Analyse
Turn samples into a number you can act on.
Raw answers are noisy. Ondit reduces them to a citation share per engine, a ranked list of who is cited instead of you, the questions you are absent from, and the claims the engines get wrong — each with the rule that produced it.
What it does
Four readings from one week of samples.
Citation share, median then mean
Cited in a majority of the five replications counts as cited for that question; the share is the fraction of questions. Two steps, so one lucky answer does not move the number.
Competitors by domain and page
Every domain the engines cited across your questions, ranked, and inside each domain the exact pages — the list a content plan comes from.
Gaps and fact-check
A gap is a question you were absent from plus the competitor who was cited; a fact-check is a claim an engine made about you, checked against your own sourced facts.
How it is built
What comes out
Visibility per engine with a two-week trend; a competitor table; a gap list; a fact-check list with the correction and its source.
What happens next
Each gap can become a brief in one click — priced before you confirm.
- Significance has floorsA difference counts only past five usable samples a side and twenty percentage points. Below that the panel says "no conclusion" rather than inventing one.
- Facts are the ground truthFact-check compares an answer against the datapoints on your listing — licence, price, last release — never against the model's own knowledge, which is how confident wrong corrections happen.
- The answer is untrusted textAn engine's answer may have read a page written by someone with an interest in what we say. It is wrapped as evidence, never as instruction.
- Recomputed, never editedShares and gaps are derived from samples on demand. Change a rule and the history follows it.