AI Scribe for Therapists: Ambient Listening vs. Dictation

What an ambient AI scribe actually feels like in session

What is an ambient scribe?

Ambient just means it runs in the background, picking everything up without you having to do anything during the session.

So in practice, you've opened your AI scribe tool on your device, hit record, and let it run while your session happens as it normally would. It picks up the conversation as you and your client talk, just like you always do. When the session ends, you stop the recording and the AI automatically generates a draft note for you to review. You can make any edits, sign off on it, and delete the recording right away if you'd like, or set it to auto-delete depending on your preferences.

The appeal with this method is that you can hit record, run your session, and have a draft note waiting by the time your client's out the door. No dictating afterward, no trying to remember what happened hours later. For therapists running back-to-back sessions all day, that can be a genuinely helpful thing.

When it works really well

Structured, protocol-based approaches like CBT, DBT, and trauma-focused work with clear stages can produce cleaner output. The transcript is clear, the themes are easy to identify, and the note usually comes out pretty close to what you'd write yourself.

The tradeoff

More relational, process-oriented sessions are a different story. The AI picks up what was said, but maybe not what mattered, like the silence before a disclosure, the moment the therapeutic alliance shifted, what you noticed as a clinician but didn't name out loud. That clinical layer still has to come from you, which is absolutely as it should be, but it's worth knowing going in.

One thing an ambient AI scribe doesn't always do

One thing to know about ambient AI scribes is that filtering varies quite a bit depending on the tool. Some are getting quite good at focusing on the clinical content and leaving out things like the small talk at the start or the offhand comment mid-session that probably doesn't belong in anyone's chart. Others pick up everything and leave the sorting to you. Either way, it's worth knowing that most tools let you adjust your prompt to guide how much detail gets included, so you can set it to capture what matters to your practice style and leave out what doesn't. That balance is yours to find, which honestly, is how it should be.

In summary, using an AI scribe as an ambient listener:

• Records your session live, in the background
• Records you and your client
• Client consent is required
• Tends to work best for structured, protocol-based work
• Main tradeoff: depending on the tool, it may capture everything including casual conversation, though most let you adjust your prompt to guide what level of detail gets included

What post-session dictation with an AI scribe actually feels like

What is post-session dictation?

Post-session dictation means your device stays in your pocket during the session. No app is open, nothing is recording, no AI is listening in the background. You meet with your client as you normally would.

Once the session ends and everything is still fresh, you open your AI scribe tool, hit record, and talk through what happened. Your synthesis, your clinical observations, anything significant that came up, what you want to focus on next time. Think of it as a verbal brain dump, just for your own notes. The AI takes that recording and spins up a structured therapy note for your review, working from your summary instead of a live transcript.

When it works really well

It often feels more like your note right from the start. Rather than trying to capture the whole conversation, you're speaking from your clinical point of view and guiding the AI as you go.

The tradeoff

It does require a small window between sessions. If your day doesn’t leave much space between clients, that window can be tough to hold onto. Some therapists find that dictating right after a session becomes a nice transition ritual.

A potential upside

With post-session dictation, only your voice is being recorded. That can be a meaningful difference when it comes to what you need to collect from clients, more on that just below.

In summary, using an AI scribe as a post-session dictation tool:

• You dictate to your scribe, summarizing the session after it ends
• Records only your voice
• Client consent requirements are simpler (though transparency is always good practice)
• Tends to work well for relational or more complex sessions
• Main tradeoff: requires about a 5-minute window between clients while things are still fresh

Using a medical AI scribe for therapy: what consent looks like

Questions around consent for AI use are one of the most common things that come up with AI notes for therapists, and it’s worth a closer look.

With ambient listening

With an ambient AI scribe, your client's voice is being recorded. Under HIPAA, that means you must obtain patient consent before recording. Though this doesn't have to happen at the start of every single session, many therapists cover it through their intake paperwork. Other therapists will do a quick verbal check-in at the start of each appointment. Consent is also an ongoing conversation rather than a one-time checkbox. Clients can ask you to pause or stop the recording at any point during a session, and that's completely okay. It's also worth knowing that some states require all-party consent for audio recordings, so your state law may add another layer to consider.

With post-session dictation

You're only recording yourself, your client isn't on the recording, so the consent picture is quite a bit simpler. If some of your clients might hesitate around being recorded (and in a therapy setting, that's a completely reasonable thing), dictation sidesteps that worry. Even if it’s not legally required, it’s still a good idea to give your clients a quick heads-up about your use of AI, so they feel informed and comfortable with their care. The APA's Ethics Code has guidance on privacy and record keeping standards for psychologists if you want a professional reference point on that.

💡 For a deeper look at how to approach the consent conversation, take a peek at our guide on using an AI scribe with clients, which walks through it step by step, including sample scripts.

What happens to the recording after the session?

And just to clarify, with most medical AI scribe tools, the recording itself doesn't become part of the clinical record, only the note does. The recording is used to generate the note, and then it's yours to manage. Most tools let you store it, delete it immediately, or set it to auto-delete. Once deleted, it's gone permanently, usually wiped from backups within a set amount of time as well. It's also worth asking your provider directly whether your clinical data is ever used to train their AI models. With reputable tools, the answer should be no, and that's something worth confirming before you get started.

How to choose the right AI scribe method for your practice

Neither method is universally better, it really does come down to how you work. A few questions worth sitting with before you decide which to try first:

How structured are your sessions?

