If you’re a speech-language pathologist, the number that follows you around isn’t your caseload — it’s the paperwork behind it. ASHA’s 2024 Schools Survey found school-based SLPs spend a mean of 6 hours a week on documentation, and 76% of them rank “large amount of paperwork” as their single biggest challenge — the top-ranked complaint in every setting surveyed. It’s the reason 27% of school SLPs in that same survey said they’re considering leaving the profession.
So when SLPs on Reddit talk about AI, the most popular use isn’t anything clinical. It’s turning a wall of clipped session notes into a progress summary a parent can actually read — fast. One SLP described their workflow plainly: feed it “a blurb of my session and some data,” and it returns a clean SOAP note “and also a parent friendly version too.” Here’s how to do that safely, including the one step you can never skip.
Read this first: the privacy rule that protects your license
Before any prompt, the non-negotiable part. Never paste a student’s or patient’s identifiable information into ChatGPT.
ASHA’s guidance is blunt about it: “Assume that all generative AI tools are not HIPAA compliant, unless vetted by your facility’s privacy, legal, or compliance team,” and “even if your version of that tool is a paid account, do not share clients’ or staff members’ PII.” The free and Plus tiers of ChatGPT have no Business Associate Agreement, which means they aren’t eligible to handle protected health information at all.
And “de-identify” means more than crossing out a name. Under HIPAA’s Safe Harbor standard there are 18 categories of identifiers to strip — names, dates (anything more specific than a year), locations smaller than a state, record numbers, and “any other unique identifying… characteristic.” ASHA adds a sharp clinical caveat: if a child has a rare disorder or is identifiable by their treatment protocol, even naming the disorder can reveal who they are.
The practical version for a busy SLP: describe the skill, the data, and the progress — never the child. “A 3rd-grade student” not “Marcus in Ms. Lee’s class.” Strip names, dates of birth, school, and anything specific enough to point back to one kid. Then, and only then, open ChatGPT.
The 5-minute workflow
Here’s the move that turns a 20–30 minute writing task into a 5-minute review.
Step 1 — Write your clinical shorthand as usual. Your real notes, with real names, stay in your district’s secure system (your FERPA-compliant IEP platform or HIPAA-compliant EHR). You’re only going to copy the clinical content, stripped of identifiers.
Step 2 — Paste a de-identified version into ChatGPT with this prompt:
You are helping a school speech-language pathologist write a warm,
parent-friendly progress update. I will give you de-identified clinical
data. Do NOT add any names, dates, or invented details — only rephrase
what I provide.
Rewrite the notes below into a short progress summary a parent with no
medical background can understand. Keep it encouraging and specific,
explain any term in plain language, and end with one simple thing the
family can practice at home.
De-identified data:
"A 3rd-grade student increased correct /r/ production from 40% to 72%
accuracy in structured word-level tasks over 8 sessions. Still
inconsistent at the sentence level. Responds well to visual cues."
Step 3 — Review every line for clinical accuracy. This is the part that keeps it ethical and correct. Read the draft as the professional you are: does it match what actually happened? Did it overstate progress? Cut anything that drifts.
Step 4 — Add the personal details back manually, inside your secure system — the child’s name, the specific dates, anything identifying. ChatGPT never sees them.
That’s it. The AI supplies the structure and the plain-language tone; you supply the data and the judgment. As one 14-year veteran put it: “You supply the data, AI supplies the structure.”
What this means for you
- If you’re a school SLP drowning in progress reports: This is your highest-value first use. Start with one report this week. The summary-writing is the part AI genuinely speeds up.
- If you’re in private practice or a clinic: The same workflow applies, but your bar is HIPAA, not FERPA — be even stricter about de-identification, and ask whether your employer has an approved, BAA-backed tool for anything involving real patient data.
- If your district gave you a tool like Magic School: Useful, but verify its compliance status before trusting it with anything identifiable — SLPs in the community have flagged that some district tools still aren’t HIPAA compliant.
- If you’ve never pasted a prompt into ChatGPT: Start here, with de-identified data and a low-stakes summary. It’s the gentlest possible on-ramp, and the privacy habit you build now protects everything you do later.
What this can’t do
- It can’t touch identifiable data — ever. That’s not a limitation to work around; it’s the rule that keeps you compliant. If a task requires real names inside the AI, it’s the wrong task for free ChatGPT.
- It can’t replace your clinical read. ChatGPT will confidently produce a summary that sounds right and is subtly wrong. One SLP reported a colleague who turned in all-ChatGPT progress reports that were “all wrong and contradictory.” You are the accuracy check.
- It can’t write the actual report or IEP for you. Even enthusiastic AI-using SLPs draw this line — great for summaries and materials, off-limits for the eval report itself.
- It can’t invent good data. Garbage in, confident garbage out. The summary is only as honest as the de-identified notes you give it.
The bottom line
Used right, ChatGPT turns the most-hated part of your week — translating clinical shorthand into something a parent understands — into a quick review job. The whole thing rests on one habit: de-identify first, every time. Describe the skill and the progress, never the child; review every line; add the names back in your secure system.
If you want to build that privacy-first habit properly — so it’s automatic, not something you have to think about under deadline — our AI for Clinical Notes: A HIPAA-Safe Workflow course walks through exactly where the line is and how to stay on the safe side of it. And if you’re a school-based clinician, AI for Teachers covers the same de-identified-drafting skills for the school context.
Wondering whether all this AI talk means your job is at risk? We answer that honestly in Is AI Coming for Speech Therapy Jobs?
Sources
- ASHA 2024 Schools Survey: SLP Caseload and Workload Characteristics — ASHA
- Generative Artificial Intelligence for Clinicians in Audiology and Speech-Language Pathology — ASHA
- Methods for De-identification of PHI (the 18 identifiers) — U.S. Dept. of Health & Human Services
- Perceptions of Artificial Intelligence and ChatGPT by Speech-Language Pathologists — American Journal of Speech-Language Pathology (Nov 2024)
- Empowering School-Based SLPs with AI Tools — Speech Therapy PD