AI Search 101 · Generative Engine Optimization
What is GEO?
- Last updated
- First published
- Sources
- 6 sources
- Reading time
- 6 min read
Generative Engine Optimization (GEO) is the practice of shaping content so that generative AI search engines — systems like Google AI Mode, ChatGPT search and Perplexity, which write an answer from the sources they retrieve — use it, quote it and cite it.
The term comes from a November 2023 paper led by researchers at Princeton University and IIT Delhi, published at KDD 2024, which measured how nine kinds of content edits changed a source's visibility inside generated answers.[1] It has since become a common name, alongside AEO, for optimizing for AI search — which Google counts as SEO.[2]
Key takeaways
- GEO is about being a source a generative engine draws on and credits: your facts, your words and your link inside its answer.
- In the original study, adding citations, quotations and statistics raised a source's visibility in generated answers by up to about 40%. Keyword stuffing did little or nothing.[1]
- The gains were largest for pages that ranked lower in search: citing sources lifted the fifth-ranked result's visibility by 115%.[1]
- Search rank and AI citation are separating. In early 2026, only 37.9% of the URLs cited in Google's AI Overviews also ranked in its top 10.[3]
- The study used a lab engine built on GPT-3.5, plus a test on Perplexity. Read its numbers as direction, not a guarantee on today's engines.[1]
What does GEO stand for?
GEO stands for Generative Engine Optimization.
- GenerativeThe engine writes new text. It summarizes, paraphrases and quotes several sources at once, instead of returning a ranked list of pages.
- EngineSearch systems that pair retrieval with a language model: Google's AI Overviews and AI Mode, ChatGPT search, Perplexity and Microsoft Copilot.
- OptimizationChanging your content so the engine is more likely to use it and cite it — measured by how much of the answer your source supplies.
How does GEO work?
A generative engine retrieves a handful of sources for a question, then has a language model write one answer from them, citing some along the way. Which sources it picks, and how much of each it uses, decides who is visible. A page can be retrieved and still contribute nothing to the answer.
The GEO paper measured that contribution directly. Its authors built GEO-bench, 10,000 queries drawn from nine sources. On its 1,000-query test set they answered each question from the top five Google results, rewrote one source at a time, and recorded how much more of the answer came from it, weighted by how early in the answer it appeared.[1]
Retrieve
The engine searches for the question — in Google's case, also a fan-out of related sub-questions — and collects candidate sources.[4]
Synthesize
A language model writes one answer from those sources, quoting some passages, paraphrasing others and leaving the rest out.
Attribute
The answer links some of the sources it used. GEO is the work of being one of them, and of supplying more of the answer.
GEO vs SEO: what's the difference?
SEO decides whether a page is found; GEO is about how much of it makes it into the answer. Google's AI features only use pages that are indexed and eligible to show a snippet, so the first is a prerequisite for the second.[4]
| Dimension | GEO | SEO |
|---|---|---|
| Optimizes for | Being used and cited inside a generated answer | Ranking position in a list of results |
| Success looks like | A large share of the answer, and the citation | A top position, impressions and clicks |
| What moved it in research | Citations, quotations, statistics and clear writing[1] | Relevance, quality, links and technical health |
| What the user reads | The engine's synthesis of your content | Your page, after clicking through |
| Origin | Academic research, 2023 | The search industry, 1990s |
What the GEO research found
The paper tested nine edits on the same sources. Three stood out, two helped less, and two — keyword stuffing and an authoritative tone — made little or no significant difference.[1]
Quotations
Adding relevant quotes from credible people scored highest in the main benchmark, and was the best performer on Perplexity, where it raised visibility by 22%.[1]
Statistics
Replacing general claims with specific numbers was one of the top three edits, with 30–40% relative gains in the main benchmark.[1]
Citing sources
Citing credible sources completed the top three — and helped lower-ranked pages most, lifting the fifth-ranked result by 115%.[1]
Fluent, easy-to-read text
Improving fluency and simplifying the language gave smaller but real gains, around 15–30%.[1]
An authoritative tone: no clear effect
The paper reports no significant improvement from rewriting a source to sound more persuasive and authoritative.[1]
Keyword stuffing: no help
Adding more of the query's keywords gave little or no improvement, and scored about 10% below the baseline on Perplexity.[1]
GEO by the numbers
From the original study, and from what has happened to AI search since.
