ChatGPT's local business recommendations come from two sources: its training data (which has a cutoff date) and, when enabled, real-time web browsing. For local business queries, the balance between these sources depends on the query type and the user's ChatGPT version.
In training data, ChatGPT learned from web crawls that included business directories, review sites, local news coverage, and business websites. Businesses that appeared frequently and consistently across these sources in the training corpus are more likely to appear in recommendations.
For real-time browsing queries, ChatGPT functions more like Perplexity — assembling an answer from live sources. In this mode, verified data networks and structured website content play a larger role than historical training frequency.