AI Agents Can Independently Replicate Academic Papers, Mollick Says Original Studies Have Higher Error Rate

Ethan Mollick, a professor at the University of Pennsylvania’s Wharton School, says AI agents can now autonomously replicate complex academic studies using publicly available descriptions of their methods. The capability could lower the labor and cost barriers to research validation, subject more papers to rapid peer scrutiny and challenge current rules at some journals that prohibit the use of AI in peer review. With transparent processes in place, AI could become a tool for assessing research credibility.

As of July 20, 2026, Mollick had observed that when AI agents failed to reproduce research findings, the problem often lay with the original human-authored paper rather than an error by the AI. However, reports did not disclose the sample size, error rates, research completion dates or funding involved, making it impossible to quantify the gap between AI and human researchers. For now, the key development is that AI can independently conduct low-cost validation.