Comment by simonw

Something I've found surprisingly effective is telling ChatGPT to "use credible sources" - you can then watch its thinking trace and see it do things like ruling out random blogs, considering media publications with a good reputation for fact checking, and double-checking information that seems unlikely.