AI Clinical Notes in Ontario and Quebec: What Therapists Need to Check
AI tools that listen to a session and draft your clinical note are no longer a novelty. Many Canadian therapists already use them, and many more are wondering whether they’re allowed to. The short answer: yes, but you’re responsible for how they’re used — and in 2026 that responsibility got a lot more specific.
In the last few months, Ontario’s privacy regulator published formal guidance on AI scribes, Ontario’s psychotherapy college spelled out what it expects from registrants using AI, and Quebec rolled out its own approved-vendor program under a privacy law with real financial teeth. If you’re an RP, psychologist, or social worker figuring out what you actually have to do before letting an AI draft your notes, this is the practical version.
We build Ignia, a practice management platform with AI notes, so we’ve had to learn these rules closely. This piece is useful whether or not you ever use our product — every regulatory claim below links to its primary source.
TL;DR: AI notes are not automatically prohibited, but the therapist or clinic remains responsible for consent, privacy, vendor due diligence, draft review, and province-specific requirements. In Ontario, IPC and CRPO guidance point toward explicit consent, data minimization, vendor review, and clinician accountability. In Quebec, confirm TGV certification, EFVP/privacy assessment requirements, and applicable professional-order guidance before using a tool.
What “AI clinical notes” actually means
AI scribe, ambient documentation, AI transcription — the labels differ, the mechanics don’t. The tool captures the session (from a recording or in real time), transcribes the audio, and drafts a structured note in your format, usually SOAP, DAP, or BIRP.
Be clear-eyed about one thing: the AI produces a draft. It decides nothing clinical and should never be treated as if it does. Ontario’s IPC, in its January 2026 news release, frames these tools as a way to reduce administrative burden, not as a clinical decision-maker. That distinction is the foundation everything else sits on.
What the rules now require in Ontario
Two Ontario bodies matter here, and they are looking at different things.
The Information and Privacy Commissioner (IPC) published AI Scribes: Key Considerations for the Health Sector on January 28, 2026, alongside a companion checklist. It is aimed at organizations that hold personal health information, and it reads as a practical guide for adopting these tools responsibly. A few points that affect a working clinician directly:
- Consent is explicit, and refusal cannot cost the client anything. A client who declines the AI scribe has to receive the same standard of care as one who agrees. Consent buried in a long intake form is not the standard regulators are moving toward; they want the client to actually understand that AI is involved and what happens to their information.
- Data minimization. Collect only what is reasonably necessary, and question whether full session audio needs to be retained at all once the note is written.
- Vendor due diligence. If you are using a third-party tool rather than building your own, you are expected to vet the vendor and have an agreement that holds them to obligations around intended use, lawful training data, accuracy, monitoring, and security.
The College of Registered Psychotherapists of Ontario (CRPO) approaches it from the professional-conduct side. Its position on AI in therapy is direct: if an RP uses AI to create or summarize notes, develop proposed treatment plans, or recommend AI products to clients, the RP is responsible for meeting all the usual CRPO requirements. That means providing services only within your competence, protecting privacy, obtaining consent, and reviewing AI outputs using your own judgment.
Read those two together and the picture is clear. The AI can draft. You remain accountable for consent, for privacy, and for the accuracy of anything that ends up in the record. The tool does not absorb any of your professional responsibility.
What changes in Quebec
Quebec is a different regime. If you practice there, or your clinic crosses both provinces, the differences are worth knowing.
Quebec’s Law 25 governs how private-sector organizations handle personal information, with duties that touch AI directly. When a decision relies on automated processing of personal information, the organization has to tell the person and, on request, explain the processing — and consent has to be clear and specific. It’s not enough that a client knows AI is in the room; they need to understand how their information will be processed. Law 25 also expects a privacy impact assessment in higher-risk situations, and the penalties aren’t symbolic. (Confirm exact section references and current penalty figures with your own counsel; specifics here come from legal commentary, not the statute directly.)
On the health side, Quebec has a more formal path. The province’s AI transcription program for health professionals sets two requirements before an AI transcription tool can be used in health and social services settings: TGV certification for the product, and a privacy impact assessment (EFVP) by the acquiring organization. Santé Québec may complete EFVPs for its own context, but other organizations still confirm their own obligations. The program also points clinicians to their professional orders and is explicit about transparency: clients must be told AI is being used — verbally at the start of a consultation or via a visible notice in the office — with the orders recommending explicit consent.
