Mathematicians can now use AI to prove theorems — and the field is arguing about whether that matters

AI systems have begun proving research-level mathematics theorems, both formally (in ways that computer verification systems accept) and informally (in ways mathematicians can read and check). The shift means mathematicians now have to decide whether to incorporate AI into their practice or resist it — and either way, the field's working methods are about to change.

For decades, mathematics has been one domain where human expertise looked irreplaceable — proof requires originality, intuition, and a kind of creative rigor that seemed to belong to human mathematicians. If AI can now do this work at research level, it doesn't replace mathematicians, but it does change what the job is. The real question isn't whether AI will prove all theorems faster than humans. It's whether mathematicians who ignore this tool will be at a disadvantage compared to those who use it — and whether the field's standards for what counts as a valid proof will shift.