GEO glossary: key terms in generative engine optimisation explained

GEO (Generative Engine Optimisation) has spawned a vocabulary that is unfamiliar even to experienced marketers. This glossary explains every key term in plain English, with notes on what each concept means in practice for your AI visibility.

Key takeaway: The GEO vocabulary overlaps significantly with SEO but has distinct additions - particularly around AI-specific crawl signals (GPTBot, llms.txt), citation mechanisms (answer engine, retrieval-augmented generation), and content structure requirements (answer-first, question-headings). Bookmark this page as your reference.

Core GEO concepts

GEO (Generative Engine Optimisation)

The practice of optimising web content to be discovered, cited, and referenced by AI engines when they generate responses to user queries. GEO encompasses content structure, entity clarity, schema markup, and citation signal optimisation.

Practical note: If SEO is about ranking in Google, GEO is about being cited by ChatGPT, Gemini, Perplexity, and Copilot. The two practices reinforce each other and should be run together, not in isolation.

AEO (Answer Engine Optimisation)

A subset of GEO focused specifically on having your content selected as the direct answer to a query - the answer that appears in position zero, the featured snippet equivalent for AI assistants. AEO prioritises concise, direct, self-contained answers.

Practical note: AEO is to GEO what featured snippets are to SEO: a specific mechanism within a broader discipline. Most GEO programmes target AEO through answer-first paragraphs and FAQ schema.

AI Visibility

A composite score that measures how effectively a domain can be discovered and cited by AI engines. Like SEO visibility but specific to AI assistant responses. Measured across signals including entity clarity, answer-first structure, schema completeness, and source authority.

Practical note: SearchScore’s AI visibility score ranges from 0 to 100 and is benchmarked against 850,000+ audited sites. Most first-time audited domains score between 5 and 20.

Citation (AI Citation)

When an AI engine references a specific source in its generated response. Unlike a traditional search result click, a citation is an attribution within the AI’s answer - the equivalent of naming a source in a footnote. Citations are the primary currency of GEO.

Practical note: A citation does not always mean the user clicks through to your site. It means your brand or content was used as a source by the AI. Brand visibility from citations can drive awareness even without direct traffic.

Retrieval-Augmented Generation (RAG)

The architecture used by most modern AI engines to answer queries. The AI retrieves relevant content from its indexed sources, then generates a response using that content as context. Your content must be retrievable before it can be used in generation.

Practical note: RAG means your content needs to be both retrievable (indexed, structured, relevant) and suitable for generation (clear, factual, self-contained answers). Thin or ambiguous content fails at the retrieval stage.

llms_full.txt

A proposed standard (not yet universally adopted) for a text file located at the root of a domain (yourdomain.com/llms_full.txt) that lists all content an organisation consents to have used for AI training and indexing. The counterpart to robots.txt for AI crawlers.

Practical note: As of early 2026, llms_full.txt adoption is growing but not yet universal. It is worth creating one listing your key content URLs as a signal of intent. Do not rely on it as a replacement for robots.txt or as a guarantee of exclusion if you want to be indexed.

GPTBot

OpenAI’s web crawler, analogous to Googlebot for Google Search. GPTBot visits websites to train AI models and to retrieve content for ChatGPT’s live responses. It respects robots.txt directives.

Practical note: Allow GPTBot in your robots.txt unless you object to your content being used for AI training. Blocking GPTBot removes you from ChatGPT’s potential citation pool. The directive is: User-agent: GPTBot Allow: /

Google-Extended (Gemini)

Google’s crawler specifically used to index content for Gemini and AI Overviews. Previously combined with the standard Googlebot, Google-Extended was separated to allow site operators to control whether their content is used for Gemini training and indexing independently of standard search indexing.

Practical note: Allow Google-Extended if you want to be eligible for AI Overviews and Gemini citations. Blocking it keeps you out of Gemini but does not affect your Google Search presence. Directive: User-agent: Google-Extended Allow: /

Answer Engine

An AI-powered search interface that answers queries directly rather than returning a list of links. Perplexity, ChatGPT Search, and Gemini are all answer engines. The shift from link lists to direct answers is the fundamental change that GEO responds to.

Practical note: When optimising for answer engines, your goal is not to appear at position 1 in a list - it is to be cited as the authoritative source for a specific factual assertion. The optimisation signals are different from traditional ranking.

Entity (and Entity Clarity)

An entity is a distinct, identifiable thing in the world - a person, organisation, product, place, or concept. Entity clarity means your content clearly identifies and describes the relevant entities it discusses, enabling AI engines to map your content to their knowledge graphs.

Practical note: Write entity-rich content: name the specific products, people, companies, figures, and dates relevant to your topic. “Our platform processes 10 million queries daily” is entity-rich. “Our platform is widely used” is not.

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness)

Google’s quality evaluator framework, originally introduced as E-A-T and expanded to include Experience. It is applied both by human quality raters and, in modified form, algorithmically. E-E-A-T signals are also used by AI engines to evaluate source credibility for citation.

