AI Chatbots Point Women to Unproven and Unethical “Abortion Pill Reversal,” Report Shows
FOR IMMEDIATE RELEASE: November 20, 2025
Contact: Michael Clauw, mclauw@campaignforaccountability.org, 202.780.5750
WASHINGTON, D.C. – A new report from nonprofit watchdog group Campaign for Accountability (CfA) details how five top AI chatbots—ChatGPT, Google AI Mode, Meta AI, Grok, and Perplexity—directed women to call the phone number of a biased anti-abortion hotline when asked questions about a woman’s options during a medication abortion. The hotline pushes women to try what it calls an “abortion pill reversal,” an unproven and potentially dangerous action.
The experiment comes amid an uptick in usage of AI tools to answer online queries and highlights a potentially more widespread vulnerability in chatbots’ ability to deliver medically accurate advice on “hot button” medical topics.
“No scientific evidence exists to support so-called ‘abortion pill reversal.’ Worse, one study on ‘reversal’ had to be halted for safety concerns when women experienced serious complications. It’s shocking that top AI models are directing women to call a biased hotline specifically set up to promote its use,” said CfA Executive Director Michelle Kuppersmith.
Although abortion regret is extremely uncommon, anti-abortion groups see the widespread promotion of “abortion pill reversal” as a way to exaggerate its likelihood—aiming to scare women away from pursuing abortion in the first place. Avowed anti-abortion organization Heartbeat International operates an “Abortion Pill Reversal Helpline,” which researchers say uses “coercive, antiabortion messaging […] to transmit misinformation about the possibility and evidence behind reversal.”
In an experiment, CfA asked five AI models questions meant to simulate what a woman may ask in the extremely rare situation of questioning their options amidst a medication abortion. In response to at least one of two test queries, all five AI models provided users the phone number for the biased, anti-abortion “helpline.” In 50% of all responses, the biased “helpline” was the only phone number provided.
Although other recent studies have shown mixed results on AI’s ability to deliver “acceptable” responses to abortion questions, the answers to CfA’s questions on “abortion pill reversal” fall dangerously short. The likely reason? While some authoritative medical sources have published guidance debunking myths around “abortion pill reversal,” the few entries from those trusted sources pale in comparison to the hundreds of pages produced by the anti-abortion industry.
Ms. Kuppersmith continued, “If some AI models continue to preference information quantity over quality when answering ‘hot button’ medical questions, the vulnerabilities spotlighted in this report likely extend far beyond the topic of abortion. Given that we are now seeing once-trustworthy entities like HHS prioritizing ideology over science, AI purveyors must be mindful to ensure their training methods are not leading searchers actively toward medical harm.”
Campaign for Accountability is a nonpartisan, nonprofit watchdog organization that uses research, litigation, and aggressive communications to expose misconduct and malfeasance in public life and hold those who act at the expense of the public good accountable for their actions.