Glossary
GEO terms, without the waffle
Every definition is short enough to quote and precise enough to be useful. Missing a term? Tell us.
- GEO (generative engine optimization)
- The field that makes brands visible in the answers of AI assistants such as ChatGPT, Gemini and Claude. Where SEO is about positions in search results, GEO is about being named in generated answers. In search engines the term sometimes clashes with geographic software; writing it out in full prevents confusion.
- AI visibility
- The extent to which a brand appears in AI answers to questions from potential customers. Measurable by repeatedly asking assistants the same questions and tallying how often the brand is named.
- Presence rate
- The percentage of answers in which a brand is named, measured over a series of repeated questions. A presence rate without a confidence interval is a claim without coverage: with few measurements the true percentage can lie far from the measured percentage.
- Confidence interval (95%)
- The range within which the true value lies with 95% certainty. Being named in 5 of 15 answers sounds like 33%, but the interval then runs from 15% to 58%: the honest answer is therefore 'somewhere in between'. More measurements make the band narrower.
- Wilson interval
- The calculation method for confidence intervals on percentages that also holds up with small numbers of measurements and extreme outcomes (0% or 100%). Hence the standard for AI visibility measurements, where samples are small.
- Visibility Score
- A composite score that, besides being named, also weighs how prominently a brand appears in the answer: at the top of a list counts more heavily than a passing mention at the bottom.
- Share of Voice (SoV)
- A brand's share of all brand mentions within a set of answers. Where the presence rate says how often you are named, Share of Voice says how much of the conversation you are compared to competitors.
- Citation (source reference)
- A reference from an AI assistant to a web page as a source for its answer. Citations determine which sites feed the answer; whoever is cited indirectly steers what is said about a market.
- Source overlap
- The extent to which different AI assistants cite the same sources. In practice low: in our measurements 316 of the 398 cited domains appeared with exactly one assistant. Visibility with one therefore says little about the other.
- AI Overviews
- The AI summaries at the top of Google search results. They often answer the question completely, so fewer people click through to websites; a major reason organic traffic falls while positions stay the same.
- Prompt (measurement question)
- The customer question put to an AI assistant to measure visibility, such as 'what is a good marketing agency in the Netherlands?'. Good measurement questions are customer language, not jargon, and contain no brand names.
- Brandless category question
- A measurement question without a brand name in it ('best accounting software') instead of with one ('is Exact better than Moneybird?'). Brandless questions measure the playing field and are reusable; brand questions measure reputation and are customer-specific.
- Measurement round (run)
- One complete measurement of all a project's questions across all chosen assistants, with repetition per question. Weekly rounds build a trend line in which real shifts can be told apart from chance.
- Uncharted territory
- Questions for which no brand from the comparison is named. There the first mention is cheapest: no one needs to be pushed aside.
- AI Readiness
- How well a website is technically readable for AI assistants: from structure and loadability to llms.txt. Necessary but not sufficient: the answer is mostly determined by external sources the assistant cites.
- llms.txt
- A text file on a website that tells AI systems where the most important, quotable content is; comparable to what robots.txt is for search engines. A small signal, but cheap to get right.
- White-label reporting
- Reports an agency delivers to clients under its own name, logo and domain, while the measurement technology comes from a platform. The end client sees the agency, not the platform.
- Grounding (web anchoring)
- An AI assistant's ability to consult live web sources while answering. Grounded answers contain citations and change faster; ungrounded answers lean on training data and lag months behind.
Within a day you'll know whether AI mentions or ignores your brand
The first measurement runs on your own prompts, across three assistants, with real figures. No demo data, no sales call first.
Prefer to talk it over first? Book a call