Earn with Nexo
MIT study finds AI deteriorates users’ ability to spot fake news
A four-week experiment shows AI chatbots boost misinformation detection by 21% while actively eroding the critical thinking skills needed to do it without help.
Here’s a fun paradox for the information age: the tool designed to help you spot lies is quietly making you worse at spotting them on your own.
A new study from the MIT Media Lab, published June 9, 2026, found that participants who used AI chatbots to evaluate news credibility saw their independent detection skills crater once the AI was taken away. Their unassisted performance didn’t just return to baseline. It dropped 15.3 percentage points below where they started.
The experiment and its uncomfortable findings
The study tracked 67 participants over four weeks as they assessed the credibility of news headline-image pairs. During the initial phases, when AI assistance was available, participants got measurably better at flagging misinformation. Accuracy jumped 21%.
By week four, with no AI to lean on, participants performed significantly worse than they had before the experiment even began. Not worse than they were with AI help. Worse than they were before they ever touched the tool. The 15.3 percentage point decline below their original baseline means the AI didn’t just fail to teach lasting skills. It actively degraded the ones participants already had.
The researchers framed this as a “dependency paradox,” drawing an analogy to GPS technology weakening people’s innate navigation skills. Short-term convenience, long-term cognitive atrophy.
Why this matters beyond academia
This research builds on earlier work from the MIT Sloan School of Management, which found that AI interactions could reduce belief in false information while users were actively engaged with the system. The new study adds the critical missing chapter: what happens when the AI goes away.
What this means for investors and the broader digital economy
The study’s sample size of 67 participants is relatively small, which is worth noting. Larger-scale replications will be needed before anyone should treat these numbers as settled science.
The 21% accuracy boost is seductive. The 15.3 percentage point skill degradation is the price you don’t see on the label.