AI & Abortion Care: How people are using AI to find abortion care, and how this might be dangerous

Abortion is a common healthcare procedure in the UK: 1 in 3 women will have one in their lifetime.[1] Despite this prevalence, it remains a heavily stigmatised and taboo topic. That gap, between how routine it is and how shameful it's made to feel, might explain why so many people now turn to Artificial Intelligence (AI) chatbots when they need information.

AI chatbots like ChatGPT, Claude, Google Gemini, Meta AI, and Grok have become a new way that people find information about abortion care.[2] These tools are fast, available at any hour, and they feel private [3]. You don't have to sift through pages of search results. For many people, asking a chatbot about abortion probably feels emotionally easier than calling a clinic. There is no receptionist, no unknown of how whoever picks up a helpline will react, no worry about ‘burdening’ a healthcare provider with questions. You just ask a question, and something answers, and it does so in a confident, authoritative tone that can make it easy to take the information at face value.

But that’s also part of the problem. The confidence these tools project has almost no relationship with their accuracy. And on abortion specifically, the consequences of that gap are real.

A 2025 study by Xu et al. evaluated ChatGPT-3.5 on common abortion questions, assessed by a team of healthcare providers. Responses were rated acceptable in only 65% of cases, and complete in just 8%. The researchers concluded that the tool “can regurgitate facts found online, but still lacks the ability to provide understanding and context to clinical scenarios.”[4]

Incomplete information about abortion (especially in contexts like the UK where there are strict gestational limits on different methods) can push someone toward delays, dangerous alternatives, or services that are designed to mislead rather than help.

A 2024 study found that ChatGPT was actively spreading misinformation about self-managed medication abortion, overstating health risks in ways that directly contradict the evidence. Most seriously, it falsely implied that self-managing an abortion might cause or delay the detection of an ectopic pregnancy, which is not true. Delivered in a calm, authoritative voice, that kind of claim can be hard to question.[5]

The Campaign for Accountability's November 2025 report tested five of the biggest AI platforms (ChatGPT, Google AI Mode, Meta AI, Grok, and Perplexity) asking about options during a medication abortion. All five directed users to an anti-abortion helpline. In 50% of cases, it was the only number given. These helplines promote so-called “abortion pill reversal”: a practice the Royal College of Obstetricians and Gynecologists has said i

The presence of anti-abortion misinformation in AI outputs is the result of a deliberate strategy by the anti-choice movement.

Crisis Pregnancy Centres (CPCs) are organisations that present themselves as healthcare clinics while systematically discouraging abortion. They have spent years producing large volumes of well-structured content designed to appear in abortion-related search results. They were already doing this with traditional search engines. Now they have moved on to AI.

The tactic is called Generative Engine Optimisation (GEO): structuring content so that AI systems cite and repeat it. Unlike traditional Search Engine Optimisation (SEO) (which gets you a link in a results list), GEO gets the AI to speak your words directly, in its own voice, to someone asking a vulnerable question. These systems cannot distinguish between authority and volume. If an organisation produces enough well-formatted content, the AI treats it as credible, regardless of who funded it or what the evidence actually says.

Here is a real example illustrates how this plays out. When you search "can I change my mind after taking mifepristone" on Google, the AI Overview that appears at the top of the page cites three sources. Two of them are Crisis Pregnancy Centres, including Lifeline Pregnancy Help Clinic, which promotes abortion pill reversal. Only one is a reputable medical source. The AI doesn't flag the difference. It presents all three with equal authority. The CPC content even appears first.

There is another problem that gets less attention: privacy. The apparent anonymity of an AI conversation is not guaranteed. Most major platforms store conversation histories. In any legal context where abortion has been restricted or criminalised (and we are living through a period of significant rollback of reproductive rights globally) that data could potentially be sought by law enforcement or third parties. Someone who feels they are asking a question in private may be leaving a record they do not know about.[7]

Turning off location services, using incognito browsing, and knowing what platforms store are all worthwhile steps: but they're not enough. The burden of staying safe online shouldn't fall on individuals quietly adjusting their settings. Tech companies need to take ownership of how their products handle sensitive health data and safeguard their users.

Not all AI is bad. The qualities that make chatbots appealing, like accessibility, anonymity, 24/7 availability, a non-judgmental tone, can genuinely serve people who need abortion information, particularly in hostile or restrictive contexts. And pro-choice organisations have been building tools that do exactly this.

Chat with Charley, and Ally, developed by Women on Web, are designed to provide accurate, evidence-based, non-judgmental information about medication abortion. Women on Web has reported that ChatGPT referrals have quickly become a significant proportion of their traffic (+300% in one month). This means people are already using AI to reach safe services.[8] The question is which AI they end up with.

Individual actions you can take include protecting your own digital security, such as turning off location based services, using incognito mode and knowing what data AI platforms store are great ways to reduce risk. Continuing to educate yourself and others on how to recognise misinformation, understand how AI generates answers and why those answers can be wrong or biased, is another important step, as is encouraging people to use trusted, pro-choice AI tools such as Chat with Charley and Ally instead.

We also know that part of the reason anti-abortion appears to be winning the GEO war right now is they produce more content, and GEO cares more about quantity, not quality, or truth for that matter. Tactics such as publishing lots of well-structured FAQs and accurate clinical information online and reviewing trustworthy services all help push reliable information into the systems AI draws from.

Collectively, we need to be working better to connect reputable organisations in the reproductive health sector such as BPAS and MSI, with those doing important watchdog work in this space, such as Accountable Tech, Privacy International, and the Campaign for Accountability. We also urgently need more research on the scale, positives and potential harm of the use of AI in abortion care, as this helps strengthen the case for regulation, transparency, and change.

With the rise of the anti-rights movement in the UK and their attempt to undermine so many of our hard-won freedoms, staying vigilant to threats to our reproductive rights is more important than ever. At Amnesty Feminist’s, our My Body, My Rights campaign is doing just that, from tackling the rise of so-called ‘crisis-pregnancy centres’ in communities across the UK to holding monthly online Activist Assemblies that share important updates on reproductive justice issues in the UK and ways you can be part of the resistance.

Sign up to the next monthly assembly here or get in touch with us at [email protected].