Can an AI Chatbot Help Women Recognize Heart Disease?

Heart disease remains the leading cause of death among women in the United States, yet public awareness of its risks and symptoms has experienced notable declines over recent decades. To address this persistent public health challenge, researchers at the University of California, San Francisco, have developed an interactive tool known as HeartBot. Led by Yoshimi Fukuoka, a nurse scientist and professor at the UC San Francisco School of Nursing, the project explores whether a specialized conversational application can improve women’s understanding of cardiovascular health and encourage prompt medical action.

The Growing Gap in Women’s Heart Health Awareness

According to data from the American Heart Association cited in the project background, cardiovascular conditions affect more than 60 million women across the nation, with the highest prevalence observed among non-Hispanic Black women. Despite extensive public health outreach over the years, surveys indicate that overall awareness has waned. Statistics show that the percentage of women who recognize that heart disease is the primary cause of mortality dropped from 65 percent in 2009 down to 44 percent in 2019, with the sharpest decreases seen among younger women, Hispanic women, and Black women.

Traditional media campaigns have fallen short of engaging these populations effectively. Fukuoka and an interdisciplinary team consisting of cardiovascular nursing scientists, linguists, communication specialists, computer scientists, biostatisticians, and programmers set out to test an innovative, technology-driven approach. Originally launched in 2001 and subsequently enhanced into an app-based format named HeartBot II in 2025, the tool is designed to deliver tailored educational content directly to users’ mobile devices.

How the HeartBot Application Operates

Unlike standard generative artificial intelligence systems that draw from the entire internet, HeartBot’s underlying architecture relies on a curated knowledge bank created by Fukuoka’s research team. This structure aims to minimize the risk of producing inaccurate or misleading medical guidance. The developers compiled approximately 600 potential user questions and formulated research-based responses, taking care to analyze the language, cadence, and rhythm of how women converse about cardiovascular issues.

The current iteration utilizes a large language model application to facilitate natural, personalized dialogues. During interactions, the app can recall previous conversations—such as an earlier mention of a family member who experienced a cardiac event—and guide users through learning modules covering:

• The fundamental mechanics of heart disease
• Common and atypical heart attack symptoms
• Recognized risk factors
• Effective prevention strategies

Researchers noted that the app translates complex medical terminology into everyday language. For instance, initial feedback revealed that terms like “risk factors” were not always well understood, prompting the team to refine the vocabulary to make the platform more accessible and engaging for everyday users.

Evaluating Safety and Efficacy Through Clinical Trials

While the tool is currently available on the Apple App Store and Google Play Store for study participants, its broader clinical efficacy and safety have not yet been definitively established. To measure its impact, Fukuoka is conducting a six-month randomized controlled trial involving 201 female patients at UCSF Health.

In this trial, half of the participants receive immediate access to the HeartBot II educational modules, while the control group receives access at a later date. Investigators are tracking changes in the participants’ retention of heart disease knowledge over time, specifically focusing on whether the application improves their ability to identify subtle heart attack symptoms—such as shortness of breath, nausea, vomiting, and pain radiating to the back or jaw—which are frequently overlooked or mistaken for other conditions.

Fukuoka’s long-term vision involves expanding testing to thousands of women over extended periods and potentially integrating physical access points, such as QR codes placed in clinical settings and public transit spaces. However, researchers emphasize that rigorous trial data are required first to determine whether conversational artificial intelligence can reliably influence care-seeking behaviors, such as calling emergency services when critical minutes matter.