Research suggests that the combination of incentives to publish and the use of large language models will lead to more papers, but they will be less refined.