10 · Documentation
Glossary
The terms that appear in augenmerk, in two sentences each.
AI assistant — The assistant in every project view. Answers questions about the project's data only with evidence and points to the view the answer comes from. 30 units per request. See The application.
Answer box — Google's AI answer above the search results (AI Overviews). A channel type of its own, because it is fed from Google's search, not from the sources of the chat systems.
Channel — A queried system in a channel type, such as "Perplexity · live search". Ten channels in three channel types. See Channels and systems.
Channel contrast — The difference between model knowledge and live search for the same brand. The diagnosis a single score cannot deliver.
Cited — A source the model used in the text: linked from a sentence or with a footnote. The stronger form compared with "named".
Competitor perception — The strengths/weaknesses questions about the three strongest competitors in the question set. Without them the topics view would lack the comparison with the market. See Questions.
Compute units — The unit of account for measurements and assistant requests. Every measurement books units before the run; every booking is in the usage ledger. See The application.
Counting question — A question with no brand in the wording and no segment. Only these form presence.
Domain check — The check of your own website with every run: reachability, AI bots, llms.txt, home page, Lighthouse, Google standing.
Evidence — The literal passage in the original answer on which a mention or statement rests. Without a findable passage there is no verdict.
Evidence rule — The rule that every mention and every statement needs a passage verified character by character. See Evidence and the judge.
File — The detail view of a finding under "Next steps": the finding, why it matters, what to do, how to see the effect. See Findings catalogue.
Finding — A rule over the measurements that applies. Appears under "Next steps" with reason, action and effect number.
Free test — One project, ten questions, four channels, one run after approval; a free second run after two weeks. Plus the frozen-pizza example project for viewing. See The application.
Gap view — Pages found on Google set against pages an AI cites. Pages with demand and without a citation are the task list; with Search Console, from real impressions. See Sources and website.
Judge — The second model that extracts mentions and statements with passages from every answer. Does not form opinions.
Lens — Viewing a project from a competitor's perspective. Shows the same evaluations, but no actions.
Live search — An answer with a current web search. The channel in which sources arise.
Measurement — One question on one channel in one run. A question produces as many measurements per run as there are active channels.
Model knowledge — An answer without web search, from the model's training alone. Never cites.
Named — A source the search presented to the model. Equivalent to "retrieved" in an answer's source list.
Need question — A question about a concrete task ("What does diesel cost in Hamburg today?"). Measures whether the AI takes on the task, not presence.
Noise band — The uncertainty of presence between two runs, depending on the number of counting questions. Movements below it do not count as movement. See Noise band and runs.
Perception question — The fixed question about your own brand in the blueprint ("What are the strengths and weaknesses of X?"); "What is X known for?" or "Why not X?" are perception questions too. Delivers the brand image, does not count towards presence.
Position (average place) — A brand's mean list position in ranking questions. Empty when no ranking question names the brand.
Presence — The share of answers in which a brand appears, across the counting questions. See Presence and position.
Prominence — How prominently a brand appears in an answer: first mention, list position, in passing.
Purchase check — The three questions buyers use to cross-check a brand before buying: drawbacks, "not for whom", worth the price. Name the brand, do not count towards presence, feed the brand image. See Questions.
Range presence — The presence across the segment questions, separate from the brand value.
Ranking question — A question about "the best", in the plural. The only question type from which a position arises.
Resolved — A finding whose number has moved since the last run. Stays visible for one run and then disappears.
Role — Owner, admin or member of an account. Members read, admins change projects, owners everything. See The application.
Run — One complete measurement of a project: all active questions on all channels, plus the domain check.
Search Console — Google's report about your own website: impressions, clicks, position per page and query. Once connected, augenmerk shows these numbers in the website view and derives questions from them. See The application.
Search query — The search a channel ran before answering, as far as it discloses it. Shows what the AI searches for when a person asks something.
Segment — A product line or niche to which questions are assigned. Segment questions count only in the segment.
Share of voice — A brand's share of all brand mentions, at most once per brand per answer.
Statement — A passage from an answer with topic and verdict, praise or criticism, verified character by character. The basis of the topics view and the brand image.
Tracked brand — A brand entered in the project. Only tracked brands get a place and appear in the topic matrix. Brands cost nothing, only questions.
Version — The state of the methodology with which a run was judged. Applies from its filing date and is never overwritten.
Updated: 2026-09-16