UNC study: AI-generated novel characters are flat, safety rules are to blame

UNC Chapel Hill researchers used an automated tool called CASPER to look at thousands of AI-written novels. They found the characters are often flat and predictable.

The team checked characters across 8 areas, like how natural or consistent they feel. They found AI characters often rely on stereotypes and stories almost always end happily. The main reason? AI's built-in safety rules.

Lead researcher Annelise Blay says human writers leave things unclear and let readers wonder. That mystery is what makes stories memorable. The study shows a key weakness in AI storytelling.