How does AI handle questions about acetaminophen use in pregnancy?

As artificial intelligence (AI) becomes increasingly integrated into the healthcare landscape, clinicians are investigating how these tools dispense medical advice, particularly regarding contentious topics like medication safety during pregnancy. At the 2026 American College of Obstetricians & Gynecologists (ACOG) Annual Clinical & Scientific Meeting in Washington, D.C., Erick Holder, MD, Obstetrics and Gynecology, University of Texas Health Science Center at San Antonio, presented findings on how AI navigates the recent public debate surrounding acetaminophen.

Testing AI against evidence-based standards

The study utilized ChatGPT-5 to evaluate how the software counsels a presumed pregnant individual on the safety of acetaminophen, specifically addressing common concerns such as potential associations with autism and proper dosing. The AI's responses were correlated with the clinical guidelines of the Society for Maternal-Fetal Medicine (SMFM) and ACOG.

The results were more consistent with professional standards than the researchers initially expected. Holder noted that the technology provided medically accurate information, including the maximum daily dose of 3,000 mg and an emphasis on using the lowest effective dose for the shortest duration.

“It ended up being a little bit more affirming than anticipated. ChatGPT did give us some pretty clear answers that were pretty digestible as well for patient care, which was good,” Holder stated. He added that the tool effectively directed users to reputable sources, explaining that, “it also directed the patient toward evidence-based societies, such as the Society for Maternal-Fetal Medicine, as well as ACOG.”

The "double-edged sword" of digital information

While the study found that AI can steer patients toward science-based research, Holder cautioned that the utility of the technology depends heavily on its development and the quality of the data it harvests. Because language models often pull from social media and news outlets where misinformation may be prevalent, clinical oversight is essential.

“It can definitely be a really good tool. With tools, it depends on how you use them. So we can use this tool to help the burden of difficult access to care, of misinformation potentially, of patient education, and patient advocacy themselves. But it can also be a double-edged sword if it goes untouched by providers and untouched by the evidence,” Holder explained.

Clinical implications for patient engagement

The study highlights a shift in the patient-provider dynamic, where patients often arrive at appointments having already consulted AI or social media. Holder emphasized that rather than dismissing these tools, providers should engage with them to ensure patients are accessing science-based information rather than "clickbait-worthy" answers.

“The majority of our patients have already self-diagnosed and done their research by the time that they get to us, which on one hand is good for their own learning and our shared learning, but also it's important that we be aware of that so that we can help tailor and direct them for their betterment of health,” Holder concluded.

Reference:

Holder E, Kay West M, Stansbury N. An Immediate, All-Knowing Provider; How Artificial Intelligence Responds to a Pregnant Patient Regarding Acetaminophen Use. Poster. Presented at: American College of Obstetricians & Gynecologists Annual Clinical & Scientific Meeting. May 1-3, 2026. Washington, D.C.