Original data, updated continuously
How often do AI assistants actually name a local business?
We ask web-grounded AI assistants the questions real customers ask, like “who's the best plumber near me?”, about the local businesses we measure, every week. This page publishes the aggregate results: 2650 engine answers about 10 US local businesses over the trailing 90 days.
of answers named the business they were asked about (695 of 2650)
US local businesses measured, 83 weekly measurement runs
engines probed, reported per engine, never blended into one score
Per-engine naming rate
Cross-engine overlap in AI answers is known to be low, so a blended score would mislead. Each row is one engine's own rate.
| Engine | Answers | Named the business | Rate | Answered with no sources |
|---|---|---|---|---|
| Perplexity | 843 | 237 | 28% | not recorded |
| ChatGPT (OpenAI) | 707 | 194 | 27% | not recorded |
| Google Gemini | 620 | 203 | 33% | 0% of 195 |
| Grok (xAI) | 480 | 61 | 13% | 63% of 141 |
Where the answers came from
When an engine cites its sources, we record them. These are the domains behind the cited answers in this dataset, by number of answers citing them. Answers that came back with no sources contribute nothing here.
Method and caveats
Each measured business gets customer-style recommendation prompts (category + city, plus its own tracked phrases) run through every enabled web-grounded engine weekly. An answer counts as naming the business only when the business is identifiably present in the answer text. An engine that fails to answer is omitted, never counted as a miss.
Not every answer is grounded in a live search. Some engines answer from their own recollection of the market and return no sources at all, and an answer like that still names a business (or fails to) exactly like a cited one, so it is counted the same way in the rate. The last column above says how often each engine did that, over the answers where we recorded it. We started recording it on 17 August 2026, so earlier answers in this window read as not recorded rather than as zero.
Caveats we want you to know: this is a small, growing network of real businesses, not a random sample of all US businesses; businesses that sign up for visibility monitoring may differ from average; AI answers vary run to run. Rates here describe this dataset over this window, nothing more. Figures update as the network grows.
Cite freely with attribution (CC BY 4.0): Radveo, Local AI Visibility dataset, 2026-08-25.
Want to know what AI says about your business? Run the free check, no signup.