Researchers use artificial intelligence to uncover hidden medical symptoms

University of Pennsylvania researchers used artificial intelligence to review over 400,000 forum posts on Reddit about popular weight loss and diabetes medications. By applying large language models, the team categorized patient discussions spanning more than five years to look for real world experiences that traditional clinical trials might have overlooked.

The automated system successfully flagged well known side effects like nausea, but it also highlighted unprompted complaints regarding temperature changes and irregular menstrual cycles. Neil Sehgal, the lead author, noted that nearly four percent of the analyzed users mentioned changes to their periods. While formal clinical studies prioritize dangerous health risks, everyday patient concerns often surface first in online support groups.

The researchers emphasized that these social media findings show correlations rather than proven direct causes. It remains uncertain how drug regulators and pharmaceutical companies will integrate informal online chatter into official safety monitoring. Using artificial intelligence to process public discussions could speed up how quickly medical experts spot potential side effects, but verifying these internet signals will still require formal medical follow up.