AI Flags Ominous Ozempic Blind Spot

Researchers fed 400,000 Reddit posts into an artificial intelligence system and found people taking Ozempic and similar drugs are reporting menstrual changes and body temperature swings far more than official drug trial data ever showed.

Quick Take

University of Pennsylvania researchers analyzed 410,198 Reddit posts from May 2019 to June 2025 about semaglutide and tirzepatide, the drugs behind Ozempic, Wegovy, and Mounjaro.
About 67,000 users self-reported taking the medications, and 43.5% of them described at least one side effect in their posts.
Common complaints like nausea and constipation matched known clinical trial results, but menstrual irregularities and temperature problems stood out as under-discussed symptoms.
The study’s lead author says the findings show patterns, not proof that the drugs cause these symptoms.

An AI Combs Through a Massive Pile of Patient Complaints

Neil Sehgal and his team at the University of Pennsylvania built an AI model to scan hundreds of thousands of Reddit posts written by people taking GLP-1 drugs. The goal was simple: find out what patients say about their bodies when no doctor is in the room. The scale of the dataset, over 400,000 posts, gives the findings real statistical weight.

The AI didn’t just count complaints. It sorted them into categories, then compared the frequency of each symptom against what shows up in official drug trials. Nausea, vomiting, and constipation topped the list, which lined up well with what clinical trials already documented. That match gave researchers confidence the tool was reading the data correctly.

The Symptoms Nobody Was Talking About

Buried inside those complaints, the AI found something trial data barely mentions. Nearly 4% of users reported menstrual irregularities, including heavy bleeding and disrupted cycles. Another cluster of posts described chills and hot flashes, pointing to problems with body temperature regulation that current drug labels don’t fully capture.

These aren’t headline-grabbing numbers on their own. But they represent thousands of real people describing symptoms that never made it into the studies drugmakers submitted for approval. For women managing both weight loss and reproductive health, that gap in the official record matters. A missed side effect is still a side effect, even if it shows up on Reddit instead of in a lab.

Why the Researchers Are Holding Back on Blame

Sehgal was careful not to overstate what the AI found. He said the model detected patterns in how often people mentioned certain experiences, but the study does not claim the drugs caused those symptoms. That’s an important distinction. Correlation in a pile of Reddit posts is not the same as a lab-confirmed cause.

Reddit posts don’t come with medical charts attached. Nobody verified dosage, how long someone had been on the drug, or whether other health conditions were in play. The study also lacks a comparison group of non-users, so there’s no clean way to know if these symptom rates are actually higher than normal background rates in the population.

Where the Research Goes From Here

Other pharmacovigilance work backs up the idea that patient forums can catch things trials miss. A separate FAERS pharmacovigilance analysis published in Obstetrics and Gynecology found semaglutide carried a broader menstrual signal than other GLP-1 drugs, though that study also stopped short of proving cause and effect. Social media listening has a track record of flagging real problems earlier than formal reporting systems, even when the data itself stays messy.

The next real test comes from matched clinical studies, not more Reddit scraping. Researchers now have a lead worth chasing: menstrual changes and temperature swings tied to two of the most widely prescribed drugs in America. Millions of patients are taking these medications right now. A follow-up study with real medical records, not just forum posts, would tell doctors and patients whether this signal is something to actually plan for.