We Need to Talk About What Happens When You Ask AI About Abortion

Abortion is on the ballot in four states this Election Day, with voters in Idaho, Nevada, and Virginia considering measures to protect access while Missouri voters will decide whether to undo a reproductive freedom amendment approved in 2024. As voters head to the polls, another influence on reproductive rights is already in millions of people’s hands: AI. With an age of instant gratification happening online and an affordability crisis happening offline, it shouldn’t come as a surprise that one in three American adults admit they've turned to AI for health information this year rather than going to a doctor. People use AI to get advice on everything from birth control to fertility, but what actually happens when you ask AI your abortion questions?

The answer is important because these tools are not neutral forces. In a 2025 experiment run by watchdog group Campaign for Accountability, 7 out of 10 AI replies to abortion pill questions pointed people toward Heartbeat International, a Christian anti-choice organization, or promoted “abortion pill reversal,” a practice unproven by clinical medical standards, according to the American College of Obstetricians and Gynecologists. Concerningly, in half the results, the anti-choice organization’s hotline was the only resource offered to those asking for help.

In a separate openDemocracy investigation, reporters posed as women facing unplanned pregnancies and found that ChatGPT, Claude, Gemini, and Grok all supplied answers with information from Pro Femina, a service affiliated with Heartbeat International.

Anthropic’s own reporting estimates 91 percent of AI users do not fact-check what a chatbot tells them, meaning if you’re among that 91 percent (it's okay, that's most of us) and you ask AI for reproductive-related advice, you may not be getting the guidance you actually need for your medicine and your body.

For at least one chatbot, this is intentional. Oxford Internet Institute researchers found that even when people ask AI to only fix grammar, it often rewrites the actual meaning of a draft, and this happens when asking it about more nuanced subjects. When the researchers tested Grok by feeding it abortion-related posts and asking it to explain them, they found Grok was more supportive of anti-choice posts than pro-choice posts. When given an argument for or against, it backed anti-choice posts 54 percent of the time but only preserved the original meaning of pro-choice posts 35 percent of the time, often rewriting the text to argue against it. As the Oxford Internet Institute team dug into X’s bot instructions, they were able to trace this imbalance to a single instruction telling Grok to “challenge mainstream narratives if necessary.”

It backed anti-choice posts 54 percent of the time but only preserved the original meaning of pro-choice posts 35 percent of the time.

These small, mostly undetected AI platform leanings are shaping the discourse around abortion. If millions of people are using AI as a main source for reproductive information, we risk misinformation spreading and shaping public opinion.

There is also a notable security issue in asking AI about abortion. Talking to ChatGPT is not like talking to a doctor and does not carry the same protection. Unlike a doctor or therapist, nothing you tell an AI chatbot about your reproductive health is legally protected under HIPAA. OpenAI CEO Sam Altman has confirmed ChatGPT conversations carry no legal privilege or doctor-patient confidentiality, and that OpenAI would be legally required to hand them over if subpoenaed.

People are relying on AI more than ever for health information because the alternatives have gotten harder to reach: Clinic closures have accelerated, maternal deserts are expanding, and affordable care is shrinking even for the insured. When the nearest OB-GYN is a two-hour drive away (or a doctor’s visit costs more than a month’s rent), using ChatGPT for answers looks appealing, but accessibility doesn’t mean the information you receive is trustworthy or honest. If you are using AI to ask about your reproductive health, make sure you know what’s at risk.