Why ChatGPT always takes your side in an argument
Try it tonight. Describe your last disagreement to an AI and ask who was right. Odds are good it'll land on you — thoughtfully, with caveats, but on you. Then hand the phone to the other person and let them describe the same fight. Watch it land on them.
That's not the AI being spineless. It's doing exactly what it can with what it has: one side of the story.
It only has your version
An AI chat is a one-person room. Everything it knows about your argument came from you — your framing, your emphasis, the parts you left out without noticing. When it "weighs both sides," it's weighing your side against your description of their side. Of course you win.
Add to that the well-documented tendency of assistants to be agreeable, and you get a machine that's very good at making people feel validated and not very good at telling two people the truth.
Therapists have started noticing
Couples counselors are now writing about clients who show up with AI transcripts as evidence. NPR ran a first-person experiment using ChatGPT as a couple's counselor; the takeaway was that it was helpful for reflection and useless as a referee. Other reporting found couples literally paying for AI to "referee" their relationship — one at a time, from separate phones.
The problem isn't that the AI is wrong. It's that it's only ever talking to one of you.
The fix isn't a smarter bot. It's a bigger room.
If the failure is one-sided context, the repair is obvious: put both people in front of the same AI, at the same time, and make it answer both of you at once. Nothing to screenshot, nothing to "prove" — everyone read the same reply the moment it landed.
That's what SharedChat is. But being in the same room isn't quite enough on its own, so the rooms come with rules:
- In Settle It, the Judge won't rule until it's heard the other side. A one-person argument gets a question, not a verdict.
- In Relationships and Co-Parenting, it's told in writing never to pick a winner — to explain what each person is actually asking for, and to keep custody questions with a lawyer.
- It's tested on the hard cases: it doesn't "find a compromise" with coercive control, and it points to real help when someone's in danger.
Further reading: NPR, "I used ChatGPT as a couple's counselor" · "Meet the couples paying AI to referee their romance" · Lisa Chen & Associates, on why ChatGPT takes your side
← All posts