How we measure AI visibility

AI answers change from one ask to the next, so a single screenshot proves very little. This is how we turn them into numbers you can plan with.

  1. 1

    You set up a project

    A project is one brand: its name, website, alternative spellings, and the competitors you want to compare with. You also set the market and the question language, Hungarian or English. Hungarian questions are asked with the context of a person living in Hungary.

  2. 2

    We collect the questions worth tracking

    From your Search Console searches, your website and your competitors' websites. An AI model scores every question for relevance from 1 to 5 and filters out branded and navigational ones. Products and partner brands you sell, like iPhone or Netflix, are kept, because you want to show up for those.

  3. 3

    Every question goes to every assistant, several times

    We use the official APIs of ChatGPT, Claude, Gemini and Perplexity with web search turned on, the way people use them for buying questions. Each question is asked two or more times per assistant, on a weekly or daily schedule.

  4. 4

    We read every answer

    Which brands are mentioned and in what order, which links are cited, what the assistant searched for, and in what tone it talks about you. Brand matching handles Hungarian suffixes, so “Telekomnál” counts as a Telekom mention.

  5. 5

    You get rates, trends and actions

    The dashboard updates after every run. Once a month the report puts it together: where you stand, what changed, and what to do next.

What the numbers mean

Mention rate
The share of answers that name your brand. 50% means every second answer recommends you.
Average position
Where you appear among the brands named, when you appear. 1 means first.
Share of voice
Your mentions as a share of all brand mentions across you and your competitors.
Own site cited
How often the answer links to your website as a source.

What it can and can't tell you

Reliable: trends over weeks, your position against competitors, the topics where you're missing, and the sources the assistants rely on. Repeated asks smooth out the randomness.

Not exact: a single answer one customer sees. Personal accounts add memory, history and location, and the apps sometimes use different model versions from the APIs. Treat one answer as an example, and the rate as the measurement.

Not a guarantee: no tool can make an AI assistant recommend you. We show what moves the numbers and whether your changes worked.

Get your free AI check See the features