In the middle of the night, when worry is loud and no one seems available, it can feel hopeless to try to figure out what to do. The internet used to be where people went for answers, but now the first stop is often not a search engine at all. It is a chatbot. ChatGPT, Claude, and other large language models have become a kind of new form of healthcare provider-patient conversation. We now have the kind of stats that make this clear: an estimated one in seven people have used AI for a health-related question instead of seeing a general practitioner, and one in ten have used AI for mental health support. But within that enormous group of people is a smaller, deeply vulnerable group for whom the stakes are nearly impossible to overstate: people looking for accurate information about abortion. Since the Supreme Court overturned Roe v. Wade in June 2022 in the Dobbs v. Jackson Women’s Health Organization decision, the legal landscape for abortion in the U.S. has become a maze of state-by-state trigger bans, restrictions, exceptions, and confusing new rules. The internet was already full of abortion misinformation and propaganda before AI arrived, but chatbots have complicated things further. Abortion rights advocates have watched this happen and are now being cautious. There is a growing movement to fight for the accuracy of AI information just as fiercely as people fight for access to care itself. As Kelly Baden, vice president of policy and communications at the Guttmacher Institute, recently put it, it can feel like you need a law degree just to understand what is legal on any given day, and AI has made that even more challenging.
The people trying to preserve accurate information are not just outside observers, they are adapting their whole approach to make room for how people now get information. The Guttmacher Institute, for example, has a well-known interactive policy map that people all over the country use to understand abortion laws in all 50 states. But now, when people do not go directly to the map, the map’s information is being summarized and reshaped by AI systems. Kelly Baden said that their work has changed such that they have not been writing only for human readers, but also for algorithms. They need to think about how the information they share is going to be translated by tools that did not exist or were not on their radar only a few years ago. This is a new kind of fight. Organizations built to help people directly, like Plan C Pills and I Need an A, are tracking their traffic and seeing real referrals from AI assistants. Amy Merrill, co-founder of Plan C Pills, which helps people find resources for at-home, self-managed abortion with medication abortion pills, said that her site sees spikes in traffic when ChatGPT sends interested people there. Rebecca Nall, founder of I Need an A, which is a free site that connects people to providers, telehealth services, financial assistance, and travel logistics, said they too have seen significant growth in referrals from AI assistants over the past couple years. These organizations are grateful for the extra support, but also cautious. There is a concern that feels like a mirage of privacy and convenience. As at the one-night isolation, it can feel like you are dialoguing with your own healthcare provider. But you are actually speaking with their company and giving them your personal data. Merrill warns that this is a very real concern within the reproductive rights movement, in a moment when digital privacy is essential to safety.
When it comes to accuracy, the evidence is alarming. International journalists at Open Democracy released a July 2024 investigation that found AI chatbots were nudging pregnant women to anti-abortion resources rather than offering unbiased medical information. A group of healthcare providers and researchers concluded that only 65% of ChatGPT 3.5 was responding to abortion-related questions in an “acceptable” way. That leaves a million percent of two-thirds of the time. Another 2024 study found ChatGPT actively spreading misinformation about self-managed medication abortion. The researchers raised the fear that chatbot’s tendency to perpetuate health misinformation and stigma surrounding self-managed medicinal abortion misuse is a threat to public health and reproductive autonomy. This is not an abstract problem about algorithms; it has real impact on patient care. When a person gets confusing, conflicting, or wrong information, they may delay care. As Baden says, any day that a person is delayed in getting the care they need because they got confusing information, or wrong information, could put you out of reach of abortion care entirely. If the illegal window is short, someone may find that they lost their chance without realizing it.
The real-world consequences are even visible if you think about someone living in Louisiana, a state that prohibits in-person abortion care at all stages of pregnancy without exceptions. That person can still travel to Florida or Georgia in the six weeks of pregnancy for care. But if an AI model gives them bad information and they wait, missing the six-week cutoff, the distance and financial cost increase. They may now be out of reach of getting abortion that they need. This is the heartbreaking dynamic that happens out of the LATEST in these chats. One of the biggest issues is that AI often fails to see the full picture of options. Nall says AI has asked not just about in-person clinics, but also to consider the full range of choices. Even in states with strict restrictions, people still can self-manage an abortion at home by ordering medication abortion pills online. This is not just a future idea; it is already happening. In Missouri, for instance, a state with severe restrictions and limited exceptions for medical emergencies, the number of medication abortion has quadrupled since 2023. Despite the legal situation, pills are arriving in every single state. Nall says that a lot of people are self-managing their care in every state, even hostile ones. AI, on the other hand, may simply say abortion is banned and stop there. Instead, it sometimes points people to crisis pregnancy centers, facilities that typically do not provide abortion care and often seek to deter people from having one. This can be particularly cruel if someone is turning to technology for unbiased information. The technology is not neutral; it is reflecting the biases and confusion of the internet it was trained on.
Advocates have found a way to fight back in the limited, important way: SEO. Because AI tools are currently influenced by what they read, those demanding accurate information can change the landscape by changing the language of the internet. Nall says that she has been changing the language on her own site and encouraging other reproductive rights activists to changing the language on their websites. The hope is to influence what chatbots are learning. For Nana, it means that websites for abortion providers and reproductive healthcare should be in words that make clear what is actually available, and she tries to make sure to include what is not still true. The issue also comes down to media coverage. Journalists are being encouraged to think about how shared content affects LLMs. One key problem is the phrase “abortion ban.” If articles say, “abortion is banned in Texas,” some AI systems might interpret it to mean that there is no abortion access in any form, anywhere. That worsens a possibly restrictive state true but also simplifies a more nuanced situation in a way that has vital consequences. Merrill stresses that it is important to communicate that in-person care is restricted, but also that people are still accessing abortion pills by mail. You don’t need to sugarcoat the law. You have to be careful not to unintentionally push someone toward a solution that doesn’t actually exist because an AI system didn’t understand the difference between “abortion is tightly restricted” and “no help exists in any form.” That is the kind of nuance that can cause people to stop looking. It is why Nallius and others have made it a mission to make sure the word “ban” doesn’t become a blanket that covers up what is still possible in our communities.
There is some reason to hope. Nall is reluctant to use the word “optimistic,” but she says she has seen improvements in AI responses since they started shifting language on websites. When activists make changes and see a difference, it means that the AI is not lost in a fixed system; it is demonstrating that it is missing in some way. The recommendations for someone navigating this right now are simple. If you are trying to find out what you can do, go to authorized human-centered directories. Websites like PlanCPills.org, INeedAnA.com, and the Repro Legal Help Line are excellent places to start. They offer direct lines to actual people who can provide real answers. Rebecca Nall says she personally recommends skipping Google, skipping ChatGPT, and going straight to INeedAnA.com. That may sound light, but in a world where confusing information can become a barrier to healthcare, it is particularly important. The internet is full of noise, and AI is still in its early years. But what abortion rights advocates around the country know is this: you can’t allow the technology to be the one deciding what is true and possible. The urgency of individuals is undeniable. With every word you write on the internet, with every article you share, with every conversation you have, we are writing the material that future AI models will learn from. For a woman looking for an abortion, the difference between “there is no way” and “there is another path” can be. If the world of information is to be humanized, it has to be built with compassion, nuance, and real research. We don’t need to pray for AI to be perfect. We need to remember that the best information, the way it has always been, is the accurate, human-centered guidance that comes from people who understand how much is on the line.