Here's the problem with AI overviews. Regular AI is just predicting the next word so you can see how it goes to the riddle and gets this wrong. RAG AI is supposed to ground this by summarizing search results, but the search results are so broad and generic right now that those too misunderstand the query and go right to the popular riddle - and you get a crap answer. This is over exaggerated by my example, but it holds true for all complex non-trivial queries.
100% and RAG is definitely not providing the grounding they were hoping for and has gotten consistently worse in general.
Other databases adding further value via RAG too?
Founder, CEO, Fractional CMO
6moThe Google Search team has been doing a great job with patch updates since the launch of SGE. There is just no quick fix to this problem. They're improving SGE faster now that it's in production but at what cost? It looks like the AI overview LLMs are leaning too hard on topical authority.