Should We Be Using AI Scribes in the Doctor's Office? by Dr. Timothy Smith

Photo Source: Unsplash

Many who have visited the doctor recently has had to answer the question of whether they agree to the physician using an AI scribe to automatically capture notes for the clinical record of the patient’s visit. The AI scribe will “listen” to the patient-doctor conversation and capture in written words the details of the visit, including the doctor’s questions and the patient’s responses. Additionally, the AI scribe will capture the doctor's voice notes on details of the patient’s condition, current medications, and suggested treatment regimens and put these details in a reviewable report for inclusion in the patient’s electronic medical records. In a recent journal article titled “Use of Ambient AI Scribes to Reduce Administrative Burden and Professional Burnout,” the authors KD Olson of the Yale School of Medicine et al studied over 260 physicians and advanced practice practitioners and the effects of ambient AI scribes. The study concluded that the AI scribes across six different health care systems significantly reduced doctor burnout from 51.9% to 38.8%. (jamanetwork.com) The reduction of the doctor’s time interfacing with a computer to add patient notes arguably increases the time a doctor can look at and connect with their patient.

The ethereal AI scribe, however, does make mistakes which, if not corrected by the physician, will become part of the permanent record. For example, the Australian medical indemnity insurer and medical defense organization Avant listed some alarming errors produced by AI scribes such as “Worryingly, one AI interaction made up a whole neurological exam when I didn’t perform one. Probably based on a statistical prediction given one to two neuro findings I described.” In another case, the AI scribe missed a very important word, “not.” The AI scribe incorrectly captured, “has epilepsy vs does not have epilepsy.” (avant.org.au) The problem gets compounded by the fact that the voice recording does not get saved, leaving the clinical notes as the final word.

Apart from these anecdotal examples of ambient scribe errors, researchers from the UCLA School of Medicine endeavored to test the effects of AI scribes in clinical practice in a randomized trial. In the trial, the authors, including Paul J. Lukac, randomly assigned physicians to one of two AI scribes—DAX and Nabla-- or to the standard, non-AI note capture methods. (nih.gov) In this study titled “Ambient AI Scribes in Clinical Practice: A Randomized Trial,” published in 2025 in the New England Journal of Medicine, the researchers found that both the AI scribes helped to reduce physician fatigue, but many of the physicians in the trial did not think the scribe’s notes rose to the quality of the physicians’ own notes. Additionally, the scribes made mistakes, including speech recognition errors, including misunderstanding ‘not drinking alcohol’ as ‘now drinking alcohol’.” The AI scribes also misattributed parents’ medical history to the patient's current history and emphasized secondary issues above the primary diagnosis.  The physicians liked the reduced time needed to capture notes, but the study suggested more research remains before cementing the benefits of AI scribes as real.

AI scribes enter into a long line of physician assistants beginning with the development of clinical notes and checklists for paper patient records, the introduction of dictation services for doctors to record their notes for later transcription by trained medical transcriptionists, and the massive undertaking of electronic health records that digitized patient records for easy sharing. The introduction of electronic health records ushered in the era of the doctor typing on a PC while conducting a patient examination. The AI scribes have promised to reduce the distraction of note-taking and to reduce physician burnout due to administrative burdens, but the risk of AI hallucinations adding incorrect information to permanent health records can have negative implications if not corrected in the system. Note-taking has always been vulnerable to errors, but in the big picture, will the physician benefit from the reduced administrative burden? Or will there be other collateral losses such as loss of memory and case detail retention? A meta-analysis by the University of Edinburgh of 27 scientific articles examining AI scribes in medical settings suggested that manual note-taking helps in clinical reasoning and that some physicians reported that AI scribes reduce the physician's recall of individual patients. (ed.ac.uk) The introduction of AI scribes may reduce doctor burden, but it will introduce other issues, such as recall and a complex system for correcting errors in the electronic health care record system.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.