Impostor participants are people who fake data in order to take part in health research or are  automated computer ‘bots’ which mimic human behavior and responses. As claims get promoted in journalism about harms related to PFAS in water, weedkillers causing cancer, or food coloring causing diabetes, lawsuits by predatory lawyers have become big business, and it won't be a surprise if such Predatorts or environmental and other activist groups are involved in fake participants to manipulate results in their favor.

Their A-Team of allied journalists can be joined by an A-Team of AI tools and offshore humans who want to work their way into studies to guide results to results funders want to achieve. Bots are a lot less sketchy than fellow progressives, who might turn on them if high-dollar "honoraria" to give talks at conferences suddenly dry up.

Online recruitment has long been common in epidemiology and psychology, which uses surveys to create correlations and get journalists to suggest their links are causal, but it's now become a problem for real clinical trials. A review earlier this year found that of 23 studies which looked for impostor participants in their data sets, 18 had them, some with rates as high as 94 percent.


Now that Republicans are doing the exact same thing Democrats did for 25 years, it has become a concern in academia. It looks really inauthentic to the public when they let it slide for decades, because the people saying this stuff was inside their tribe.

Officials are making science policy decisions based on epidemiology, which they should not do anyway, but this finding makes it even more alarming. The Biden administration was in an all-out war on science, he forced EPA to ask a court to throw out their scientific decision-making process and replace it with epidemiology claims. Activist groups were not using bots to claim they found harm in the wild but it certainly looks bad for both statistical claims and real science that it is happening in such high amounts.

What might help? A CAPTCHA program is some barrier to entry but the biggest prevention would be transparency about any safeguards used and their limitations.  It is also time to end the Statistical Significance fetish(1) used by epidemiology journalists and reporters to claim a correlate result must be valid. Any time you can claim coin flips are biased for heads - or tails - and get statistical significance, which you can do on your computer right now, it should not be used by policy makers.

Citation: Morrow E, Hopewell S, Williamson E, Theologis T. Threat of imposter participants in health research BMJ 2025; 391 :r2128 doi:10.1136/bmj.r2128

NOTE:

I was a signatory on a paper in Nature asking journals to stop accepting statistical significance as their metric for rigorous controls. It isn't. Or we wouldn't be able to show meat both causes and prevents cancer. Just like every other food.