A test comparing ChatGPT and Perplexity's local business recommendations across California cities found only 25% agreement on recommendations, with the two tools agreeing on the #1 result just once in 20 searches. The tools appear to rely on different web sources, with Perplexity favoring business websites and ChatGPT emphasizing third-party review directories.
An AI visibility tool creator discovers that mention counts mislead: being named in AI answers differs fundamentally from being recommended. They develop a decision model using Jev through OpenRouter to distinguish between recommended, listed, dismissed, and absent mentions, finding that over half of answers recommend nobody despite naming products.