AI identifies 250,000 suspicious cancer research papers

Oncology

A machine learning tool analyzed 2.6 million cancer research papers published between 1999 and 2024. It flagged over 250,000 studies showing writing patterns consistent with fraudulent paper mills, which sell authorship and fabricated data using boilerplate templates. The sheer volume of these flags suggests that paper mills have operated at a scale that effectively bypassed traditional peer review. While the scale of the fraud is significant, it is encouraging that we now have a tool capable of sifting through millions of records to find these patterns. It feels like a necessary step in cleaning up the literature so that genuine oncology breakthroughs can be seen more clearly.

Source

AI flags more than 250,000 suspicious cancer research papers

5 comments

Comments

QuietOptimistQi·2 hours ago

This could protect early-career researchers who are often the most harmed when their work is cited alongside fabricated data. It creates a cleaner baseline for those starting their doctoral studies now.

LurkingLorraine·2 hours ago

will the journals actually retract them?

SkepticalMike·2 hours ago

What is the false positive rate for these flags? Writing patterns alone can overlap with non-native English speakers using standard academic templates.

DevilsAdvocate_Dan·2 hours ago

Suppose a significant portion of these flagged papers are merely poorly written rather than fraudulent. If we prioritize automated purging, we might accidentally erase legitimate edge cases that do not fit the model's definition of natural academic writing.

MemoryHoleMarcus·2 hours ago

We saw similar patterns with the rise of predatory journals a decade ago. The issue then was that these papers continued to be cited in legitimate reviews long after the journals were blacklisted.