Your questions, exactly as you wrote them
Each tracked question is used word for word, with no rewording on our side.
What we collect, how each indicator is calculated, what our measurements don’t tell you, and what we do with your data.
Method version 1.4 · Oct 10, 2026
Each tracked question is used word for word, with no rewording on our side.
You see the answers exactly as they were given, with their date and the market concerned: what your customers see, not what a user somewhere else would see.
Every answer is kept with its date: the full answer, the entities detected, the cited sources and the searches the engine ran.
| Indicator | Question it answers | Formula | Denominator |
|---|---|---|---|
| Visibility | Am I cited, and how often? | answers with a “My brand” entity ÷ answers collected | Answers collected over the period and engines shown (e.g. 214 answers). “My brand”: your brand and what you attach to it — products, people, events. A “cited” entity is an entity mentioned in the answer. |
| Share of voice | How much weight do I carry against the other players cited? | your mentions ÷ mentions of the same type | Mentions of every entity of the same type detected on your tracked prompts — brands on one side, products on the other — one per answer. |
| Position | Where does my name appear in the answer? | average of mention ranks | Only the answers where your brand is mentioned. |
| Citation share | Which sources do AI engines rely on? | domain citations ÷ total citations | Citations of every domain over the period (e.g. 212 citations). |
| Sentiment | What tone is used about me? | average tone of the passages | Passages about your brand, not whole answers. |
| Optimization score | Can AI engines read my page? | average of the criteria, weighted by importance | Criteria that apply to the analyzed page, out of the 24 in the framework. Importance high 3, medium 2, low 1; criterion good 1, average 0.5, poor 0. |
They describe a sample of questions, not every conversation your customers have. They are for comparing periods, and competitors, with each other.
Interfaces, models and sources change. Any change of method on our side is dated in the history below.
Data hosted and processed in Europe.
Collected answers and indicators are kept for as long as you subscribe, then for 12 months after your subscription ends.
Your prompts and results serve only your own tracking and are never used to train models. Access, export and deletion on request.
| Date | Version | Change |
|---|---|---|
| Oct 10, 2026 | 1.4 | Definitions clarified: visibility counts “My brand” entities, share of voice compares entities of the same type, optimization score weighted by criterion importance. |
| Sep 16, 2026 | 1.3 | Google AI Mode added to the tracked engines, with separate Overviews / Mode indicators. |
| Jul 2, 2026 | 1.2 | Average position calculated only on the answers where the brand is mentioned. |
| Apr 14, 2026 | 1.1 | Sentiment calculated on the passages about the brand, not on the whole answer. |
| Feb 3, 2026 | 1.0 | First public version of the measurement method. |
You can decline: the whole site stays available.
Remembers your theme choice and your cookie choice.
Pages viewed, referrer, device; aggregated statistics.