I'll tell you what I've learned about AI. It's awesome for creative stuff--I asked Grok to write a judicial opinion overturning the verdict in The Trial from Side 4 of The Wall, using 80's pop hits as precedent, and its response floored me. I thought it was awesome.
But as a research tool, you have to double-check its results, because they're frequently wrong. Just as one example, when I asked it to calculate the total tariff for the tricorder's HTS code, it was correct except for the Section 301 tariff, which it mistakenly told me was on List 3 at 25%. Checking the Section 301 database, I found it's actually on List 4, at 0%. I could give you other examples. Sabine Hossenfelder, a German physicist who does YT videos, has said she frequently gets scientific paper citations from every AI that simply don't exist. AI's literally make up quotes and citations. She said this happens all the time.
So AI, as it is right now, is hardly the game-changing technology it's purported to be. Don't use it to summarize research without double-checking its findings with original sources.