When a new drug spreads faster than the evidence that should accompany it, the consequences can be hard to track. That is the situation with certain unapproved weight-loss treatments that have circulated widely, particularly the glp-1 class of medications. Regulators cannot easily monitor a product that was never formally approved, and clinical trials do not capture every effect once a drug reaches the real world.
To fill that gap, engineers have turned to an unusual source: public online conversations. By combining artificial intelligence with discussions on Reddit, where users candidly describe their experiences, the researchers set out to identify potential side effects that might otherwise go unnoticed. The approach treats the internet as an enormous, unstructured listening post, using machine learning to sift genuine signals from the noise of anecdote and exaggeration. The work is a reminder that real-world evidence can come from unexpected places, and that patients themselves are often the first to notice problems. It also raises questions about accuracy, privacy and the ethics of mining personal stories for medical insight.
