Using AI Session Notes as Clinicians

For behavior technicians, BCBAs, and therapists writing session notes

It does: read everything on the note you're working on — your checkboxes, your data, and the observations you type — and turn that into a written narrative in professional clinical language.

It doesn't: know your client. It can't see prior sessions, the treatment plan, or anything outside the note in front of you. It doesn't make clinical decisions, and it can't tell whether something you entered is accurate.

You own what you sign. When using the AI Session Note Narrative Generator, the process is simple: review, edit, and sign. The AI draft is a starting point — you are responsible for reviewing the output and making any necessary edits. Every statement in the final narrative is yours.

Here are some best practices to implement when using the AI Narrative Generator.

Complete your checkboxes, selections, and data before you click Generate. The AI builds the narrative from what's on the note, so a half-finished note produces a vague or incomplete narrative — and you'll spend longer fixing it than you would have spent filling in the fields first.

The top box in the purple section is where you add what happened. This is the part that makes the difference between a generic draft and a useful one.

Write it however is fastest. Fragments, abbreviations, and bullet points are all fine. You're not writing the note — you're giving the AI the raw material. It handles the sentences.

Focus on the context of what happened during the session:

Skip what's already on the note. You don't need to re-type trial counts, percentages, program names, or anything you already checked a box for. The AI can see all of it.

Don't ask for formatting. Your organization has already configured things like professional wording, narrative length, objective language, and past tense in the template. You don't need to request these — asking again just adds length to your entry without improving the draft.

Here's what a useful entry looks like. Short, specific, unpolished:

Your entry:

That's under a minute of typing. The AI turns it into a narrative covering how the client presented, how they responded to each target, the behaviors and your responses, and how the session progressed — pulling the trial data and program names from the note itself.

Reminder: The AI-generated narrative will also incorporate information pulled from other sections of the session note — not just what you typed in the observation box.

Here are two examples of prompts that may produce underwhelming results.

Example: Too Little Input

The AI has almost nothing to work with. The draft will be thin, generic, and you'll end up writing it yourself anyway.

Example: Too Much Input

These prompt details are already part of the note template and do not need to be restated in your AI prompt. These instructions take longer to type than the observations, and they crowd out information that would have improved the draft.

Read the whole thing. Professional-sounding writing is easy to skim past — that's exactly why it needs a careful read.

Check for these five things:

The bottom box is fully editable. Fixing a sentence directly is faster than regenerating, and you keep everything else that was already correct.

If your narrative requires extensive correction, you may need to regenerate the AI's output. To do this, make your edits directly in the bottom box, then click Generate again to produce a revised narrative. Your edits are sent back to the AI as context, so it will try to carry your corrections into the new version. This is useful when something is wrong throughout — fix it in one place, regenerate, and the rest usually follows.

There is no undo. Regenerating overwrites the narrative in the bottom box, and the previous version cannot be recovered. You'll get a confirmation prompt first. If you've made edits you don't want to lose, copy the text somewhere before regenerating.

If the draft keeps coming out wrong in the same way, that's almost always a template or prompt issue — not something you're doing wrong. Tell your supervisor or whoever manages your templates. This is a fixable issue, and they can't fix it if nobody reports it.

When you sign a note where you used AI generation, you'll see a checkbox: I reviewed the AI-generated text and confirm this note is accurate.

That's a professional attestation. Checking it means you personally read the narrative and confirmed it's accurate. Don't check it on a narrative you haven't read.

Once signed, the note exports with just the final narrative. The observations box and the AI interface don't appear in the printed or exported record.

Bring it up with your supervisor if any of the following occur:

These aren't complaints, and noticing them is the review process working. Patterns in the drafts usually mean something needs adjusting in the template or the prompt, and your supervisor or template owner is the person who can do that. Your feedback is what keeps the tool working well for your whole team.

Last Updated: 9/22/26 by Tatum Winslow