

This is frustratingly common. Who has the resources and incentive to run such studies?
Finding the same thing with evidence for the effectiveness of AI copilots. Everyone doing wonk on it has a finger in the pie. So in order to believe the research you have to trust the integrity and discipline of people you never met, not to mention trust a whole organizational to be culturally respectful of the scientific method (which probability near zero for a large corporate).
Do they send you hallucinations as if they were fact all the time? That’s my current favourite.