- 10,000queries in GEO-bench, the benchmark the original study introducedAggarwal et al., KDD 2024, Nov 2023, revised Jun 2024 [1]
- ~40%the largest average visibility lift the study reported, from adding citations, quotations or statisticsAggarwal et al., KDD 2024, Nov 2023, revised Jun 2024 [1]
- +115%visibility for the fifth-ranked source after it cited its sources; the same edit cut the top-ranked source's by 30%Aggarwal et al., KDD 2024, Nov 2023, revised Jun 2024 [1]
- 37.9%of URLs cited in Google AI Overviews also ranked in its top 10 in early 2026, down from about 76% in mid-2025Ahrefs, Mar 2026 [3]
How to apply GEO
Start with pages that are already found
An engine can only quote what it retrieves. Begin with pages that already earn impressions for the questions you care about.
Add the evidence
Put specific figures, named sources and attributable quotes beside the claims they support — the three edits that moved visibility most in the study.[1]
Write plainly
Improving fluency and readability gave a measured 15–30% gain in the study.[1]
Don't rewrite for the robot
Google's 2026 guide says you don't need to write in a special way for AI search, and that pages made for fan-out queries to manipulate its answers break its spam policies.[2]
Measure citations, not just rankings
Track which questions cite your pages on each engine, and compare that with where they rank — the two now diverge.[3]
GEO questions, answered
What does GEO stand for?
GEO stands for generative engine optimization: shaping content so that generative AI search engines, which write an answer from the sources they retrieve, use it and cite it. The term comes from a 2023 research paper by Aggarwal and colleagues, published at KDD 2024.
Is GEO the same as AEO?
They describe the same goal today. GEO comes from academic research on generative search; AEO is a marketing term from the voice-search era around 2018 that now covers AI assistants. Google's 2026 guidance names both and says that, for Google Search, both are still SEO.
What did the original GEO paper find?
Testing nine content edits on the 1,000-query test set of its 10,000-query benchmark, it found that adding quotations, statistics and citations to credible sources raised a source's visibility in generated answers by up to about 40%, with the biggest gains for lower-ranked pages. Keyword stuffing gave little or no improvement, and an authoritative tone made no significant difference.
Do the GEO paper's results hold on ChatGPT or Google today?
Treat them as direction, not a guarantee. The study used an engine built on GPT-3.5 and a 200-query test on Perplexity; today's engines use newer models and different retrieval. Its broad lesson — specific, well-sourced, clearly written content gets used more — is consistent with Google's and Microsoft's published guidance.
Does GEO replace SEO?
No. A generative engine can only use pages it can find, and Google's AI features only draw on pages that are indexed and eligible to show a snippet. GEO adds a second question on top of SEO's: once your page is retrieved, how much of the answer does it supply?
How do you measure GEO?
The paper measured the share of an answer's words that came from a source, weighted by how early they appeared. In practice, teams track whether their pages are cited, and whether their brand is mentioned, across a fixed set of questions on each engine, and watch how that changes over time.
Why did lower-ranked pages gain most in the GEO study?
The study found its edits helped sources further down the search results the most. Citing sources raised the fifth-ranked result's visibility by 115%, while the same edit applied to the top-ranked result lowered its visibility by about 30%. The authors attribute the gap to traditional rankings depending on other signals, such as backlinks.
Sources
Every figure on this page was checked against its source on .
- 1.Aggarwal et al., KDD 2024. GEO: Generative Engine Optimization. Nov 2023, revised Jun 2024.
- 2.Google Search Central. Optimizing your website for generative AI features on Google Search. May 2026, updated Jul 2026.
- 3.Ahrefs. Update: 38% of AI Overview Citations Pull From The Top 10. Mar 2026.
- 4.Google Search Central. AI features and your website. Dec 2025.
- 5.Wikipedia. Generative engine optimization. Oct 2026.
- 6.Microsoft. Optimizing Your Content for Inclusion in AI Search Answers. Oct 2025.