The takeaway is simple: don’t assume a tool that’s acceptable in Ontario is cleared for Quebec. Confirm its TGV status, the EFVP requirements, your professional-order guidance, and your clinic’s obligations first.
A note on AI “therapy” versus AI documentation
Two things get blurred here. Using AI to help with documentation is a workflow and privacy question. Using AI as a substitute for therapy is a different, more fraught issue, and CRPO has been pointed about it: AI tools aren’t regulated to provide therapy in Ontario, and treating them as if they were carries real risks — privacy concerns, unhealthy dependence, and the reinforcement of negative thinking. This post is about the first thing. The second deserves its own careful conversation, and it’s not one to hand to a chatbot.
A practical checklist before you adopt any tool
If you are evaluating an AI notes tool, these are the questions worth answering first. They apply to any vendor, ours included.
- Where does the data live? Canadian data residency can reduce cross-border privacy complexity, but it does not make a tool compliant by itself. Confirm where audio and transcripts are stored.
- What is the consent flow? You need a clean way to get explicit, understandable consent, and to offer a real alternative to clients who decline.
- Can you see and edit the draft before it is final? You are reviewing the AI’s output with your own judgment, so the workflow has to make that the default, not an afterthought.
- What happens to the audio? Ask whether recordings are retained, for how long, and whether you can turn retention off.
- Can the vendor produce documentation? Privacy impact assessments, security posture, audit logs, and a data processing agreement. If a vendor cannot show you these, that is your answer.
- Does it fit your province? Ontario and Quebec have different expectations. Confirm the tool actually meets the one you practice under.
What this looks like in practice
To make it concrete, here’s roughly how the workflow runs on Ignia — one example of building these requirements into the day-to-day rather than treating them as paperwork.
Consent is the part most tools treat as an afterthought, so it’s where Ignia starts. It lives in its own module inside a customizable intake form the client completes online, on their phone if they want, before they come in. If consent isn’t captured as a session approaches, the platform flags it to the clinician beforehand; if it still needs to happen in the moment, the therapist can capture and record it in a single click. The responsible step is built into the normal flow, not left to memory.
The rest follows. The session is transcribed, the platform drafts a note in your preferred format, and you review and edit that draft before it’s saved — so the final note is yours, not the model’s. The transcript is processed by an AI model provider to produce the draft, while the stored record stays in Canada, with an audit trail of who accessed what.
The goal is to make the responsible path the default. For any AI notes tool you evaluate, Ignia included, that’s the standard to look for: consent before use, clinician review before charting, clear retention settings, and documentation you can rely on. The tools that hold up are the ones that make the responsible path the easy one.
Frequently asked questions
Can therapists in Ontario use AI to write clinical notes? Yes. CRPO’s guidance allows registrants to use AI to help create or summarize notes, as long as the registrant works within their competence, protects privacy, obtains consent, and reviews the AI’s output with their own judgment. The responsibility stays with the clinician.
Do clients have to consent to an AI scribe? Yes, and consent should be explicit and understandable, not buried in a long form. A client who declines should still receive the same standard of care, so you need a real alternative. In Quebec, professionals must also inform clients that AI is being used.
Is a tool that is approved in Ontario automatically allowed in Quebec? No. Quebec has its own path, including TGV certification for the product and a privacy impact assessment (EFVP) by the acquiring organization. Confirm a tool’s Quebec status separately rather than assuming Ontario acceptance carries over.
Does the AI make the final clinical note? It should not. The note the AI produces is a draft. The clinician reviews and edits it before it becomes part of the record, and the clinician remains accountable for what is charted.
What should I check before adopting an AI notes tool? Where the data is stored, the consent flow and the alternative for clients who decline, whether you can edit the draft before it is saved, what happens to the audio, whether the vendor can produce privacy and security documentation, and whether the tool meets your province’s specific requirements.
Sources
- Information and Privacy Commissioner of Ontario, AI Scribes: Key Considerations for the Health Sector and the companion checklist (January 28, 2026).
- Information and Privacy Commissioner of Ontario, news release on the AI scribes guidance (January 28, 2026).
- College of Registered Psychotherapists of Ontario, AI Therapy.
- Gouvernement du Québec, AI transcription program for health professionals.
This article is general information, not legal or compliance advice. Privacy law and college expectations change, and they differ by province and profession, so confirm the current requirements that apply to your practice before you act on them.