LLMs cannot search, this is just responding with something it may or may not have read. Have you verified these numbers are in fact correct? Have you tried "searching" for anything outside of its training data? ChatGPT will happily hallucinate references. If you have not experienced this yet, you haven't seen the "true" side of LLMs.

There will be plugins for search, but will still need all the crawling and scraping and serving to be available. LLMs will offer an alternative interface to search, but not always the best one. One thing the LLM does not give you is a spectrum of options. How do you know it didn't miss something. This is where having to "troll" through search results is actually worth the effort.