I think this is the strongest protection against this attack. You’d need to identify enough people at enough varied polling locations to be significant enough to sway an election.
Do too many at one location and it raises flags. Cast a vote with a name that also votes absentee or at another location, raises flags.
You’d have to distribute enough fake votes over a large enough area and across enough different shifts to not get anyone’s attention. And that’s expensive and hard to keep secret due to how many people would be involved.
Don’t rely too much on ChatGPT here. It’s not a human and isn’t going to respond the same way a human would.
If you over practice with that tool you might become great at debating ChatGPT but that may entrench bad habits that hurt you when debating a human.