I routinely use duckai, Grok, chatgpt and gemini for research. I will ask all 4 the same question, and it is amazing how the answers can be different and often wrong.
I’ve found many of the answers regarding historical subjects wrong, subjects I was a part of back in the day and knew the correct answer. The questions were asked to verify my past knowledge and quite disappointingly they do give out wrong info.
I’ve noticed they search the web for answers. If they go to places like Wiki they can and will get wrong info. I’m certain other websites are full of lies too.
I’ve been doing historical research for many years on a relatively arcane 19th century subject area - most being examination of the historiography and prior referenced sources, as well as previously undiscovered primary resources.
It becomes very apparent that AI (rather, LLM’s) are just giant plagiarizing tools, and they regurgitate the most common information available via the web. Much of which is garbage - AI results will be weighted based on mass of material processed, not inherent quality or accuracy.
In this area, it’s a menace.