Loading...
Loading...
10Module
Metrics, KPIs, and methods for tracking your brand's presence in AI responses. Understanding mention rates, sentiment, and competitive positioning.
Available3 lessons18 min
Measuring AI visibility means measuring your noise floor first: how much the same question disagrees with its own rerun, computed per platform, excluding pairs where neither run cited anything. Every later result is a comparison against that figure rather than against zero. The record is a spreadsheet with one row per run, carrying the date, market, question, platform, run number, whether you were named, the full list of sites that were, and who ran it, with a fresh session per run and no edits to a question's wording. Before attributing any change to your own work, check whether the other sites named moved with you: yours moves you, and a model release, an index refresh or a question changing meaning moves everyone.
Module nine was a diagnosis run once. This is the measurement kept, which is a different discipline: what has to stay identical between runs, what a change is allowed to mean, and how to tell your own effect from the platform's.
It starts with the control, because a visibility number compared against nothing has no scale attached to it, and the control is the thing most likely to be quietly wrong. The study on this site had it wrong by 14 points in its first version, which is larger than most differences anyone reports in this field.