Why Large Language Models Won't Fix Google

Google's problems are not technical but business.

They started out with ads clearly separated from the search results because they thought otherwise it would look illegitimate.

They discovered that if the search results were good in that case, nobody would look at the ads. They brought in a substantial amount of money, but they were under pressure to put up better numbers quarter after quarter so they (1) degrade the search results, and (2) blend ads into the search results so frequently you don't see any real search results on the page.

If LLMs or some other technology improves the search results and displaces advertising, that's a problem for Google, not a solution.

I don't think Google is afraid LLMs will "fix" all its issues. I think they're more afraid that their seemingly unassailable moat is suddenly exposed to disruption.

People use search engines to find answers to questions. If ChatGPT can give an answer faster and in a more natural format then people will flock to it. Most won't care that the quality of the response is as low or lower than what a search engine returns.