I really think this is a "drink the koolaid" phenomenon. Personally, I don't want an LLM synthesized result to a query. I want to read original source material

I really think this is a "drink the koolaid" phenomenon. Personally, I don't want an LLM synthesized result to a query. I want to read original source material on websites, preferably written by experts, in the field in which my search is targeted. What I find in serious regression in search, is interpretation of the search query. If I search for something like "Systems containing A but not B" I just get results that contain the words A and B. The logical semantics of asking for "not B" is completely ignored. Using "-B" doesn't work, since many discussions of something that doesn't have B, will mention the word B. These errors didn't seem to be so egregious historically. There seemed to be more correct semantic interpretation of the query. I don't know if this has to do with applying LLMs in the backend of search, but if LLMs could more accurately interpret what I'm asking for, then I would be happy to have them parse my queries and return links that meet my query specifications more accurately. But again, I don't want a synthesized result, I want to read original source material. I see the push to make everything LLM synthesized prose, to be just another attempt to put "Big Tech" between me and the info I'm trying to access. Just link me to the original info please... p.s. Something like the "semantic web" which would eliminate any 3rd party search agent completely would be the ideal solution.