I’m not an anti AI zealot either, but frankly humans can skim search results and “link dive” way better than current LLMs can, if they’re familiar with the web.

Sometimes I use GLM 4.6's deep research mode with a link to go off as a supplement (I’d recommend that over Claude; it’s great! And open weights!), but 90% of the time just poking through Google/DDG with a filter to block SEO/AI spam yields better reviews. We can just parse more links and make better snap “credibility” judgements than LLMs can.