GEO / AI search

GEO (Generative Engine Optimisation)

Q: Which Schema.org types matter most for GEO?

Four types dominate: (1) Article or BlogPosting with author, datePublished and sameAs pointing to Wikidata. (2) FAQPage for Q&A sections — LLMs frequently use question-and-answer pairs directly as an answer structure. (3) HowTo for instructional and process-like content. (4) DefinedTerm for glossary entries. Supplemented by sameAs links from all relevant entities (people, organisations, concepts) to Wikidata QIDs, this creates a dense entity network that LLMs treat as a trust signal.

Also known as: Generative Engine Optimisation, GEO, AI SEO, AI search engine optimisation

Generative Engine Optimisation (GEO) is the systematic optimisation of web content for citability in the answers of generative AI systems — that is, ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Unlike classic SEO, which aims at clicks from a SERP listing, GEO optimises for being named as a source within a single summary generated by the model. The term was coined at the end of 2023 by researchers of the Princeton group (Aggarwal et al.) and established itself between 2024 and 2026 as a discipline in its own right alongside SEO.

Why GEO exists alongside SEO

Between 2024 and 2026 the share of search queries answered by AI reached 20-35 %, depending on the estimate (Adobe Digital Insights, Gartner). These answers no longer require a click — users receive the information directly. Anyone who does not appear among the AI sources loses visibility without any classic SERP loss. SEO delivers rankings, GEO delivers mention within the answer. The two complement each other; they do not replace each other.

The most important GEO levers

Citability structure — compact answers in the opening sentences, a clear H2 hierarchy, facts presented as lists.
Entity completeness — naming all entities relevant to the topic, ideally linked to a Wikidata QID via schema.org/sameAs.
Source references — linked external sources (studies, statistics, primary sources) signal trustworthiness to the AI.
Schema.org markup — above all Article, FAQPage, HowTo and DefinedTerm with an author byline and sources.
AI crawler access — OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended must be permitted in robots.txt, otherwise there is no crawl and no citation.
Freshness — AI models favour current sources, and regular content updates increase the chances of a citation.

GEO is not a black-box trick

GEO does not work through ”LLM hallucination tricks” or hidden prompts within the page — such attempts are recognised and filtered out by the major providers. At its core it is editorialisation: writing more clearly, substantiating statements, making structure readable for machines. Precisely the same best practices that featured snippets and voice search optimisation already called for — now with considerably more leverage on real traffic.

Example from practice

Example: A B2B SaaS provider adds to its product pages (a) a TL;DR section with a two-sentence answer to the most common search question, (b) Wikidata QID sameAs entries for the tool itself and three compared competitors, (c) an FAQ section with schema markup. After 8 weeks the citation rate in ChatGPT Search rises from 4 % to 19 % with stable Google rankings — the share of direct AI traffic doubles.

Frequently asked questions

What is the difference between SEO and GEO?

SEO (search engine optimisation) optimises for clicks from a SERP listing — position 1 is the goal. GEO (generative engine optimisation) optimises for being named as a source within a single answer generated by an LLM — regardless of where the page ranks in classic terms. The two disciplines complement each other: SEO brings click traffic from classic search, GEO brings visibility in AI answers. A page can sit at SERP position 15 and still be cited in Perplexity — or, conversely, hold SERP position 1 and go unmentioned in ChatGPT.

Does GEO work through hidden prompts or LLM tricks?

No. Attempts to manipulate AI models through hidden prompts, white text on a white background or other black-hat techniques are recognised and filtered out by OpenAI, Google, Anthropic and Perplexity — often with a penalty effect. At its core GEO is editorialisation: writing more clearly, substantiating statements with sources, maintaining Schema.org markup, permitting AI crawler access in robots.txt. These are the same best practices Google has been encouraging since featured snippets — only with considerably more leverage on real reach.

Which Schema.org types matter most for GEO?

Four types dominate: (1) Article or BlogPosting with author, datePublished and sameAs pointing to Wikidata. (2) FAQPage for Q&A sections — LLMs frequently use question-and-answer pairs directly as an answer structure. (3) HowTo for instructional and process-like content. (4) DefinedTerm for glossary entries. Supplemented by sameAs links from all relevant entities (people, organisations, concepts) to Wikidata QIDs, this creates a dense entity network that LLMs treat as a trust signal.

How do I measure whether GEO optimisations are working?

Through AI citation tracking: weekly monitoring of whether and how often your own domain is cited as a source by ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews for defined target questions. The key KPIs are citation rate (cited answers / total queries), mention rate (mention without a link) and citation share compared with competitors. Sound trackers work with a majority vote (querying each question three times) because of the non-deterministic nature of LLM answers. Rankmio integrates such tracking across all 5 engines in the GEO module.

Used in Rankmio for

GEO audit + citation tracking in Rankmio

Go to the feature →

Last updated: 2026-06-29 · Browse all glossary entries