For six months, my husband convinced everyone that I had stolen millions from the company I built, including our own daughter. As I sat in federal court, certain I was about to lose my freedom, the doors suddenly opened. My nine-year-old son walked in carrying an old blue backpack and said, “Your Honour. I know who framed my mom. They’re sitting in this room.”

This is all part of a thread on X that appeared one day. The first-person narrator probably didn’t exist; they were just a 21-year-old stock trader in Nigeria using AI to generate stories and get paid for it by X. The algorithm noticed my interest and fed me more of these melodramas—stories of redemption, where good triumphs over evil.

It’s pretty obvious that most of the stories are fabricated. They’re presented as true first-person narratives but contain tropes like an innocent person being wrongfully fired or evicted, only to be redeemed in the end. These mini-melodramas routinely garner hundreds of thousands, sometimes millions, of views. That’s a lot of dopamine, it seems.

Professor Rob Cover from RMIT Digital Ethnography Research Centre broke down their structure for me: they start with rage-baiting setup and then turn around to show the underdog redeemed while the adversary is humiliated. Jenna Russell, a PhD candidate at the University of Maryland studying AI writing, says it’s simple to generate these mini-melodramas. They’re optimized for engagement using easy, well-known tropes.