the work of making your content more visible inside those written answers. "Visible" here means two things: your page is cited, and your content shapes what the answer says.
A note on spelling: this guide uses British spelling, but "generative engine optimization" keeps its "z" because that is how the original paper and most of the industry write the name.
Sources: Aggarwal et al., GEO: Generative Engine Optimization, arXiv 2311.09735; Wikipedia, Generative engine optimization
Where the term came from
The name comes from a paper titled "GEO: Generative Engine Optimization" by Pranjal Aggarwal and five co-authors. It was first posted on arXiv on 16 November 2023. The third version followed on 28 June 2024, and the paper was accepted to KDD 2024, a major data science conference.
The authors built a test set called GEO-bench, with 10,000 queries from nine sources. They then changed source pages in set ways and measured how much of each page showed up in the AI answer.
Sources: Aggarwal et al., GEO: Generative Engine Optimization, arXiv 2311.09735
What the research found
The headline number is that GEO methods could boost visibility by up to 40% in generative engine responses. That figure comes from the authors' own test engine. On Perplexity, a live AI search tool, the gains were up to 37%.
The methods that helped most were simple and honest:
Method tested in the paper What it changes Result in the paper Cite Sources Adds citations from credible sources Among the strongest Quotation Addition Adds relevant quotes from credible sources Among the strongest Statistics Addition Uses numbers instead of vague claims Among the strongest Keyword Stuffing Repeats more search keywords Little to no gain; on Perplexity, 10% worse than doing nothing The authors also found that results varied by subject. A method that worked for one domain did less for another.
Read the numbers with care. Most tests ran on a benchmark and test engine the authors built themselves. Real AI search systems change often and do not publish how they choose sources.
Sources: Aggarwal et al., GEO: Generative Engine Optimization, arXiv 2311.09735; Aggarwal et al., GEO paper full text (PDF)
GEO, AEO and SEO compared
Three labels get mixed up. Here is how they are usually used.
Term Short for Usual goal SEO Search engine optimisation Rank pages in classic search results and earn clicks AEO Answer engine optimisation Be the direct answer to a question, such as in a featured answer or voice reply GEO Generative engine optimization Be cited or used in AI-written answers In practice, they overlap a great deal. Google's advice for its AI features is its normal SEO advice, and the GEO paper's best tested methods were adding citations, quotations and statistics. Neither source promises a result. Our SEO vs SEM guide covers how organic and paid search differ.
Sources: Google Search Central, Optimizing for generative AI features; Aggarwal et al., GEO: Generative Engine Optimization, arXiv 2311.09735
What Google says
Google has a guide for site owners on its generative AI features. It mentions AEO and GEO by name. It says many suggested "hacks" for them are not effective or supported by how Google Search works.
Google lists things you can ignore for its Search:
- LLMS.txt and other special files: Google Search does not use them. Making one neither helps nor harms your Google visibility, though other services may read them.
- Chunking content: there is no need to split pages into tiny pieces. Google says there is no ideal page length.
- Rewriting just for AI: AI systems understand synonyms and meaning, so you do not need every keyword variant.
- Chasing fake mentions: seeking inauthentic mentions across the web is not as helpful as it seems, and spam systems apply.
Google's main advice is to create valuable, "non-commodity" content: content that people find unique, compelling and useful. Its separate page on AI features says there are no extra technical requirements to appear in AI Overviews or AI Mode beyond normal Search ones: a page needs to be indexed and eligible to show with a snippet. Google's newer guide then sets a condition at site level, tied to Search Console: a site needs inclusion in the generative AI features of Search before its pages are eligible.
These rules are for Google. Other AI tools may work differently, and few publish their rules.
Sources: Google Search Central, Optimizing for generative AI features; Google Search Central, AI features and your website
Claims to test before you hire a GEO agency
Agencies in Singapore and elsewhere now sell GEO. Some offers are sound. Some promise what no one can deliver. This table sets common pitches against the evidence above.
What the pitch says What the evidence says Question to ask "We guarantee you will be cited in AI answers" Google says meeting every rule does not guarantee a place What happens to the fee if we are not cited? "GEO lifts visibility 40%" Up to 40% in a lab test of single pages; up to 37% on Perplexity Which engine, which queries, and what was our baseline? "You need an llms.txt file for Google" Google Search ignores it Which AI service will read it, and how will we know? "We will add your keywords everywhere" Keyword stuffing gave little or no gain, and worse on Perplexity How will you add sources, quotes and data instead? "We will get you mentioned on many sites" Google says inauthentic mentions are not that helpful Are these real, earned mentions? "We track your AI visibility" Possible, but tools sample a set of prompts Which prompts, how often, and can we see the raw list? A fair GEO offer looks a lot like good SEO and good writing. It improves facts, sources and clarity, then measures with a method you can check. The editors' notes on how Best SEO Campaign is selected and the public methodology show the kind of evidence worth asking for. Our AI Overviews guide explains how Google's own AI answers pick sources.
Sources: Google Search Central, AI features and your website; Google Search Central, Optimizing for generative AI features; Aggarwal et al., GEO paper full text (PDF)
Frequently asked questions
What is generative engine optimization?
Generative engine optimization, or GEO, is the work of making content more likely to be used and cited in AI-written answers. The term comes from a 2023 research paper by Aggarwal and colleagues.
Is GEO replacing SEO?
No. Google says its normal SEO best practices still apply to its AI features, and no special hacks are needed. On top of normal Search rules, there is one site level condition tied to Search Console: the site needs inclusion in the generative AI features of Search. GEO mostly builds on good SEO.
Is generative engine optimization a real thing?
Yes, as a research idea and a marketing practice. The 2023 GEO paper measured visibility gains of up to 40% in its tests, but Google warns that many GEO "hacks" are not effective for its Search.
What is the difference between GEO and AEO?
AEO, answer engine optimisation, usually aims to be the direct answer to a question. GEO aims to be cited or used inside AI-written answers. Both overlap heavily with SEO.
Does an llms.txt file help with Google?
No. Google says its Search does not use llms.txt or similar files, so adding one neither helps nor harms Google visibility. Other services may use them.
What GEO methods worked best in the research?
Adding citations, quotations from credible sources and statistics. Keyword stuffing gave little or no improvement and did worse than doing nothing on Perplexity.
Sources