Assistants and answers
How often you appear among the sources for a set of questions — a number whose entire meaning lives in the set, which is the part suppliers omit.
Also called Share of voice in AI answers · Доля цитирования
Formula
Citation share = Answers citing you ÷ Answers sampled
- Answers citing you
- responses naming or linking your domain, counted by reading them
- Answers sampled
- the prompts asked, the assistant used, and the date — all three published, or the denominator is unknown
There is no log and no API that reports this; every figure is produced by asking questions and reading answers. So the prompt list is not methodology to be summarised, it is the denominator. A share quoted without it cannot be reproduced, contradicted, or compared with next month's.
The number is also unstable in ways a search ranking is not. The same prompt asked twice can return different sources, models are replaced without notice, and a change of a few points between months may be the assistant rather than the site. A single measurement is therefore not a baseline; several, over a fixed prompt set, are.
This is the honest state of measurement in answer engines today, and the reason to say it plainly is that the alternative is being sold a dashboard. A tool reporting your citation share to one decimal place across “thousands of prompts” is reporting a number whose prompts you have not seen, produced by sampling a system that does not promise to be stable.
Ask for the prompts, the assistant, the dates and the raw answers. If the prompt set is proprietary, the figure is unfalsifiable and cannot go in front of anyone making a decision. And check what the prompts are for: a set built from questions your buyers do not ask will report an excellent share of an audience that does not exist.
Quantities derived from the counts. Each is only as sound as the definitions beneath it.
Cannot be computed without
- Citation
- Being named or linked inside an answer a machine assembled — the unit of answer-engine visibility, and the one your own server cannot see.
Knowing the definition is not the same as being able to check the figure. These are the procedures that do the second thing.
- Testing a promise about AI answers
- “We will get you cited by ChatGPT and the other assistants.” · 20 minutes, 5 questions.
The definitions are the easy part. Whether the figure on your dashboard was computed this way is a different question, and usually the more expensive one.