Can you use AI for NDIS progress notes? (2026 guide)

Yes, you can use AI to help write NDIS progress notes. How you use it matters enormously for compliance and for participant privacy. This guide covers the rules support workers need to follow before pasting anything into ChatGPT, Claude.ai, or any other AI tool.

The short version: AI is fine as a structuring tool. AI is not fine as a ghostwriter. And there are specific things you must never do with participant data, no matter which tool you use.

What the NDIS Commission says about AI

The NDIS Commission has not banned AI tools. It has not endorsed them either. The relevant rules sit in the existing NDIS Practice Standards and the NDIS Code of Conduct, both of which apply regardless of how you produce a document.

Three things flow from those rules.

The worker is always responsible for the content of the note. If your note says you spent two hours on community access and you actually spent forty minutes, the fact that an AI helped phrase it does not protect you. You are attesting to the content.

A note written entirely by AI from nothing is a compliance breach. The note is supposed to record what you personally observed and delivered. If you did not observe it, AI cannot invent it for you, and pretending otherwise is dishonest documentation.

AHPRA made its position clear in 2024: practitioners are accountable for every word in their records, regardless of how it got there. If an auditor or regulator asks you to explain a sentence in a note and you cannot, the fact that AI wrote it is not a defence. The practical lesson is simple: AI is a tool for expressing your observations, not a tool for inventing them.

The privacy rule you cannot break

This is the single most important section. Read it twice.

Do not paste participant names, NDIS numbers, disability information, health details, home addresses, or any identifying information into a public AI tool. That includes ChatGPT, Google Gemini, Claude.ai, Microsoft Copilot, and the chat interfaces of any other large language model that you have not signed a specific data-processing agreement with.

The reason is the Australian Privacy Principles under the Privacy Act 1988. Participant information is sensitive personal information. When you paste it into a public AI tool, that data is processed on servers that may be outside Australia, may be retained in logs, and may in some cases be used in training data. None of that is consistent with the obligations a support worker has to a participant.

What this looks like in practice:

Tools that are built specifically for NDIS documentation, like Billa, handle this differently. Participant data does not leave a compliant Australian-hosted environment, and you do not need to type identifying information into any public chat box. That is the meaningful difference between using a generic AI tool and a purpose-built one.

What good AI-assisted note writing looks like

Five steps.

1. Observe and capture during or right after the shift. Bullet points on your phone. A voice memo you transcribe. Shorthand that only you can read. The point is the observations are yours.

2. Strip identifying details before you use a public AI tool. Replace names with "the participant", remove NDIS numbers, replace specific addresses with "their home" or "the local supermarket".

3. Prompt the AI to structure, not to invent. A good prompt sounds like: "Turn these shift notes into a compliant NDIS progress note. Keep all facts exactly as I have written them. Add structure and professional language. Do not add anything not in my notes."

4. Read the output carefully. Add back details the AI missed. Remove anything that did not happen. Anything the AI invented to "round out" the note has to go. You are the author of every sentence that ends up in the participant's file.

5. Add the participant's name and details in your actual notes system. Not in the AI chat box.

Here is what that workflow looks like in practice.

That transformation is what AI is good at, and it is exactly what Billa does.

What AI cannot do for your notes

Three honest limits.

AI cannot observe what actually happened. If you did not write it down, AI cannot retrieve it. The quality of the output is bounded by the quality of your input.

AI does not know your participant's specific NDIS goals unless you tell it. Either include the goal in your prompt or add the goal reference in by hand at the end.

AI-generated text has a flat, generic quality that experienced auditors notice. Your specific observations and your voice are what make a note feel like a real record. The best AI-assisted notes still sound like you because you fed in real, specific observations.

The best framing: AI is spell-check for your observations, not a ghostwriter for your shift.

Compliance summary

A quick reference you can come back to:

✓ AI can help structure and improve your notes
✓ You must still personally observe and record what happened
✓ You are responsible for every word in your final note
✗ Never paste participant names or identifying details into public AI tools
✗ AI cannot write a note from nothing, that is a compliance breach
✗ Do not submit an AI-structured note without reading it carefully first

Get the prompt template

The downloadable template is a set of de-identification rules and ready-to-use prompts for the five most common note types: community access, personal care and daily living, skill development, social participation, and incidents or concerns. Use these prompts with any AI tool to stay on the right side of the privacy rules and produce a compliant note.

Next steps

If you have not already, the NDIS progress note template is the right starting point for the structure of a compliant note. It pairs naturally with the prompts in the AI guide.

Workers and small providers who are heading into an audit should also read the NDIS audit checklist which walks through what auditors examine beyond the notes themselves.