Methodology v1
How we measure your UP Score
AI visibility metrics aren't standardized, so we explain exactly what our score measures, and what it doesn't.
- 01We read your website
- 02You pick the questions
- 03We ask AI, like a customer
- 04We read every answer
- 05You get your UP Score
We build your business profile
We read your website (homepage and a few key pages) and pick out your name, other names customers use, what you offer, where you are and who asks AI about you. You check and correct this profile before anything runs.
We write the questions, you pick them
We write realistic questions in your customers' words, grouped in topics (what customers come to you for) and written for each market: a country and language your customers ask from. They never contain your name: they are the questions people ask before they know you. Every question names the place you serve (each city on its own when you serve several), unless you work online. You choose which ones to ask, and the same ones are asked again in your next checks, so results stay comparable. Each gets an importance weight based on how common and decision-close it is.
- “Best boutique hotel in the old town for a cultural experience?”
- “Quiet hotel near the main sights in the city center?”
We ask AI, like a customer would
Each question is sent exactly as written, with web search on and no hint about your business, and asked from its market: the city it is about, the market's city, or its country. We use the platforms' official APIs, so answers can differ from what the consumer apps show. The free check asks ChatGPT; paid plans ask every assistant below.
- Claude · web search
- OpenAI GPT · web search
- More coming
We read every answer
Mention
Is your business named? We match your name, other names and website, then double-check unclear cases. Your own branches and sub-brands count as you.
Position
Where you appear when the answer is a list.
Citations
Which websites the answer cites, and whether yours is one of them.
Competitors
Other businesses of the same kind recommended in the answer. A business only counts if its name really appears in the answer.
Answers that recommend no business at all are shown in your report but left out of the score, because they say nothing about who AI recommends.
The UP Score
Each answer is weighted by its question's importance. AI share of voice is your share of all business recommendations in the answers. The score bands:
- Stuck0–19
- Low20–39
- Moving40–59
- Rising60–79
- Up80–100
Several markets: each market gets its own score from its own answers, and your UP Score is their average, so a market with more questions doesn't outweigh the others. Scores by topic, by city and for any filter are worked out the same way from the answers they cover. “Across your markets” points out what differs between them (who AI names first, which sites it reads, your website's languages) by fixed rules, without AI.
ConfidenceLow · under 20 answersMedium · 20–59High · 60 or more
Website readiness
A separate score: can AI reach, read and trust your website? We check access for AI crawlers, HTTPS, speed, your title and description, structured data, contact details and your presence on the sites AI cites. It doesn't change your UP Score, and passing it doesn't guarantee AI visibility.
Limitations
- AI answers change between runs and over time. A single check is a snapshot.
- Results depend on the questions and platforms tested.
- We test official APIs; consumer apps (ChatGPT app, Google) can answer differently.
- Automatic detection can make mistakes. Every report shows the answers so you can verify.
- The UP Score is our own measurement, not an industry standard, and not a prediction of traffic.
When the method changes, we publish a new version. Every report records the version it used.
See it on your business.
Check my business