The GEO glossary
The terms of AI search optimization, each defined in a few lines. Dates and figures link to their primary source.
- GEO (Generative Engine Optimization)
GEO covers the techniques that help a website get cited in the answers of generative AI engines: ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews. The term was formalized by researchers from Princeton and Georgia Tech in a paper presented at KDD 2024. SEO aims for a well-ranked link; GEO aims for a citation inside the answer.
- AEO (Answer Engine Optimization)
AEO means optimizing content for answer engines, the tools that answer a question directly instead of showing a list of links: Google featured snippets, voice assistants, then conversational AI. The term predates GEO and the two largely overlap: short, structured and verifiable answers that a machine can lift from the page without rewriting them.
- LLMO (Large Language Model Optimization)
LLMO is another name for optimizing content for large language models. It covers the same practices as GEO: making content accessible to AI crawlers, easy to extract and backed by sources. Some people also use it to describe how present a brand is in the data that models are trained on, before any web search happens.
- LLM (large language model)
An LLM is an artificial intelligence model trained on very large amounts of text to understand and generate language. GPT (OpenAI), Gemini (Google), Claude (Anthropic) and Llama (Meta) are LLMs. On its own, an LLM answers from what it has learned; connected to a search engine, it can read and cite recent pages.
- RAG (Retrieval-Augmented Generation)
RAG is the method by which an AI retrieves documents, places them in its context, then writes its answer from them. Described by Lewis et al. (Facebook AI Research, NeurIPS 2020), it explains why a clear, self-contained and sourced passage is more likely to be picked up and cited than a vague or promotional one.
Source: Lewis et al., NeurIPS 2020How ChatGPT chooses its sources
- AI Overviews
AI Overviews are the summaries written by Google's AI that appear at the top of some results pages, with links to the pages used as sources. They have been available in France since July 22, 2026, together with AI Mode. Being cited as a source in an AI Overview has become an objective in its own right for search optimization.
- Google AI Mode
AI Mode is the conversational interface of Google Search: the user asks a question, Google answers with a text written by its AI, suggests follow-up questions and shows links to its sources. Unlike AI Overviews, which sit on top of the results page, AI Mode replaces the page with a conversation. It has been available in France since July 2026.
Source: Abondance, 22 juillet 2026
- AI share of voice
AI share of voice measures a brand's place in AI answers compared with its competitors: across a set of questions from its industry, the share of mentions it gets out of all the mentions of every brand cited. It is the AI equivalent of advertising share of voice or of SEO visibility, and it only makes sense against named competitors.
- Mention rate
Mention rate is the share of an AI engine's answers in which a brand is cited, over a set of queries: a brand cited in 30 answers out of 100 has a 30% mention rate. It is the basic metric of an AI visibility audit, to be read with the position of the mention, its sentiment and the competitors cited.
- Citability
Citability is the ability of a piece of content to be reused as is by an AI engine. A citable passage answers a specific question, makes sense without the rest of the page, contains verifiable facts and names its sources. It is the first criterion of the Detekia score, which checks in particular for direct answers placed right under headings.
- Answer capsule
An answer capsule is a short paragraph placed right under a heading that answers the question asked by the heading in a few sentences. This format makes the passage easy for an AI engine to extract and reuse without rephrasing it. The Detekia score looks for capsules of 20 to 50 words as part of its citability criterion.
Source: Search Engine Land
- llms.txt
The llms.txt file, placed at the root of a website, summarizes in Markdown the pages most useful to language models. Proposed in September 2024 by Jeremy Howard, it is not a standard adopted by the major engines, and a SE Ranking study (2025) found no measurable effect on citations by ChatGPT. It remains cheap to publish.
Source: llmstxt.org (Jeremy Howard, 2024) · SE Ranking, 2025llms.txt, robots.txt and AI access
- AI crawlers (GPTBot, OAI-SearchBot…)
AI crawlers browse the web on behalf of AI companies. At OpenAI, GPTBot collects training data, OAI-SearchBot indexes pages for ChatGPT search and ChatGPT-User reads a page when a user asks for it. Blocking OAI-SearchBot in robots.txt keeps your site out of ChatGPT's sourced answers, while blocking GPTBot does not.
Source: OpenAI, Overview of OpenAI CrawlersSitemap, robots.txt and AI crawlers
- E-E-A-T
E-E-A-T stands for Experience, Expertise, Authoritativeness and Trustworthiness. This framework, taken from Google's guidelines for assessing page quality, is also a useful benchmark for AI engines: an identified author, cited sources, trust pages and an external reputation all make content easier for an engine to recommend and to cite with confidence in an answer.
- Structured data (Schema.org)
Structured data describes a page's content in a standard vocabulary, Schema.org, most often in JSON-LD format: organization, article, author, product, frequently asked questions. It helps engines identify unambiguously who is speaking, about what and since when. It does not guarantee a citation, but it removes doubts about identity and freshness.
Source: Schema.orgSchema.org for AI
- Zero-click search
A zero-click search ends without a visit to any website: the answer is displayed directly on the results page or in the conversation with the AI. With AI Overviews and conversational assistants, this behavior now extends to more and more complex questions. Hence the value of being cited in the answer itself, not only ranked among the links.
- Hallucination
A hallucination is a false statement that an AI engine produces with confidence: an invented price, a service that does not exist, a misattributed quote, a wrong date. For a brand, it is a reputation risk that is hard to spot without measurement. Clear information, consistent across pages and repeated by third-party sources, leaves less room for invention.
- GEO score
The Detekia GEO score is a score out of 100 that evaluates whether a page is ready to be cited by AI engines. It adds up 7 criteria measured deterministically: citability, verifiability, authority, AI crawler access, neutrality, external presence and freshness. Below 40 it is low; from 40 to 69, fair; from 70 up, good.
- Citation test
A citation test asks an AI engine questions your customers might ask and checks whether your site appears in the answers. The Detekia test asks one of OpenAI's models 5 questions about your page's topic, without web search: it gives a quick indication, not a measure of your visibility across your whole market.
- AI visibility audit
An AI visibility audit measures what ChatGPT, Gemini, Perplexity and Claude answer on an industry's queries: a brand's mention rate, position, sentiment, competitors and cited sources. It covers a large number of questions asked to several AI engines, whereas a manual test only gives a one-off snapshot that is hard to compare over time.
Written by Guillaume Bourdon, founder of Beeleven. Updated on September 30, 2026.
Is your page ready to be cited?
Free GEO score: 7 criteria and a citation test, in under a minute. You get the result in exchange for your email.