AI search & GEO glossary
Plain-English definitions of the terms this site uses, written to stand on their own — the same discipline the rest of Foreground's copy tries to follow: specific facts, not vague marketing mood.
What is GEO (Generative Engine Optimization)?
GEO is the practice of making a business more likely to be mentioned when an AI system — ChatGPT, Gemini, Claude, Grok, Perplexity, and similar tools — answers a question in that business's category. It's the same underlying goal as traditional SEO (being found), aimed at a different surface: instead of ranking in a list of links, GEO is about being one of the few names an AI model actually says out loud in a generated answer.
What is AI search visibility?
AI search visibility is whether a business shows up when someone asks an AI system a question in that business's category — "best plumber near me," "who does custom furniture in Rotterdam." Unlike a traditional search results page, an AI-generated answer has no fixed number of slots and no way to scroll to page two: it's a short paragraph with room for only a few names, and a business is either one of them or it isn't.
What is a customer niche (in AI search visibility)?
A customer niche is a specific type of customer, what they want, and where — not a whole business category. "Emergency plumber" is a category; "a Rotterdam renter who needs a burst pipe fixed tonight" is a niche. AI visibility is measured per niche because the answer an AI model gives genuinely changes based on who's asking and what they need, not just what industry a business is in.
What is a shortlist (in AI search visibility)?
A shortlist is the small set of businesses an AI model actually names often enough, across a set of answers for one customer niche, to count as a real contender rather than a one-off mention. Being "in the shortlist" means an AI model names a business regularly enough for that niche; not being in it isn't a penalty, it's a starting point.
What is a citation in AI search?
A citation is the specific web page an AI model points to as the source for a claim in its answer, when that model's answer is grounded in live web search rather than only its training data. Which pages get cited is a real, actionable signal: a page that never gets cited as a source is a page these models effectively can't see, no matter how good the business is.
What is presence detection?
Presence detection is checking whether a business's name appears in an AI-generated answer at all — by name, a known alias, or its domain. A business mentioned only in passing, or only as "beaten" by a competitor's recommendation, still counts as present. It's a plain mechanical check, not a judgment of how favorably the business was described.
What is an action plan (in AI search visibility)?
An action plan is a set of specific, fact-checked actions for getting named more often in one customer niche — page copy to publish, directories to list on, a review request to send. "Fact-checked" means every claim in it is checked against the business's own site or the pages AI actually cited before it's shown; claims that don't check out are removed rather than softened.
What is an AI hallucination, in this context?
A hallucination is when an AI model states something as fact that isn't true — inventing a business, misattributing a service, or citing a source that doesn't actually say what the model claims. It's a real, known limitation of these systems and one reason a single AI-generated answer is treated as a directional signal worth tracking over time, not a certified, always-correct measurement.
What is an AI answer engine?
An AI answer engine is any AI system people actually ask questions to expecting a direct answer instead of a list of links — ChatGPT, Gemini, Claude, Grok, Perplexity, and Microsoft Copilot are the best-known examples. The distinction from a traditional search engine matters: an answer engine returns one synthesized response instead of ranked results, so there's no "page one" to check by hand.
What is web-search grounding?
Web-search grounding is when an AI model's answer is generated using live search results fetched at the moment of the question, instead of relying only on whatever it learned during training. A grounded answer can cite real, current pages as sources; an ungrounded answer can only draw on facts baked into the model's training data, which may be stale or simply never included a given business at all.
See how Foreground's own scan works for how these ideas turn into "named in X of 12 answers" for a real customer niche, or check your own business for free.