Faculty and students from nursing and computer science use AI and University resources to create a free chatbot that protects user privacy.
When a patient said that she had used a chatbot to decide which birth control method to choose, Kandyce Brennan felt excited and alarmed.
“I immediately saw an opportunity but also a real risk to patient outcomes,” said the assistant professor in UNC’s School of Nursing.
That was in 2023. Brennan, also a midwife, knew the moment mattered. More than 19 million women in the U.S. live in “contraceptive deserts” with limited or no reproductive health services. People were increasingly turning to chatbots for sexual and reproductive health advice, but general-purpose chatbots can hallucinate, oversimplify or provide unsafe guidance.
“What if a chatbot could be designed for sexual and reproductive health, built with clinical rigor, empathy and safety at its core?” Brennan wondered.
The answer is SARHAchat, the Sexual and Reproductive Health Assistant.
The project began in 2024, when Brennan started working with Tianlong Chen, assistant professor in the UNC College of Arts and Sciences’ computer science department. The goal was to create a socially responsible chatbot to provide health education without replacing clinicians.
They gained momentum after receiving a $36,000 seed grant from UNC’s School of Data Science and Society later that year. Chen’s technical approach emphasized accuracy, safety and transparency. Brennan’s clinical approach focused on health equity and patient-centered care.
In August 2025, the Eshelman Innovation Venture Studio in the UNC Eshelman School of Pharmacy selected the project for a six-month development sprint. Second-year PharmD student Alexandra Kelly joined as venture builder, mentored by Amrita Bhowmick, entrepreneur-in-residence at Eshelman Innovation. Hong Liu, a junior who double-majors in computer science and economics, was technical venture builder, mentored by Xinyu Zhao, a doctoral student in Chen’s lab. Zhao led technical development, working with clinicians to ensure SARHAchat’s language, tone and structure reflected real clinical communication.
The sprint’s customer discovery phase included in-depth interviews with clinicians, public health departments, school counselors, executives and community members. The conversations helped identify pain points in sexual and reproductive health education, refine SARHAchat’s use cases and ensure the tool delivered meaningful value at minimal cost.
Chen’s team chose ChatGPT-4’s application programming interface as the foundation and built the system with University-provided tools like Red Hat open-source software. A clinical advisory board defined essential content, safety guardrails and the appropriate balance between specialized and general guidance. And they loaded in Centers for Disease Control and Prevention medical eligibility criteria and Food and Drug Administration recommendations.
After a national pilot with users who learned about SARHAchat from their care providers, the team refined sample conversations. They focused on factual accuracy, which is consistently around 98%. Brennan also rated conversations on an empathy scale, a process the technical team called “dialing in the temperature,” helping train SARHAchat to respond warmly and respectfully.
SARHAchat does not collect personally identifiable information or require accounts and does not store conversations for users. Web-based, it avoids app-store tracking. Developers view conversations only through encrypted systems with two-factor authentication.
“Chatbots have historically failed on privacy,” Brennan said. “We wanted users to trust that their information was protected.”
Chen said that having nursing and computer science experts working together reminded the team that a person is behind every data point. “Collaborating across disciplines meant we weren’t just building an AI tool. We were building a supportive companion that understands real-world barriers patients face,” he said.
Users can download their PDF chat summaries to keep or share with healthcare providers. Clinicians report that such summaries help patients start conversations and make visits more efficient.
“I hope that SARHAchat transforms AI for reproductive health by providing a hallucination-free space where clinical precision meets genuine empathy, ensuring every user receives safe, personalized and stigma-free guidance,” Chen said.
The team plans to expand content for adolescents, strengthen trust and reliability, and share lessons learned to help others build ethical, equitable health AI tools.
By Scott Jared, University Communications and Marketing