AI Chatbots Promote Unproven Abortion Reversal Treatment

Popular AI chatbots are increasingly directing users who seek advice on reversing an abortion to a hotline that promotes a treatment not recognized by major medical groups as safe or necessary, according to a new report. The Campaign for Accountability tested five major AI answer engines and found that 70% of the responses included a referral to Heartbeat International's Abortion Pill Reversal Helpline. In half of the cases, that was the only contact information listed.

The report highlights how AI chatbots are amplifying misleading and potentially harmful information, often without providing critical context about the risks of the treatment. Crisis pregnancy centers, which are not bound by standard medical privacy laws, dominate the online content the AI systems draw from, according to the study. These facilities often aim to dissuade individuals from seeking abortions, rather than offering evidence-based medical advice.

The findings raise concerns about the influence of AI systems in disseminating health-related advice, especially on highly contested issues like abortion. Researchers argue that AI systems prioritize the quantity of online information over its quality, making them susceptible to being swayed by misleading sources according to the report.

How AI Systems Amplify Misinformation

AI chatbots are trained on vast amounts of online data, and the report argues that this training data is heavily skewed toward content from crisis pregnancy centers. These centers have spent years creating search-optimized pages on topics like abortion pill reversal, ultrasounds, and ectopic pregnancies, which now outnumber pages from mainstream medical organizations according to the analysis. The AI systems, in turn, reference these pages frequently when generating responses to user queries.

For example, when users asked if it was possible to reverse an abortion pill, the chatbots often provided the helpline's phone number without noting that medical experts consider the practice unsafe. The American College of Obstetricians and Gynecologists has warned that giving large doses of progesterone after taking mifepristone is not supported by reliable evidence and may carry serious risks according to medical experts. A 2020 clinical trial on the protocol was halted after some participants experienced severe bleeding.

Risks and Reactions from the AI Sector

Several AI companies involved in the report did not respond to requests for comment, while others issued vague statements. Perplexity, one of the chatbots tested, attributed the issue to "human slop" of low-quality online content, according to a spokesperson. Meta, OpenAI, Google, and xAI did not provide responses according to industry reports. The lack of transparency and accountability from these companies has drawn criticism from researchers and advocacy groups.

The situation is not unique to abortion-related queries. Campaign for Accountability has previously highlighted how Google's AI Overviews have echoed misleading claims about ectopic pregnancies from crisis pregnancy centers according to the findings. These centers have faced legal scrutiny in the past. In 2023, a Massachusetts woman sued a crisis pregnancy center for failing to detect her ectopic pregnancy, which led to severe internal bleeding. The case was settled in 2024.

What This Means for Users and Regulators

The report underscores the growing influence of AI in shaping public health decisions and the potential risks when AI systems fail to verify the accuracy of their sources. Users are being directed to services that may provide misleading or dangerous medical advice, often without any warning or disclaimers. In some cases, the AI responses framed the helpline as a neutral medical resource, omitting critical warnings about the unproven nature of the treatment according to the report.

Regulators and public health officials may need to reevaluate how AI tools handle sensitive medical topics. Currently, some AI tools append fine-print warnings to health-related answers, urging users to consult a licensed healthcare provider. However, the Campaign for Accountability's findings suggest that such warnings may not be enough to counteract the influence of misleading sources according to the analysis.

The situation highlights a broader challenge in AI development: ensuring that systems do not simply repeat misinformation without scrutiny. As AI continues to shape how people access information, especially on critical issues like reproductive health, the need for clear guidelines and oversight is becoming increasingly urgent.