Walk into a GP's office and a therapy practice and ask both clinicians to show you their notes. You'd be looking at two completely different documents — different structures, different vocabulary, different purposes. Yet most AI scribe tools treat them the same.
Understanding why different specialties document differently is the first step to understanding why a one-size-fits-all approach to AI clinical notes doesn't work.
SOAP notes: the clinical workhorse
SOAP — Subjective, Objective, Assessment, Plan — is the dominant format in acute and primary care settings. It was designed to be scannable, defensible, and transferable between providers. If a patient arrives in an ER, an on-call physician should be able to read a SOAP note from a GP and understand the clinical picture in under a minute.
The structure maps well to how clinical encounters flow:
- Subjective — what the patient reports (chief complaint, HPI, symptoms)
- Objective — what the clinician observes and measures (vitals, exam findings, lab results)
- Assessment — the clinical interpretation (diagnosis, differential)
- Plan — what happens next (prescriptions, referrals, follow-up)
For an AI scribe, SOAP is relatively tractable. The conversational cues map onto the sections fairly directly. When a patient says "my knee has been hurting for three weeks," that goes in Subjective. When the clinician says "range of motion is restricted to 90 degrees," that goes in Objective.
Narrative notes: the therapy model
Therapy and psychology documentation is a different beast. A psychodynamic session note doesn't have a clean structure of symptoms and diagnoses. It's a narrative account of what happened in the room — the themes that emerged, the interventions used, the patient's response, and the clinician's formulation.
A CBT session note looks different again: it might document the homework reviewed, the belief being targeted, the cognitive techniques applied, and the patient's engagement level. A DBT note tracks skills practiced and chain analysis of behaviors.
Getting AI to generate a good therapy note requires the model to understand therapeutic modalities, not just transcribe what was said.
Specialty-specific formats
Beyond SOAP and narrative, there are formats unique to specific specialties:
- Psychiatric MSE — Mental Status Examination, including appearance, speech, mood, affect, thought process, cognition, insight, and judgment. Plus risk stratification.
- Dental charting — Structured around the oral cavity, quadrant by quadrant, with specific notation for surfaces, conditions, and planned procedures.
- Aesthetic consultation notes — Document baseline anatomy, patient goals, treatment plan, product and dosing, and post-procedure care instructions.
- Veterinary SOAP — Similar structure to human SOAP but with species, breed, age, weight, and owner information as key fields.
What this means for AI
A model that produces great GP notes and mediocre psychiatry notes isn't a clinical AI scribe — it's a GP AI scribe that will frustrate everyone who isn't a GP.
This is why we built Patien with specialty-specific note generation from the ground up. When you select your specialty, you're not just changing a template — you're changing how the model parses the conversation, what it considers clinically significant, and how it structures the output.
The result is notes that sound like you wrote them, not like a general-purpose transcription tool made a best guess.