Skip to content

How we measure

AI answers change every time. So we count, not guess.

The method behind every number we show you, and what it does not cover. Every report links here.

4
assistants asked, each with web search on
20
buyer questions per business
1
scan a week, on Friday
80
answers a week, every one kept

What happens in a scan

  1. 01

    We read your website

    We fetch your homepage, robots.txt, llms.txt and sitemap, and note the facts the page itself states, such as services, hours and prices. A fact is kept only if the page says it.

  2. 02

    We write the questions your buyers ask

    Buyer questions for your category and city, the things someone would type into an AI assistant to find a business like yours.

  3. 03

    We ask the assistants

    Each question goes to each assistant with web search on and your city stated, and every answer is saved as it was given.

  1. 04

    We record who is named

    For each answer: which businesses were named, whether you were one of them, and which source the assistant used. Every number links back to these rows.

  2. 05

    We count

    The result is a count: named in this many answers, out of this many. AI visibility is that share, always shown with its count.

  3. 06

    We check the facts and list the fixes

    The accuracy check compares what AI says about you with the facts on your website. Each wrong fact becomes a fix, and a fix is confirmed only when later scans get it right.

The free scan and a plan

Free scan

6 buyer questions, each asked once on ChatGPT and Gemini: 12 answers. About two minutes. No card and no account.

A plan

20 buyer questions on all 4 assistants, once a week: 80 answers a week. Wrong facts and fixes are tracked until later scans confirm them.

The method, part by part

4 assistants, with web search on

On a paid plan we ask ChatGPT, Gemini, Claude and Perplexity, each with web search switched on, the way their apps work when they look something up. The free scan asks ChatGPT and Gemini.

Every answer records the assistant version that gave it. When a provider changes that version, or we change a setting, a new series starts, and we only compare weeks inside one series.

The API, not the app

We ask through each assistant's API, not by typing into the ChatGPT or Gemini apps. It is the only way to ask the same question the same way every week and keep the full answer and its sources.

Your own ChatGPT may answer differently: the apps know your history, your location and your past chats. Every report says so.

Set to your city

Local questions depend on where they are asked from. ChatGPT gets your city as its search location. Gemini's search ignores coordinates, so we put the city in the question itself.

We never rely on where our servers happen to be.

20 questions, once a week

Each business gets 20 buyer questions for its field, split by intent: the best in the city, a problem, a comparison, a price. We ask them every Friday, once on each assistant: 20 answers per assistant a week, 80 in all.

The free scan is smaller on purpose: 6 questions, each asked once on ChatGPT and on Gemini, 12 answers. The report says so.

Counts, and a rule for change

We report counts: how many answers named you, out of how many. AI visibility is that share, and it is never shown without its count.

A change counts as real only when it falls outside normal week to week movement and holds for 2 weeks in a row, on the same questions. We always read it next to your competitors.

Wrong facts close the same way: an error counts as fixed only when the answer is right on 2 scans in a row.

What we do not cover yet

Google AI Overviews, Google AI Mode and Microsoft Copilot are not measured yet. Until they are, every report says so.

We do not measure or promise traffic, leads or new patients. Where we show signals like visits from AI assistants, they are labelled as indicators.

A monthly check in the real apps

Once a month we ask 10 of your questions by hand in the ChatGPT and Gemini apps and compare them with the API answers, so you can see how closely our numbers follow what people see.

Why there is no overall position

An assistant's answer is a short list that changes from one ask to the next. We show where each answer placed you, as a position in that answer.

We never reduce that to one overall position or a figure out of 100. We show how often you are named, the answers behind every number, and the sources the assistants trusted.

See your own answers first.

Run the free scan