We ask ChatGPT, Gemini and Perplexity the questions real buyers ask, about real local businesses, and record who gets named. Every study states its sample, its date, its engines and its exclusions, so anyone can check the work or repeat it.
Across 17 businesses, whether the engines cited a company's own website predicted whether they recommended it (Spearman rho +0.92). Google review count did not (+0.25). Every business never recommended had zero citations to its own site.
17 businesses · 12 questions each · 3 engines · 1,306 citations
21 of 29 dental practices were named in none of the questions a patient asks before choosing a dentist. Five of the ten most-reviewed were among them, including one with 1,180 reviews and a 5.0 rating.
29 practices · 17–31 questions each · 3 engines
For each business we generate the questions a buyer actually asks before choosing — “best dentist in Austin”, “affordable plumber near me” — and put them to OpenAI, Perplexity and Gemini. We record whether the business is named in the answer, and which businesses are named instead.
Businesses whose website lists a different city are excluded, because the questions would be built around the wrong market. Results where an engine fails to return a usable answer are excluded and reported as such rather than counted as an absence. Sample sizes are small and stated; we would rather publish 29 businesses honestly than imply a national study.
The same check, on your website, in about 20 seconds. No signup.