Method
How we measure
What ChatGPT, Gemini, Perplexity and Claude say about your business changes from week to week. Here's how we measure it, and how to read the numbers.
How a measurement works
-
01
Your questions
You choose the prompts: questions your customers actually ask, like “Who is a good plumber in The Hague?”. We ask them in the language and country you choose.
-
02
Four models, every week
Every week on a fixed day (twice a week with Agency) we ask each prompt to ChatGPT, Gemini, Perplexity and Claude, the four best-known AI assistants built on large language models (LLMs). We do this through their official APIs with web search on, just like in the apps.
ChatGPTGeminiPerplexityClaude -
03
Every answer analysed
We check whether your business name or website is mentioned, in what position, which competitors appear and which sources the model cites.
-
04
Scores and trends
From that we calculate your visibility, average position and share compared with competitors, per model and per week.
One measurement at a glance
With Starter we ask 25 questions to 4 AI models: 100 answers per measurement, every week. In your dashboard you see per question and per model whether you are mentioned, in which position, and whether it got better or worse than the previous measurement.
Prompts × models — latest measurement
▲ better / ▼ worse than last week| Prompt | ChatGPT | Gemini | Perplexity | Claude |
|---|---|---|---|---|
| What is the best plumber in The Hague? | #2 of 3 ▼ | not mentioned | #1 of 4 | #1 of 3 ▲ |
| I urgently need a plumber in The Hague. Who do you recommend? | #1 of 4 ▲ | not mentioned | #1 of 3 ▲ | not mentioned ▼ |
| Reliable plumber in Scheveningen? | #1 of 4 | not mentioned | not mentioned ▼ | #1 of 4 ▲ |
| Which plumber in The Hague has the best reviews? | not mentioned ▼ | #1 of 4 | #1 of 4 | not mentioned |
| Boiler repair in The Hague: which company? | #2 of 3 ▼ | #1 of 3 ▲ | #1 of 3 | #1 of 3 ▲ |
Example. Your dashboard shows all questions, here you see 5 (20 more).
What the numbers mean
Visibility
The percentage of answers that mention you.
Average position
Where you appear when mentioned (1 = first).
You 31%
Share of voice
Your mentions compared with your competitors' combined.
Cited sources
The websites AI gets its information from. That's often where the biggest gains are.
Reading cited pages
After every measurement we look at the pages AI assistants cite most often, such as review sites, comparison articles and news stories. For each page we check whether your business and your competitors are mentioned on it. If a competitor is on a page and you are not, you see it as an opportunity in your dashboard and we add a task to your action plan.
To do this we fetch at most 15 public pages per measurement, once, with a recognisable bot (HelloGeoLabsBot). We only read what anyone can see without logging in, we do not store a copy of the page and we only keep the title and which brand names appear on it. We follow robots.txt: if a website asks bots not to read a page, we skip it.
Opportunity: competitor yes, you no
Honest about the limits
Answers vary
Ask the same question twice and the answer may differ. Look at the trend over several weeks, not a single measurement.
The app may differ
In the apps, your location, earlier chats and settings play a role. We measure neutrally, without personal context, so every week is comparable.
Models change
AI providers release new versions regularly. A sudden rise or drop can be caused by that.
No guarantee
Nobody can promise that AI will mention your business. We can show where you stand and what helps.
Why measuring weekly is enough
Some tools measure every day. We deliberately measure every week (Agency twice a week). Here is why.
AI does not change by the hour
Models and their search results are updated over days and weeks. A weekly measurement shows the real trend, without noise from random day-to-day differences.
Improvements take time
A new page, review or mention is picked up after days to weeks. Measuring every day mostly shows that nothing has changed yet.
Many questions, four models
Every question goes to ChatGPT, Gemini, Perplexity and Claude. With 25 to 250 questions per measurement you get a reliable picture, even though each answer varies a little.
Calm, focus and a low price
One moment a week with results and two concrete tasks works better than new numbers every day. And because every measurement has AI costs, this keeps the price low, from €39 per month.
Want to measure more often? Agency measures twice a week, and with a custom plan you choose the days yourself.
What we never do
We never create fake reviews, fake mentions or other tricks to mislead models. Our advice is about real improvements: clear information on your website, mentions on trusted sites and good reviews.