Sign in

How it works

From a question to a number

There is no magic and no privileged access. We ask the models the way a customer asks them, we record every answer, and we count. This page shows exactly how.

Start the analysis
  1. 1Question
  2. 4Models
  3. 3Repetitions
  4. 12Answers
  5. 1Result

The flow

What happens to a single question

One question multiplies into answers, and the answers reduce to one number.

  1. 1QuestionWritten the way a customer would write it.
  2. 4ModelsChatGPT, Gemini, Claude, Perplexity.
  3. 3RepetitionsBecause the same model does not answer the same way twice.
  4. 12AnswersStored as raw text.
  5. 1ResultOne result, calculated through four indicators.

For a full audit with 60 questions, this becomes 720 answers.

Stages

Four stages, every one of them checkable by you

At each stage you can check the work, because the material travels with you.

  1. 01

    Building the questions

    We start from the decision the customer is making. The questions cover direct recommendation, price, urgency, named comparison and the early stage.

    • A conversation about the sector, the city and the services
    • A draft list of questions
    • Your approval before the assessment begins
  2. 02

    Running the assessment

    Every question goes to the four models and is repeated within the same assessment; how many times depends on the service. We store the full text, the date and the model for every answer.

    • The same questions, with no hidden variations
    • Answers are stored before we read them
    • No cherry picking of answers
  3. 03

    Coding and counting

    We read every answer and record what it contains: who is mentioned, in what order, what is said, what is cited.

    • Mentions and position
    • Named competitors
    • Incorrect statements about your business
  4. 04

    The reading

    The score is calculated from the four components. Then we explain what it means and where it is worth acting.

    • The score and its breakdown
    • The variance across repetitions
    • Recommendations by priority

The data

What we record for every answer

This is why you can check our conclusions yourself.

FieldTypeExample
QuestiontextWhich are the best dental clinics in Tirana?
ModellistPerplexity
Repetitionnumber2 of 3
Datedate2026-03-14
Answerfull textStored without editing
Businesses mentionedlist[Clinic A], [Clinic B]
Your positionnumber or emptynot mentioned
Sources citedURL list3 sources, your site not among them

The names in brackets are placeholders.

Variance

Why one answer is not enough

The same model, the same question, the same day. The answers come back different, which is why a single assessment is not enough.

  1. Repetition 1

    Mentions [Clinic A] and [Clinic B]

    2 businesses
  2. Repetition 2

    Mentions [Clinic A], [Clinic C] and [Clinic D]

    3 businesses
  3. Repetition 3

    Declines to recommend and suggests searching yourself

    0 businesses

Across three repetitions, your business was mentioned zero times. That is the finding. From a single repetition, we would not know whether it was chance.

An illustrative example of the kind of variance we encounter.

Next step

See how it applies to your own sector.

Tell us where you operate and we will come back with the proposed questions.

Start the analysisRead the methodology