Practical note: For GEO, Experience is increasingly important - first-hand, specific, named experience with the topic. “We have helped 3,200 clients reduce churn by an average of 22%” demonstrates experience. Generic claims do not.

SAVI (SearchScore AI Visibility Index)

SearchScore’s proprietary AI visibility metric, combining signals across answer-first structure, entity clarity, schema markup, citation readiness, and source authority into a single 0-100 score, benchmarked against a database of audited domains.

Practical note: SAVI is updated quarterly for tracked domains. It is the most comprehensive single metric for AI visibility available and is the benchmark used by SearchScore’s methodology across all audited sites.

Question Heading

An H2 or H3 heading phrased as a question that a user might type into an AI assistant, rather than as a topic statement. “What pricing plans are available?” is a question heading. “Pricing Plans” is not.

Practical note: Question headings are a direct citation signal for AI engines. They create an explicit match between the heading text and the queries the AI is likely to receive. Every informational H2 should be a question.

Answer-First Paragraph

A paragraph that states the answer to the section’s question in the first 40-60 words, before any context or narrative setup. The AI citation-optimised alternative to the traditional blog introduction that builds tension before delivering the key point.

Practical note: Pages with answer-first structure are cited at approximately 3x the rate of pages with narrative-first openings. The first 40-60 words after each heading should be a complete, self-contained answer.

Source Authority

The assessment by an AI engine of how credible, authoritative, and trustworthy a source is, used to determine whether to cite it and how prominently. Source authority is derived from backlink profile, domain age, entity clarity, E-E-A-T signals, and citation history.

Practical note: Source authority is why a SearchScore report is cited more readily than an unknown blog post on the same topic. Authority is built through backlinks, citations, and consistent demonstrated expertise over time - it cannot be faked in the short term.

SEO terms relevant to GEO

These are established SEO terms that are directly applicable to GEO work. Understanding them in the GEO context helps clarify why the two disciplines are complementary rather than competing.

Schema Markup (Structured Data)

Structured data in JSON-LD format placed in a page’s head, using schema.org vocabulary to explicitly describe entities, Q&A pairs, articles, products, and other content types. Schema helps both search engines and AI engines parse content accurately.

Practical note: FAQPage schema is the single highest-ROI schema type for GEO. Article schema, Organisation schema, and Product schema are also valuable. Use Google’s Rich Results Test to validate after implementation.

Rich Results

Search results that include visual enhancements beyond the standard blue link - such as FAQ panels, review stars, product pricing, and HowTo steps. Rich results require valid structured data. They are visible signals of schema usage and correlate with AI citation potential.

Practical note: FAQ rich results in Google Search are a proxy for AI citation readiness. If your pages qualify for FAQ rich results, they almost certainly contain the structural elements that AI engines look for.

Crawl Budget

The number of pages a search engine crawler will index on your site within a given timeframe, influenced by site speed, server reliability, internal linking, and site size. For AI engines with live indexing (Gemini, Bing Copilot), crawl budget affects how quickly new content is indexed.

Practical note: Improve crawl efficiency by fixing broken links, improving page speed, submitting XML sitemaps, and using canonical tags correctly. A healthy crawl budget ensures your new content is indexed before it becomes stale.

Frequently asked questions

What is the difference between GEO and AEO?

GEO is the broader discipline of optimising for AI engine visibility across all touchpoints - content structure, entities, schema, authority signals, and more. AEO (Answer Engine Optimisation) is a subset focused specifically on being selected as the direct answer to a query. AEO is the tactic; GEO is the strategy.

Is llms.txt a replacement for robots.txt?

No. llms_full.txt is a proposed addition to robots.txt, not a replacement. Robots.txt controls crawler access for search engines and AI crawlers that respect the standard. llms_full.txt is a separate mechanism for explicitly declaring which content can be used for AI training. Both should exist on your domain. robots.txt handles the technical crawl permission; llms_full.txt handles the policy declaration.

Should I block or allow GPTBot?

In most cases, allow GPTBot. Blocking it prevents your content from being used in ChatGPT responses, which removes a significant portion of your AI citation opportunity. Only block if you have a specific legal or ethical objection to your content being used for AI training or retrieval. Run a cost-benefit analysis: the traffic and brand awareness from ChatGPT citations typically outweigh any theoretical training risk.

How is AI visibility measured?

AI visibility is measured through a combination of signal-based analysis (evaluating your content against known AI citation signals) and query simulation (testing how your brand performs for representative queries across AI engines). SearchScore’s SAVI score uses 250+ signals, benchmarked against 850,000+ audited sites, to produce a single 0-100 score with per-engine breakdowns.

Do I need different content for SEO and GEO?

No. The optimal structure for both channels is the same: clear question-headings, answer-first opening paragraphs, specific named entities, FAQ schema, and authoritative depth. Creating separate content for each channel wastes resources. Optimise one piece of content for both channels simultaneously and you get the full benefit from each.