Artificial intelligence software accelerates manuscript production but reduces total processed publications by congesting expert peer review. People assume faster writing automatically yields more published discoveries, but downstream verification forms an unyielding physical bottleneck. Extra submissions fill incoming review queues faster than human referees can read them, generating severe delays that distort submission choices and reduce system output.
When authors adopt writing assistants, they generate manuscripts at higher rates without creating any new referee hours. The resulting influx resembles widening highway entry ramps while leaving the single exit booth unchanged. Rising wait times create a congestion wedge that alters project abandonment decisions and thins the submission stream. If researchers redirect their daily hours toward producing drafts instead of evaluating peers, total completed reviews collapse.
Operations researchers constructed a stochastic service model coupling Poisson project arrivals to stationary review systems through inverse mean occupancy maps. They calculated a system independent Red Queen attenuation factor to measure the exact fraction of upstream writing gains that survives as completed evaluations. Numerical experiments with public conference data confirmed that the unique equilibrium remains stable even across variable service times and finite server architectures.
Academic organizations can now model submission behavior to design linked review credits and verification standards that prevent systemwide congestion. Planners can also evaluate how targeted capacity expansion alters submission thresholds for heterogeneous research projects before expanding intake.