Upheal uses AI to turn your session recordings into structured clinical notes. This article explains what the AI does, where pre-written text is used instead, and what to watch out for when reviewing your notes.
How the AI builds your notes
After your session is transcribed, Upheal’s AI reviews the full transcript and extracts clinically relevant information into the appropriate note sections. It follows instruction sets designed by clinicians. These define what belongs in each section and how it should be written.
The AI does two main things:
Extracts information from the transcript: identifying what was said that’s relevant to each section of your template
Writes that information into your note: organizing it into the narrative paragraphs and bulleted lists your template specifies
The result is a note grounded in what was actually discussed during your session, not generated from general knowledge.
What the AI doesn’t write
There are two cases where note content is not produced by the AI:
Custom static text sections
If you’ve added a custom static text section to your template, Upheal displays that text exactly as you wrote it. The AI isn’t involved in generating this content at all.
Pre-written phrasing in MSE and Objective sections
Mental Status Exam (MSE) and Objective sections work a bit differently. For symptom areas where there’s no relevant information in the transcript, Upheal uses pre-written default phrases that reflect assumed-normal findings.
For example, if nothing in the session suggests otherwise, you might see:
“Behavior: Client was calm and cooperative, with good eye contact and an appropriate level of activity.”
In this case, the AI’s job is simply to recognize that there’s nothing relevant in the transcript for that symptom area, then output the pre-written phrase.
Note: If there is relevant information in the transcript for an MSE or Objective item, the AI uses that information to generate the content instead of the default phrase.
How diagnoses appear in your notes
Any diagnoses that appear in your notes are generated by the AI based on the transcript from that single session only. Upheal does not pull from previous sessions, your client’s profile, or an existing diagnosis list.
This means:
The AI infers diagnostic context from what was discussed in that session
Diagnoses are a reflection of the session’s content, not a clinical determination
Note: Always review diagnostic language in your notes before finalizing. You are responsible for ensuring clinical accuracy.
When notes might be inaccurate
Because the AI works from the transcript alone, it can occasionally get things wrong. Here are the most common situations to watch for:
Topics that weren’t discussed in the session
If something like a specific diagnosis, symptom, or treatment goal wasn’t meaningfully covered in the session, the AI may exaggerate or fabricate content to fill that section. This is something that can occur with any AI, not just in Upheal notes, and is sometimes called hallucination.
This is most common in sections tied to diagnoses. If diagnostic-related content wasn’t discussed, the AI may produce language that sounds plausible but doesn’t accurately reflect the session.
Unclear or ambiguous speech
If the transcript is unclear due to crosstalk or technical issues, the AI may misinterpret what was said.
What to do
Upheal notes are a starting point, not a finished product. Before signing or submitting any note, review it carefully against your recollection of the session. You can edit any section directly within Upheal.