AI is speeding up the policing of dodgy science papers.
Peer review struggles to catch all bad research. "Paper mills" churn out quick, low-quality work. Corporations ghostwrite to push favorable agendas. Traditional checks like Retraction Watch, Data Collada, and investigative journalism help, but they’re slow and limited.
Now, AI tools like ImageTwin and Proofig scan millions of images for fakes and manipulation. Natural language processors flag weird “tortured phrases” common in fake papers. Dashboards like Semantic Scholar track whether studies get cited positively or negatively. 
The big change? AI that can do math and logic checks on papers, spotting errors human reviewers miss. The Black Spatula Project already finds algebra flaws in published proofs. These models could audit all scientific papers worldwide.
This audit will expose outright fraud but mostly show that lots of papers are just routine and rarely cited. To the public, this could look like science is breaking, feeding mistrust and misinformation. AI might be seen as an impartial auditor, and bad actors could exploit that.
Experts warn scientists must reset expectations. Most research is incremental, not earth-shattering. Overhyping results fuels disappointment. A more humble, honest portrayal of science as a collective effort will better survive AI’s scrutiny.
The upcoming AI-led audit could come from any player: governments, think tanks, or bad actors trying to shake trust. Scientists pushing transparency now might steer this moment toward renewal – waiting risks fueling a broken science narrative.
Science’s power has never been infallibility, but its willingness to fix itself. AI just raised the stakes to prove it publicly.
Image credit: Jamillah Knowles & We and AI, CC BY-SA
