For supervisors reviewing and signing off on session documentation
When your technicians or staff use AI Session Notes, the narratives you review have been drafted by the AI, then reviewed, edited, and attested to by the author. Your review is a separate check on top of that. This article covers what has changed in reviewing these notes and what to look for.
What Changes About Review
A note written hastily usually looks like one: short, unpolished, and easy to flag.
An AI-drafted note that's missing something still contains well-written narratives, correct terminology, complete sentences, and a professional tone.
This means the usual cues that tell you to slow down and look harder aren't there. Instead of reading for quality of writing, read for whether the narrative’s content is supported.
What to Check
Is Every Statement Supported?
Read the narrative against the rest of the note — the structured selections, the data, and the author's observations. You should be able to connect each claim to something recorded.
Watch particularly for statements that sound like clinical observation but have no source: why a behavior occurred, the effectiveness of an intervention, or how the session compared to previous ones.
Is Anything Missing?
This is harder to catch than a wrong statement, because nothing looks out of place. Compare the narrative against the data and the author's observations to confirm nothing that should have been included was omitted.
Are There Clinical Determinations the Author Shouldn't Be Making?
The ones to flag:
Progress or regression — "the client is progressing," "showed regression." These are determinations made from the data, and yours to make.
Generalization — claiming a skill has generalized is a clinical determination, not a session observation.
Mastery — unless the system recorded it.
Baseline comparisons — "more prompting than usual," "higher rates than previously." Comparisons require a baseline in the record.
Rationale for a clinical decision — if a narrative explains why a protocol change was appropriate, check that the reasoning came from a clinician and wasn't constructed by the AI.
If clinical determinations appear routinely rather than occasionally, the cause is likely the template's AI Prompt Instructions. Reach out to whoever manages your templates to address it.
Is Content Duplicated?
Check whether the narrative restates what the note already covers in other sections — the goals addressed, the data collection method, the recorded data, the interventions used. Duplication makes notes longer without adding anything, and in a payer review it can raise questions about whether the narrative adds clinical value beyond what the structured note already documents.
Would This Note Stand on Its Own?
Read it the way an external reviewer would, without the context you have. The note should show:
What was done
Why the service occurred
How the client responded
How the session related to the treatment plan goals
That's the standard CASP's documentation guidance sets for a session summary, and it's a reasonable test of whether a narrative is doing its job. If a narrative is fluent but doesn't answer those four questions, it isn't finished.
Does It Read Like Every Other Note?
Compare a few notes for the same client across a couple of weeks. If the narratives are close to interchangeable, something isn't working — either the authors aren't entering session-specific observations, or the prompt is producing formulaic responses. Records that look templated raise questions in an audit about whether services were individualized.
What the Review Is and Isn't
The author's attestation is theirs. When an author signs, they confirm they read and approved the narrative. That's their professional accountability and it doesn't transfer to you.
Your review is a separate checkpoint, and it's the one that catches what the author's own review missed. The two aren't redundant — the author knows what happened in the session; you know what the documentation needs to show.
A Few Notes on Feedback
When it comes to coaching authors on their AI-generated notes, a few things tend to work well:
When a narrative feels thin, it's often worth looking at the Note Author Prompt first — a sparse narrative usually reflects a sparse entry, and helping an author improve their input tends to produce better results than focusing on the output.
Specific feedback is more actionable than general feedback. Pointing to exactly what was missing — a behavior, a response, a change in the session — gives the author something concrete to work from.
When an author proactively flags something the AI got wrong, that's the review process working as intended. Treating it as a positive contribution encourages the rest of your team to do the same.
What to Escalate, and to Whom
Escalate to whoever owns your organization's AI templates when you see:
The same error across authors and clients
Narratives consistently claiming progress or generalization
Consistent duplication with other note sections
Narratives routinely too long or too repetitive
Content routinely dropped from a particular field
Scope problems — technician notes describing supervision, or the reverse
Correcting these issues note-by-note is time-consuming and doesn't address the root cause. Correcting the prompt or the field fixes it for everyone.
Tip: Make sure your organization has a named owner for prompt and template changes. If you don't know who that is, it's worth identifying them before your team's note volume grows.
Quick Reference
Use this checklist as a quick guide when reviewing any AI-assisted session note.
Check | Looking For |
Supported | Every statement traces to something on the note |
Complete | Nothing the data shows or the author entered is missing |
In Scope | No progress, regression, generalization, mastery, or baseline claims |
Not Duplicated | Narrative doesn't restate other note sections |
Stands Alone | Shows what, why, response, and relation to goals |
Individualized | Reads as this session, not any session |
Related Articles
Setting Up a Session Note Approval Workflow — How to configure approval workflows for session documentation
AI Session Note Narrative Generator Overview — Overview of the AI Session Notes feature and how it works
Using AI Session Notes as Clinicians — A guide for note authors using the generation tool
Setting Up the AI Session Note Narrative Generator — How to configure AI on your session note templates
Rolling Out AI Session Notes Across Your Org — How to stage and execute a rollout across your staff
Last Updated: 9/22/26 by Tatum Winslow