GEO glossary: AI search terms explained in plain English

Terms, A to Z

AI Overviews

Google's AI-written summaries that can appear above the usual search results for some searches. They draw on pages Google has indexed and usually show links to some of the sources used. Whether one appears, and what it says, varies by search.

Answer engine

A tool that replies to a question with a written answer rather than a list of links. ChatGPT, Perplexity, Gemini and Copilot are often described this way. The term is informal.

Citation

A link or named source an assistant shows alongside its answer, pointing to a page it relied on. Being cited is not the same as being recommended: a page can be cited for a fact while a competitor is the business named. Not every assistant shows citations every time.

Entity

A distinct thing a system can recognise: a business, person, place or product. Search engines and assistants try to connect the facts they find to the right entity. Inconsistent names or details can make that harder.

Generative engine optimisation (GEO)

The work of making a business easy for AI assistants to find, describe correctly and name in their answers. Related labels include AEO and LLMO. It improves the odds of being named; it cannot guarantee it.

Grounding

Tying an answer to specific sources, often by searching the web or a document set at the time of asking. Grounded answers tend to be more current and are more likely to show citations, though they can still be wrong.

Hallucination

When an assistant states something false as if it were fact, such as an office that does not exist or an old price. It happens because answers are generated from patterns, not looked up line by line. Clear, consistent facts on your own pages reduce the room for it but do not remove it.

llms.txt

A proposed plain-text file at the root of a website that gives AI tools a short guide to its most useful pages. It is a proposal, not an agreed standard, and support varies. Cheap to add; do not expect results from it alone.

Prompt or question set

A fixed list of questions, written the way customers ask them, used to test assistants repeatedly. Keeping the set the same over time is what makes month-to-month comparison meaningful.

robots.txt

A text file at the root of a site that tells crawlers which parts they may fetch. It is a request that well-behaved bots follow, not a lock. You can allow or block search and AI crawlers separately.

Schema markup

Code added to a page, usually as JSON-LD, that labels facts in a shared vocabulary from schema.org: this is a business, this is its address, these are its opening hours. It helps machines read facts correctly. On its own it does not get a business named.

Share of answers

An informal way some people describe how often a business is named across a set of AI answers, for example named in 3 of 10 asks. There is no agreed standard for measuring it, so always say which questions, which assistant, how many asks and when.

Training data vs live search

Training data is the text a model learned from, frozen at a cut-off date. Live search is when an assistant looks things up at the time of asking. Updating your pages can affect live search answers once they are recrawled; it does little to what a model already learned.

Zero-click search

A search where the person gets what they need on the results page, from a snippet, panel or AI summary, and does not click through to any site. It means fewer visits can happen even when your information is being used.

Honest limits

Definitions here reflect common usage at the time of writing; vendors change their products and terms. GEO Agency's tools run on Google Gemini, so scores and simulations are estimates, not a record of any assistant's answers. We do not promise rankings or a place in AI answers.