Is content convergence an incentive problem, not an AI problem?

We’re producing more content than ever, but most of it is forgettable. Different companies, different products but same feel.

People blame AI for being generic. I think that’s the wrong diagnosis.

The real change is cost. Generative AI made publishing almost free. When output is cheap, systems optimize for whatever is easiest to measure: speed, volume, engagement. Once that happens, convergence is expected.

Sameness isn’t a bug. It’s what optimization against shared proxies produces.

There’s some early evidence for this. In at least one natural experiment, limiting access to LLMs increased content diversity. Other studies suggest AI-generated communication is trusted less over time, even when quality looks fine.

This looks less like a creativity problem and more like an incentives + feedback loop problem.

Question: if publishing stays frictionless, is convergence inevitable? If not, what constraints would actually prevent it?