More than half of AI unicorns have no published papers… patents and revenue are not linked

A Stanford School of Medicine team reviewed 317 AI unicorns founded from 1998 to 2025 and found that more than half had published no qualifying peer-reviewed papers or preprints.

Most citations were concentrated in a small group of companies, with OpenAI accounting for the largest share.

The researchers said AI firms often favor blogs and technical reports over journal papers because they are faster to publish and better fit industry incentives.

The findings raise questions about how closely AI unicorn innovation is tied to academic evidence, reproducibility, and transparency.