How this was made: drafted by Miriad with AI assistance and checked against the sources listed at the end. Practice guidance, not legal or clinical advice; check your regulator's and data-protection rules.
TL;DR: Record notes are where most vets now use AI, and the profession's main worry is accuracy. Use AI to draft SOAP notes, never to sign them: the vet reviews every note against a short checklist (drugs, doses, weights, sides, owner statements), client data stays in assessed tools, and you measure time saved over two weeks before rolling it out.
It is the end of a twelve-consult afternoon, and one of your vets still has nine records to write up. She records consults with an AI scribe now, so the drafts are there. They are also long, a little too fluent, and in one case they mention a heart murmur she does not remember hearing. Does she sign them all, or check them all? That choice decides whether AI note-taking saves time safely or moves risk into your medical records. This article sets review rules for that choice. It is part of our guide to veterinary clinic marketing, because the time saved on notes is often the time a small team then has for clients.
How are vets using AI for record notes now?
Clinical documentation is the leading AI use in veterinary practice. A 2026 follow-up of the Digitail and AAHA survey found most respondents using AI, with AI scribes and SOAP notes the top use. A separate 2026 survey of general practices found medical records and SOAP notes the leading AI use there too.
The numbers, with their scope:
- 2026 follow-up (US and Canada, 1,730 respondents): 83.7% use AI; clinical documentation is the top use at 62.8% of users; client communication ranks second (dvm360, 2026).
- 2024 baseline (3,968 respondents): 39.2% used AI tools (Digitail, 2024).
- General practices, 2026: 48% use AI in some form; of those, 63% use it for medical records and SOAP notes and 38% for diagnostic assistance (Instinct Science via Santé log, 2026).
The surveys use different samples and questions, so do not add them up. What they share is the direction: if your vets do not use an AI scribe yet, some of the vets you hire next will expect one.
What are the risks of AI-drafted medical records?
The main risk is an error that looks right: a dose, a side or an examination finding the vet did not state. The second is data: client and patient details leaving your control. The third is quieter: a tool added without changing how notes are reviewed, which saves little time and adds new failure points.
The profession names the same risks. In 2026, reliability and accuracy was the top concern at 67.4%, then data security and privacy at 59% and lack of training at 47.1% (dvm360, 2026). In the 2024 baseline, reliability and accuracy led too, at 70.3%, ahead of data security and privacy at 53.9% (Digitail, 2024).
The workflow risk shows in the same 2026 follow-up: about 20% of AI users reported no measurable impact yet, and 65.2% of practices had added AI without meaningfully changing workflows (dvm360, 2026). A scribe that produces drafts nobody has a routine for reviewing does not save the time it promises.
A review checklist for AI-drafted SOAP notes
Review every AI-drafted note before signing, in the same order each time, so the check takes a minute rather than a re-read. Check the items that cause harm if wrong first, then the assessment and plan, then anything the draft added. A printed checklist next to the screen works better than good intentions.
The checklist:
- Patient: species, breed, sex, neuter status and weight match the record.
- Drugs: names, doses, routes, frequency and duration exactly as prescribed.
- Sides and sites: left and right, limb, tooth number, lesion location.
- Owner statements: anything under "Subjective" was said by the owner, not inferred.
- Findings: every examination finding was actually examined. Delete anything that was not.
- Assessment and plan: in the vet's words, including what was discussed and declined.
- Estimate and consent: if a written estimate or consent was given, the note says so.
- Sign-off: the vet's name and the time of review, not the time of the draft.
Example (illustrative, not a real case): A vet reviews 9 AI-drafted notes at the end of a session using the checklist. Each takes between 1 and 3 minutes, 17 minutes in total. In 3 notes she corrects something: a dose written in mg instead of mg/kg, a left forelimb recorded as right, and a "grade II murmur" she did not hear, which she deletes. She notes each correction in a review log: date, note, item number on the checklist, what changed. After two weeks the log shows which checklist items catch the most errors, and the practice moves those to the top.
How do you keep client data safe?
Keep identifiable client and patient data in tools you have assessed: ideally an AI scribe integrated with your practice management software, under a written agreement that says where recordings and transcripts are stored, who can access them and when they are deleted. Public chat tools with default settings are for templates, not records.
Questions to ask any vendor before a trial:
- Where are recordings and transcripts stored, and in which country?
- Are they used to train the vendor's models? Can that be switched off in writing?
- How long are recordings kept, and can you delete them yourself?
- Does the draft go straight into your PIMS, or is it copied by hand?
- What happens to the data if you stop using the service?
Tell clients, too. A short sign at the desk and a line in the new client welcome email cover most of it: what is recorded, why, and how to opt out.
How do you trial an AI scribe in a small practice?
Run a two-week trial with one vet, measure the time spent on notes before and during, and keep a correction log. Decide on numbers, not impressions: if the review takes nearly as long as writing the note did, the tool is not saving time for that vet, and that is useful to know before you pay for everyone.
A simple frame for the decision:
Minutes saved per session = (minutes writing notes before) − (minutes reviewing AI drafts + minutes fixing errors)
Example (illustrative, not a real case): Before the trial, the vet spends about 45 minutes after each session on 9 notes. During the trial, review takes 17 minutes and fixing takes 6, so 23 minutes. That is 22 minutes saved per session. With 8 sessions a week, about 3 hours a week. If the correction log shows the same error type every day, fix the template or the dictation habit before you scale up.
The time saved does not have to become more consults. In a short-staffed team it can go to the things that otherwise wait: calling back owners, writing the handout, answering reviews. Marketing a vet clinic when the team is already short-staffed sizes that routine, and cutting phone time spent explaining things to pet owners shows where to put the first hour. The veterinary clinic marketing guide puts both in a 90-day plan.
If you want your clinic's protocols, review rules and data rules written down once, so every AI tool and every new hire starts from the same facts, the Business Brain Starter Kit is built for that. Get notified when it launches.
FAQ
Is it safe for vets to use AI scribes for SOAP notes?
It can be, if the vet reviews and signs every note, the tool's data handling meets your data-protection obligations, and the client knows the consult is being recorded. The risk is not the drafting itself but an unchecked error entering the medical record, where it can follow the patient to the next vet, the insurer or a complaint.
Do we need to tell clients we use an AI scribe?
We recommend it in every case, and your regulator or data-protection law may require it. A short line at check-in and a sentence on your website are enough for most owners: what is recorded, why, who sees it and how long it is kept. Offer a way to opt out, and record the choice in the patient file.
What should a vet check in an AI-drafted note?
Check the things that cause harm if wrong: drug names and doses, weights, laterality, species and breed, dates, and anything the note attributes to the owner. Then check that the assessment and plan are the vet's own. Delete anything the AI added that was not said or examined, even if it sounds plausible.
Can we put patient records into ChatGPT?
Not into a public tool with default settings. Client names, contact details and identifiable records should only go into tools that your practice has assessed for data protection, ideally ones integrated with your practice management software and covered by a written agreement with the vendor. For general writing tasks, use templates with placeholders instead of real records.
Sources
- Artificial intelligence use among veterinary professionals more than doubles in 2 years — dvm360, 2026-10-06.
- 39.2% of veterinary professionals use AI tools in their practice — Digitail with AAHA, 2024-02-14.
- New Surveys from Instinct Science Reveal Surprising Shifts in Veterinary Practice — Instinct Science press release, hosted by Santé log, 2026-03-11.