Protocol-based approaches like CBT, DBT, and structured trauma work tend to produce cleaner ambient output. More relational, process-oriented work usually gives you more control with dictation, since you're filtering through your own clinical lens before the AI scribe ever sees it.

How do your transitions look between clients?

If you have natural buffers, dictation is pretty realistic to maintain. If you genuinely don't have a few minutes between appointments, ambient removes that need for extra time entirely.

How much of your own style do you want in your notes?

If it matters to you that your notes sound just like you, dictation tends to produce output that's already closer to your natural voice. Ambient gives you more raw material and details to work with, which can be great, but it does usually mean a bit more editing. Some AI scribes for therapists also let you create prompts, so you can guide what the AI focuses on and how your notes are structured, and refine that over time.

You don't have to pick just one

Many therapists who use AI scribes for a while don't end up committing to one method permanently. Different approaches tend to work better in different situations. Group sessions, couples sessions, or anything more complex often work better with dictation, both because the consent picture is simpler and because AI output from multi-person recordings tends to need a bit more editing to get just right. Sometimes the choice also comes down to the client. Even within the same session type, certain clients may feel more comfortable with one method over another.

A good starting point is to try one method with a session type you already know well, so you can actually tell how much work the output needs. That feedback loop will usually tell you pretty quickly whether ambient or dictation is the better fit for how you actually practice.

Reviewing your AI therapy notes: what to look for

Whatever method you go with, the AI therapy notes that come out are a draft, a starting point, not a finished product. They need your review before they go anywhere. And just to clarify, with most medical AI scribe tools, the recording itself doesn't become part of the clinical record, only the note does. The recording is used to generate the note and can be deleted right after, or set to auto-delete depending on your tool's settings.

The therapists who tend to get the most out of these tools treat the output a bit like notes from a first-year practicum student. Always reviewed with care before anything becomes part of the permanent record. Not because the tool can't be trusted, but because that review is still yours to do.

The most common things to watch for

• Over-elaboration: the note sounds more clinical or severe than the session actually was, or includes more detail than you'd typically document
• Phrasing that doesn't quite sound like you: the note is accurate, but it doesn't read the way you'd naturally write it
• Details that weren't discussed: occasionally the AI might add something that didn't actually come up in the session
• Clinical information that got garbled: medication names and clinical terminology in particular can get a little mangled, especially if they came up quickly or in a noisier environment

If you find yourself making the same edits after every session, that's usually a sign your prompt is worth revisiting. You can often get much closer to the output you want right from the start just by adjusting how you've set things up.

💡 New to AI scribes altogether? This getting-started guide for health practitioners is a good place to start before you pick a method.

Frequently asked questions

What's the best AI scribe for therapists in private practice?

The best AI scribe for therapists starts with choosing a tool that’s built for clinical work. That means strong privacy and security practices, clear documentation you can review, and a company that’s transparent about how your data is handled.

From there, most AI scribes support both ambient recording and dictation. What tends to matter more is how you use the tool in your workflow, and what feels more natural for your sessions and your clients.

Are AI scribes for therapists HIPAA compliant?

If you're practicing in the US, yes, the tool needs to be HIPAA compliant. Which means a signed Business Associate Agreement (BAA) with your provider and appropriate data security practices in place. It's worth asking your AI scribe provider directly about their data retention and training policies. For a fuller breakdown of what HIPAA compliance looks like for AI scribe tools, this article covers both HIPAA and PIPEDA requirements in straightforward language.

What's the difference between an ambient scribe and post-session dictation?

An ambient AI scribe records and transcribes your session in real time while it's happening. Post-session dictation means you record yourself summarizing the session after it ends, your voice only, no client recording involved. Both can generate structured AI therapy notes, but they feel quite different in practice and can come with different requirements around client consent.

Do I need client consent to use an AI scribe for therapy notes?

It depends on the method. With ambient listening, your client's voice is being recorded, which means you'll need informed consent under HIPAA before recording, though this can be covered through your intake paperwork rather than at the start of every session. It's also worth checking your state law, since some states require all-party consent for audio recordings. With post-session dictation, you're only recording yourself, which simplifies things quite a bit.

Either way, giving clients a heads-up about your use of AI is good practice regardless of what's legally required. It's also worth letting clients know that consent is ongoing. They can ask you to pause or stop a recording at any point, and that should always be an option. The APA's Ethics Code speaks to this directly if you want to check out some professional standards. For a full walkthrough of how to have that conversation, including scripts and what to do if a client says no, read our guide to getting client consent for AI scribe.

Can I use an AI scribe for couples or group therapy sessions?

Yes, though it's worth knowing that ambient listening tends to get more complicated with multiple people in the room. The AI has more voices to track, the output usually needs more editing to get right, and the consent picture gets a bit more involved since everyone in the session would need to be aware that recording is happening. Post-session dictation tends to be the simpler choice for couples and group work. Nothing is recorded during the session, you just talk through your notes afterward to your AI scribe tool.

What should I watch for when reviewing AI-generated therapy notes?

A few things are worth a quick scan before you file. Does the note sound like you, or has the AI gone a bit more formal or clinical than the session actually was? Are there any details in there that didn't actually come up? And if there were medications or specific clinical terms mentioned, it's worth checking that they were picked up correctly, since those tend to be where transcription gets a little wobbly.

The good news is that most of these issues are fixable at the prompt level. If you're catching the same kinds of things every time, that's usually a sign your setup just needs a tweak.

Oh, and if you're already on Jane, here's how to set up AI Scribe in your workflow, so you can try it for yourself.