How Bella measures
The questions
Thirty questions per business category, written the way a customer types or says them: "AC stopped working, who can come out today?", "best plumber in Round Rock", "how much does a furnace repair cost in Austin". They cover emergency repair, who's recommended, installation, maintenance, specific problems, cost and reviews. The set is fixed per category so results compare week to week.
The location
Every question is asked as if from the center of the business's city, on a phone. AI answers change with location, so the coordinate is recorded on the report.
The surfaces
| Surface | How it's asked | Counted as an AI answer? |
|---|---|---|
| Google AI Overviews | Live Google search result from the location; the AI Overview block | Yes |
| Google AI Mode | Google's AI Mode response from the location | Yes |
| Google Maps pack | The map results on the same search | No — recorded separately |
| Gemini | Gemini with Google Search grounding; may differ from the consumer app | Yes |
| ChatGPT | OpenAI's web search, with the city as location | Yes |
| Perplexity | Perplexity's search model, with the city as location | Yes |
| Claude | Claude with web search, with the city as location; the 10 core questions only | Yes |
How answers are counted
- Each question is asked twice per surface. Both results are kept.
- Three outcomes are recorded separately: no AI answer appeared; an AI answer appeared but didn't name the business; an AI answer named the business. Failed samples are excluded and their number is stated.
- "Named" means the business name appears in the answer text. Position is recorded when the answer is a list.
- Competitors are counted the same way, from the same answers.
- When an answer names the business and states a phone number or hours, those are compared with what the owner gave Bella. Where the surface exposes its sources, the cited page is recorded.
What the numbers mean
They are counts of test answers, not of real customer searches. A difference of one or two between weeks is noise; Bella flags a change only when it is large or when it concerns a wrong phone number or hours. Every raw answer is included in the report.
What Bella does not claim
That any change will cause AI to name a business. Listing and profile corrections usually show up in Google and Perplexity answers within a few weeks and ChatGPT can take longer, but the engines decide. Bella measures, points to the listing an error came from, and shows what moved.
Want the numbers for your business? Bella runs the 30 questions from your city and sends the report in 24 hours. $29/month, cancel anytime.
Check my business