Plan C

To be able to gather top issues/problems as well as who is leading solutions, or who wants to be leading, and be able to accurately map shared values, stewardship and solutions response.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

At Plan C we're responding to the problem of AI answers on leading platforms (chatGPT, Claude, Perplexity, Google) containing misinformation and omissions, and leading a project to coordinate content improvements across high-authority abortion websites to improve answers from the source. These strategies are distinct but complementary to efforts around policy change, addressing bias and privacy/security. From the beginning 10y ago, Plan C has prioritized SEO and search visibility because of a) the censorship we've always experienced in sharing this health information online, and b) our understanding of the importance of an accurate and complete online search journey, as we introduced the concept of safe self-managed abortion in health systems and culture. We're also tracking research to understand if and how answers improve with on-page content and structure changes, building and using feedback loops with AI companies, forging relationships to address new issues as they arise, and sharing AI Optimization best practices and learnings with the reproductive health and access movement.

I think the emergent issue of accuracy, completeness and censorship of information on leading AI answer & summary platforms is captured by a few of the above points.

In the US, search is an extremely common way to learn about reproductive health options and make informed choices about care. Google searches used to contain lists of URLs, including manipulative SEO content and "fake clinic" listings. This required the searcher to parse through resources. Now people often seek information on platforms like ChatGPT and Claude, or see AI Overviews at the top of Google results. Google search summaries are seen by 2B people/month, and and 25% of ChatGPT's 900M users/week are asking health questions.  AI answers can be helpful, even better than traditional search - but right now, they can contain misinformation and omissions. As more people turn to AI for abortion information, answers are often incomplete or inaccurate – and this can be life-altering. A recent UPenn study found that only 73% of medication abortion answers were accurate. Only 56% were accurate on ChatGPT. And responses were disproportionately inaccurate in abortion-restrictive states. Medication abortion, a rapidly-growing option for tens of millions of people of reproductive age living in restricted states, showed the highest variability across models. What someone sees in their moment of search can be life-altering. Why it's happening? Models are synthesizing answers and interpreting legal risk and clinic access, based on what's found on the open internet. This can lead to errors and omissions in the options and resources platforms include. We're optimistic that this experience can be even better for abortion seekers than old methods of searching online, but it's not going to happen organically. AI could represent a complete breakthrough in access to accurate, helpful and human-centered information pathways. Potentially millions of people will have a clearer more complete understanding of abortion options in their area (UPenn estimates 1.3M pregnant people across the US are currently at risk of encountering misinformation on AI platforms annually). Abortion stigma will be lessened, by factual no-nonsense answers to questions are served by leading AI platforms treating this as any other information. Groups like Plan C are designing strategies to improve on-page information, opening conversations with experts and AI companies that center accuracy and trust of the tools, and researching answers to understand how things shift and change and respond accordingly. Other groups are taking other approaches to improve accuracy, shape training and reinforce policy, privacy and anti-bias goals. We see this as an "all hands on deck" moment to act, while these tools hit the market and protocol gets established.

I look forward to learning from the group on their desired role! Right now there is a major opportunity for leadership, clarity of goals, coming together to eliminate duplication of efforts and move forward in a coordinated and scalable way.

There are multiple AI x Repro working groups in the US that I can share more about if helpful. The Center for Democracy & Tech (CDT) is doing good work. I'm not sure how active this group is but found it interesting: https://peoplesaiaction.com/ Perhaps looking at speakers from RightsCon https://www.rightscon.org/ happy to talk more.

See earlier answers - at Plan C we're responding to the problem of AI misinfo/omission of info on leading platforms (chatGPT, Claude, Perplexity, Google), as it relates to reproductive health and abortion, by leading a project to coordinate content improvements across high-authority abortion websites, known to be regular AI citations and sources, and improve answers from the source. These strategies are distinct but complementary to efforts around policy change, addressing bias and privacy/security. We're also tracking research via partners regularly testing prompts on these platforms, to understand if and how answers improve with on-page content and structure changes. We're building and using feedback loops with AI companies, forging relationships to address new issues as they arise, and sharing AI Optimization best practices and learnings with the reproductive health and access movement. We're most interested in answers generated on leading answer and summary platforms, because this is where we can reach a broad swath of the public, adopting these mainstream tools (or continuing to use google but relying on the AI summary) and using the tools for their health answer capabilities. We know this is already very common: 25% of chatGPT's 900m/week prompts are health related, and a recent Rock+Health study found that up to 1/3 of all prompts on leading platforms are health related: http://rockhealth.com/insights/the-tortoise-and-the-hare-of-care-health-ai-insights-from-rock-healths-2025-consumer-adoption-survey/ We're specifically starting with content areas that are historically under-explained or left vague, because of the nature of the abortion issue (sensitivity/privacy/complexity) and adding nuance and depth to details that commonly shape repro health answers.