A plastic-surgery scribe should draft what the clinician and patient say. It should not quietly turn a photograph, implant record, or procedure observation into text it never received. Test those boundaries separately. The useful product is the one that keeps each fact tied to its source and gives the surgeon a clean review queue.
One visit can leave four different records
Picture a 3:40 p.m. consult. The patient explains a prior procedure. The surgeon describes today’s findings aloud. Staff capture photographs in an approved system. A plan is discussed, but nothing has been performed yet.
Those facts should not collapse into one undifferentiated paragraph. Use this source card during a trial:
| Record lane | Source | Safe scribe job | Human checkpoint |
|---|---|---|---|
| Patient narrative | Conversation | Draft the reported history and goals | Confirm attribution and context |
| Clinician finding | Spoken observation | Draft only what the clinician states | Confirm laterality, site, and wording |
| Photograph or image | Approved image system | No inference from the image | Review image, label, consent, and access separately |
| Procedure record | Clinician and source record | Draft the dictated account | Confirm what actually occurred before signing |
The table is intentionally strict. A polished sentence with no traceable source is still a bad sentence.
The photograph needs its own privacy path
HHS explains that health information becomes protected health information when it identifies a person, or when there is a reasonable basis to believe it can identify the person. Its HIPAA de-identification guidance lists full-face photographs and comparable images among the Safe Harbor identifiers that must be removed (HHS de-identification guidance).
That does not mean every clinical photograph follows one universal rule. It means a practice cannot treat its photo library as decorative media. Privacy and legal owners should define why an image is captured, where it is stored, who can view it, how consent or authorization is handled, and how the practice handles patient access and amendment requests (HHS medical records guidance). Record retention and disclosure still follow applicable law and practice policy.
An AI-scribe pilot should not improvise this policy. Keep image capture outside the scribe test unless the vendor contract and the practice’s approved workflow expressly include it.
Run a three-encounter falsification test
Do not start with the easiest visit. Try to prove the product wrong.
First, run a consultation where the patient describes old surgery but the clinician has not verified the details. The draft should preserve that attribution. Next, use a follow-up with an image in the chart that is never described aloud. The draft should not invent an image finding. Last, dictate a procedure account with exact site and laterality, then change one detail during review and check whether the edit history remains visible in the destination system.
Score each encounter on five items:
- Every statement has a clear speaker or source.
- Unspoken image findings stay out.
- Planned care does not become completed care.
- Laterality and dates survive review.
- The clinician can correct the draft before it becomes part of the record.
The federal Health IT Playbook recommends selecting technology against the practice’s own needs, using common demonstration scenarios and a testing plan (ASTP/ONC Health IT Playbook). A three-encounter falsification test turns that advice into something a specialty practice can run in one afternoon.
Put the scribe beside the specialty system
Patient Square is an AI clinical platform. Its AI Medical Scribe drafts a structured SOAP note, ICD-10 suggestions, and a prescription draft for clinician review. Practice Copilot can work alongside an existing EHR or use the bundled Patient Square EHR, named AI Copilot EHR on the pricing page, from the Copilot tier. The annual US rates are $79 per physician per month for Assist and $119 for Copilot (Patient Square pricing).
Patient Square does not claim photo management, image interpretation, implant tracking, procedure imaging, or plastic-surgery-specific clinical modules. If one of those capabilities decides the purchase, keep the specialty system. Evaluate the scribe on spoken documentation and review burden instead.
Our opinion: the strongest plastic-surgery pilot is deliberately boring. It proves that the draft knows less than the surgeon, not more. Use the AI scribe security checklist for the vendor review and the small-practice implementation guide for a limited rollout.