03/03/2026
"AI, Medical Records, and You"
At TC Rehab, one of the cornerstones of our client/patient services are our comprehensive medical records and exam reports, which contain all of the expert findings, functional assessments, and full body range of motion and muscle mass measurements taken of your pet.
In the past year, we have begun using an AI software in clinic that will transcribe our verbal findings during an exam. These time-saving programs aid our veterinarians by transcribing the findings as they happen and organizing it for transfer to our medical templates, which become the records you receive.
It is imperative that every professional who uses AI scribe software carefully proofreads and edits these transcriptions before formulating the medical records, because AI can (and does) make mistakes. Worryingly, AI will sometimes leap to its own (incorrect) conclusions in generated summaries, despite recording everything we say verbatim.
A quick example: if one of our doctors verbally reports a patient is leaning off of a limb, AI may conclude (and therefore report in its generated summary) that the patient is limping, even though the doctor did not state that! That is a troubling medical discrepancy.
We advise extreme caution in using AI to interpret scientific data or asking AI a medical question. AI is a data crawler, rather in the vein of Google, and will gather data from high-traffic online forums and message boards such as Reddit, which is not a reputable source for qualified medical advice.
Scientific sites tend to be low traffic and therefore are not picked up as easily by AI or Google. Even more detrimentally, scientific articles and research studies are often behind paywalls, so programs like ChatGPT do not have access to the latest, credible scientific information.
Dr Julia has been working with the specialty college (ACVSMR) on their continuing education committee, and in discussing with her colleagues, there have been incidents of scientific reports submitted to peer review (specialist review) before publication that have been noted to have bogus references likely gathered via AI assistants/programs like ChatGPT, Google’s Gemini, Microsoft CoPilot, etc.
As an experiment, Dr Julia put one of her scientific publications (Muscle Action During Skijoring) into both ChatGPT and our own AI veterinary software and asked each to make as summary. Both made different mistakes in interpreting the data and turned a phrase that said “may be associated but further research is needed” into “is associated” – a huge difference!
AI is like many things of the “Age of Intelligence” – a good tool, but a bad master. Proceed with caution when using or trusting it with medical data.
What are your thoughts?
