I've seen ChatGPT and other LLMs recommend tools, products, and websites based on user queries.

As a solo maker, I'm curious,

how do these models decide what to recommend?