Stop Using AI to Write Your Hot Takes (Mine Included)
A generated opinion reads like a generated opinion, and the point of a take is that a specific person risked being wrong.
I see AI-written 'hot takes' everywhere now, and they're lukewarm. The model optimizes for the least offensive strong statement, which is a contradiction. A real take is risky, specific, and a little embarrassing in hindsight. None of that survives a prompt.
The giveaway is structure: three balanced points, a nod to the other side, a tidy close. That's a committee, not a conviction. When everyone's take has the same shape, the take is just packaging, and packaging doesn't change anyone's mind.
I'll use AI to sharpen a take I already have, to find the weak premise, to argue the other side so I'm not fooling myself. I won't ask it to have the take. The opinion has to be mine before the tool touches it, or it becomes the average of every opinion, which is noise.
If your post could have been written by anyone's assistant, it will be read by no one as yours. The risk of being wrong in public is the whole value of a take. Outsource that and you've outsourced the point.
Updated October 28, 2026: the structural tell below has only gotten easier to spot.
Frequently asked questions:
Q: Why do AI-written hot takes read flat?
A: The model optimizes for the least offensive strong statement - a contradiction. A real take is risky, specific, and a little embarrassing in hindsight.
Q: What's the structural giveaway?
A: Three balanced points, a nod to the other side, a tidy close. That's a committee, not a conviction.
Q: How should I use AI for opinions?
A: Sharpen a take you already have - find the weak premise, argue the other side. The opinion has to be yours before the tool touches it.