Lung infections such as pneumonia remain among the leading causes of death worldwide, yet accurately diagnosing them—especially in critically ill patients—has long been a major challenge. Researchers at the University of California, San Francisco have developed a promising new approach that combines genetic information with generative artificial intelligence to improve diagnosis and reduce the overuse of antibiotics. The team analyzed electronic health records from critically ill patients using a generative AI model trained to detect subtle patterns in clinical data. By integrating this AI-driven analysis with a specific biomarker linked to lower respiratory tract infections—measured through genetic testing—they were able to identify pneumonia with significantly greater accuracy than traditional methods. This dual approach helps distinguish bacterial pneumonia from other conditions that mimic it, such as viral infections or non-infectious lung inflammation. As a result, doctors can more confidently determine when antibiotics are truly needed, reducing the risk of overprescription. Antibiotic overuse is a major public health concern, driving the rise of drug-resistant bacteria and increasing the risk of severe side effects. By enabling earlier and more precise diagnosis, this method could help preserve the effectiveness of existing antibiotics while improving patient outcomes. The study highlights the potential of combining advanced AI with molecular diagnostics to transform how we approach complex, high-stakes medical conditions. If validated in larger trials, this approach could become a standard tool in intensive care units, helping clinicians make faster, more accurate decisions in life-threatening situations.