Artificial intelligence (AI) tools used by doctors to create clinical notes are missing important non-verbal cues such as facial expressions and gestures, according to a recent study. These tools, known as ambient voice technology, use speech recognition and generative AI to automatically convert spoken words into structured medical records. The goal is to save time and improve the quality of doctor-patient interactions. However, a review of 27 studies by researchers at the University of Edinburgh found that these systems only listen to audio, missing out on physical gestures, facial expressions, and the tone of voice used by both doctors and patients. The researchers described these AI tools as having a "written language bias," meaning they treat spoken conversations as if they were written text. This approach prevents them from capturing important non-verbal information that a doctor might notice during a consultation, such as a patient’s emotional state or body language. The study also raised concerns that patients might behave differently when they know an AI is listening, possibly leading to less openness or trust in the interaction. Dr. Lucas Seuren, a research fellow at the University of Edinburgh’s Centre for Biomedicine, Self and Society, noted that while many doctors are excited about the potential of ambient AI scribes to reduce paperwork, the experiences and needs of patients are often overlooked. He warned that this could lead to important patient stories being missed, particularly among those who already face challenges in accessing healthcare. Published in the journal BMJ Digital Health and AI, the study highlights the need for careful adaptation of this technology to fit within different healthcare systems and settings. It also points out that the widespread adoption of ambient scribes—estimated to be used by up to 40% of GPs in the UK—has outpaced the research on their broader impacts. While most studies have focused on productivity and performance, the researchers argue that more attention is needed on how these tools affect the overall structure and delivery of healthcare. They emphasize the importance of building a strong evidence base before expanding their use further.