When 58-year-old Maria Gutierrez found her husband collapsed on their kitchen floor, she didn’t hesitate—she dialed 911 and began CPR under the guidance of a calm, clear voice on the line. But what if that voice had been faster, more precise, and perfectly aligned with the latest medical guidelines? In San Diego, a new AI-powered tool called ChatCPR is proving it can be. Developed by researchers at UC San Diego in collaboration with the University of Pittsburgh School of Medicine, Johns Hopkins University, and others, ChatCPR recently scored 100% on guideline-based CPR checklists—outperforming real 911 dispatchers in analyzing and responding to recordings of actual emergency calls. With more than 350,000 Americans suffering out-of-hospital cardiac arrests each year and survival rates hovering around just 9%, every second and every compression counts.
The stakes couldn’t be higher. Only 2% of Americans are certified in CPR, meaning most bystanders who witness a cardiac arrest are thrust into a high-pressure situation with little training. They call 911 and wait for instructions—yet even trained dispatchers miss critical steps nearly 30% of the time, according to the study published in JAMA Internal Medicine. That’s where ChatCPR steps in. Unlike commercial AI models like ChatGPT or Gemini, which averaged 90% accuracy in simulated scenarios, ChatCPR was specifically designed for emergency resuscitation, delivering real-time, step-by-step guidance on chest compressions, breaths, and timing—tailored to the caller’s responses and the victim’s age.
In testing, researchers analyzed 12 real 911 calls where bystanders attempted CPR. When ChatCPR was run against the same audio, it consistently identified the need for compressions faster, corrected hand placement more effectively, and adhered flawlessly to American Heart Association protocols. It didn’t just match human performance—it surpassed it. “If AI is going to earn its place in medicine, it should start by helping people save the person right in front of them,” said Dr. John W. Ayers, head of AI at UC San Diego’s Altman Clinical and Translational Research Institute and co-author of the study.
Still, the team emphasizes this isn’t about replacing dispatchers. It’s about augmenting them. “The goal is to raise the floor of performance, not to replace trained professionals,” said Dr. Christopher M. Horvat of UPMC Children’s Hospital of Pittsburgh, who helped design the evaluation framework. ChatCPR is open-source, designed to be integrated into emergency response systems, potentially as a real-time assistant to dispatchers or even a direct interface for callers via smartphone apps.
As cardiac arrest remains one of the leading causes of death in the U.S., innovations like ChatCPR offer a lifeline—not just through technology, but through empowerment. The future of emergency response may not be human or machine, but a partnership where lives are saved not in spite of the chaos, but because help arrives—perfectly guided—within seconds.