How kennt.ai measures AI visibility

How we collect answers

Every project has a set of prompts: the questions your buyers ask, like “best tutoring provider in Vienna”. kennt.ai suggests them from your website and, once connected, your Google Search Console queries; you confirm each one.

Every day, a data provider (DataForSEO) asks each prompt in the consumer interfaces of each engine, from your market's location and in its language. We store what a person would see: the full answer, the brands it names in order, and the sources it cites or retrieved. This is the same approach rank trackers use for Google results. We don't log in as anyone and send no personal data.

An AI model (Anthropic's Claude) then reads every answer and lists each brand named, in order, with the tone used about it. Brands you don't track yet show up as suggestions, so new competitors don't go unnoticed.

Engines

ChatGPT

The answer in the ChatGPT app, including the sources it cites and the web searches it ran.

Gemini

The answer in the Gemini app, with its sources.

Google AI Mode

The answer in Google's AI Mode tab, with its sources.

Google AI Overviews

The AI summary above Google's results. When Google shows none for a question, that run has no AI answer and doesn't count.

Every paid plan tracks all four. The free kennt.ai Score asks the first three live; AI Overviews can't be fetched live reliably, so they are in the plans only. Perplexity and Microsoft Copilot are not tracked yet.

Markets

A market is a country and language pair, for example Austria in German. AI answers differ by location and language, so prompts are written in the market's language and each market is measured separately.

Daily collection, weekly numbers

The same question can get a different answer an hour later. One day of answers is therefore noisy, and kennt.ai reports headline numbers per week, per market and engine.

A trend needs at least two full weeks. kennt.ai calls a change real only when the 95% confidence intervals of the two weeks don't overlap; otherwise it says too early to tell. We never report day-to-day moves or single position changes as a signal.

Metrics

Visibility

Share of answers that name your brand. The headline metric.

Share of voice

Your brand's mentions as a share of all mentions of tracked brands.

Position

Average rank among the brands named, when you are named (1 = first).

Sentiment

How positively the answers talk about you, 0–100.

Used as source

Share of answers that cite or retrieve one of your own pages.

Prompts that contain your own brand name (“Is GoStudent any good?”) always name you, so they are left out of visibility and used for sentiment only. Weekly probe questions about how AI sees your brand never count in the tracking numbers either.

The kennt.ai Score

The free score is one number from 0 to 100: how often AI names you, compared with your strongest competitor. A perfect 100 also needs you named in at least half of all answers.

score = 100 × your visibility ÷ max(your visibility, best competitor's visibility, 50%)

It is built from 10 unbranded questions asked live in three engines (up to 30 answers). Failed calls and calls without an AI answer are left out. More mentions of you never lower it; more mentions of a competitor never raise it.

Sample size, stated honestly

Every number in kennt.ai comes with its n (how many answers it is based on) and a 95% confidence interval (Wilson). As a rule of thumb, for a brand named in about 40% of answers:

That is why the score shows a likely range, not just one number, and why a 44% vs. 41% lead after one week is “too close to call”.

Limits

• kennt.ai sees answers without a user's personal history. People with long chat histories or connected accounts can get different answers.
• Answers are collected once a day per prompt and engine, not continuously.
• Traffic and leads come from Google Analytics 4. AI Overviews and AI Mode visits arrive as Google search, so GA4 can't attribute them to AI.