# GEO is about being a source worth using.
> What GEO is, where the term comes from, what the research measured, what Google and OpenAI document, and what GEO changes in practice. No guarantees.
- Page: https://mohammedteto.com/generative-engine-optimization/
- Language: en — French version: https://mohammedteto.com/fr/optimisation-moteurs-generatifs/
- Author: Mohammed Teto (https://mohammedteto.com/#person)
- Published: 2026-10-07
- Updated: 2026-10-04
A generative engine does not rank ten links. It reads several sources and writes one answer. GEO is the work of making your content, and what others say about you, usable in that answer.
Generative Engine Optimization (GEO) is the practice of improving how often and how accurately a person, an organization or a piece of content is used and cited in answers written by generative systems such as ChatGPT, Gemini, Perplexity, Claude and Google's AI features. It builds on SEO and differs from it in its object: the unit that matters is the passage a system can use and attribute, and the body of independent sources that support it.
## Where the term comes from
The term was introduced in a research paper, "GEO: Generative Engine Optimization", by Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan and Ameet Deshpande, first submitted in November 2023 and published at KDD 2024 ([arXiv:2311.09735](https://arxiv.org/abs/2311.09735)).
The authors describe generative engines as systems that answer a query by synthesizing information from multiple sources and summarizing it with a language model. They observe that content creators have "little to no control over when and how their content is displayed" in such answers, and propose GEO as a framework to improve that visibility.
This page uses the term in a wider sense than the paper: it includes the sources that describe an entity, not only the wording of a page, and says so where the two differ.
## What the research measured
The paper's figures are easy to quote beyond what they show. What it reports:
| Element | What the paper states |
|---|---|
| Benchmark | GEO-bench, 10,000 queries from diverse domains |
| Main setup | Answers generated by the gpt3.5-turbo model from retrieved sources |
| Methods tested | Nine ways of rewriting a source, among them adding citations, quotations and statistics, simplifying the language, and keyword stuffing |
| Main finding | Including citations, quotations from relevant sources and statistics can boost source visibility "with an increase of over 40% across various queries" |
| Best methods | Improve on the baseline by 41% on position-adjusted word count and 28% on subjective impression |
| Second engine | On Perplexity.ai, visibility improvements "up to 37%" |
| Caveat | The efficacy of the strategies varies across domains |
Keyword stuffing, which the paper presents as a classical SEO tactic, is not among the methods the authors report as effective.
Three limits follow from the design. Visibility is measured inside the generated answer, by a position-adjusted word count and a subjective impression score, not as traffic or revenue. The main experiments used one model, gpt3.5-turbo, in a controlled setup; the paper was submitted in November 2023 and last revised in June 2024. And the methods concern how a page is written, not whether the entity behind it is trusted elsewhere. The finding to keep is directional: in this benchmark, adding sources, quotations and figures raised a source's visibility.
## What the providers document
Google and OpenAI each publish their own rules, and they are not the same.
**Google.** Its documentation states: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." To be shown as a supporting link, a page must be indexed and eligible to be shown in Google Search with a snippet. No new machine-readable file, AI text file or special markup is required ([Google Search Central](https://developers.google.com/search/docs/appearance/ai-features)).
**OpenAI.** It operates separate crawlers with separate purposes. OAI-SearchBot is used to surface websites in ChatGPT's search features, and OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers; its FAQ for publishers adds that it may still surface the link and page title of a disallowed page if it obtains the URL from a third-party search provider ([OpenAI, publishers and developers FAQ](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq)). GPTBot crawls content that may be used to train its models; disallowing it is a statement about training, not about search ([OpenAI, crawlers](https://developers.openai.com/api/docs/bots)). A site that blocks every AI crawler without distinction therefore also blocks OAI-SearchBot, and removes itself from ChatGPT search answers.
**Engines differ in the sources they use.** The available figure, and its limits, are given in [how sources and corroboration shape representation](https://mohammedteto.com/ai-visibility/#sources). It is the reason GEO is assessed engine by engine.
## What GEO changes in practice
- **Access** Pages can be crawled, indexed and shown with a snippet, and the crawlers that feed AI search are not blocked by accident.
- **Evidence on the page** Each important claim carries its source, its date and its figure where one exists. Citations, quotations and data replace adjectives. This is the part the research supports most directly.
- **An unambiguous entity** The same names, roles and relationships everywhere, so that a passage is attributed to the right person or organization. See [entity optimization](https://mohammedteto.com/entity-optimization/).
- **Independent sources** What publications, institutions and reference sources say about the entity, because an answer about you can draw on sources other than your site.
- **Measurement** A fixed set of prompts, run per engine and repeated. See [AI search intelligence](https://mohammedteto.com/ai-search-intelligence/).
Structuring a page so that a direct answer can be extracted from it is the adjacent discipline, covered in [answer engine optimization](https://mohammedteto.com/answer-engine-optimization/).
## What GEO is not
It is not a way to make a model say something. No one can edit a generated answer; what changes over time is the state of the sources it draws on.
It is not a file or a tag. An [llms.txt](https://llmstxt.org/) file is a proposed convention for presenting content to language models. Google states that no AI text file is needed to appear in its AI features.
It is not a replacement for SEO. A page a search system cannot find gives an assistant that searches nothing to retrieve.
And it is not a guarantee. A consultant or an agency that promises a mention, a citation or a position in an AI answer is promising something outside its control.
## Where to go next
GEO is one of the five disciplines compared in [AI visibility](https://mohammedteto.com/ai-visibility/#disciplines). It is applied inside [The Authority Architecture System](https://mohammedteto.com/method/), at the layer called [AI Search Optimization](https://mohammedteto.com/method/#ai-search-optimization). The three acronyms are set side by side in [SEO, AEO and GEO compared](https://mohammedteto.com/compare/seo-vs-aeo-vs-geo/). Related analyses are collected in [Insights](https://mohammedteto.com/insights/).
## Frequently asked questions
### What does GEO stand for?
Generative Engine Optimization. The term was introduced in a 2023 research paper published at KDD 2024, and describes improving the visibility of content in answers produced by generative engines.
### Does GEO work?
The founding paper reports that adding citations, quotations and statistics to a source "can significantly boost source visibility", by up to 40% according to its abstract ([arXiv:2311.09735](https://arxiv.org/abs/2311.09735)). That is a controlled experiment on one setup. It indicates a direction and does not predict a result for a given site.
### Do I need an llms.txt file?
No. It is a proposed convention for presenting content to language models. Google states that no AI text file is required to appear in AI Overviews or AI Mode.
### Should I block AI crawlers?
That depends on the crawler. OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, while GPTBot concerns training. The two can be set separately in robots.txt.
### Can anyone guarantee that ChatGPT will cite my site?
No. The wording of a generated answer, and the sources it cites, are decided by the system at the moment of the question. What can be improved is the quality, accessibility and corroboration of the sources.
---
Understand how search and AI currently interpret your authority. Request a confidential assessment of your visibility, entity signals, narrative consistency and source authority.
[Request a Private Assessment](https://mohammedteto.com/private-assessment/)