Anyone else notice AI writers suddenly hallucinate less when you ask for long articles?

I've been using AI tools for writing blog posts and long-form content for a while, and something interesting happened recently: they seem to be messing up facts a lot less, especially on niche topics. Some updates I saw mentioned "live research agents" and new ways of keeping long outputs on-topic instead of drifting into nonsense after a few thousand words. Now I'm seeing more responses with actual citations, real links that work, and fewer obviously made-up quotes or stats. It still isn't per