The first time I used Otter.ai seriously, I was on a discovery call with someone who spoke fast and thought out loud and kept interrupting herself to add a caveat. By the time we hung up I had four pages of transcript and a light headache, and I was genuinely impressed. The words were there. Not perfectly, but well enough.
That was about three years ago. I've had a more complicated relationship with it since.
Here's the honest version of what Otter.ai is, what it does for coaches specifically, and where I'd push back on the default "it's great!" consensus you find in most roundups.
What Otter.ai actually does
Otter is a transcription and notetaking service. You can record live through the app, run it alongside a Zoom or Google Meet call via their integration, or upload audio files for transcription after the fact. It produces a transcript, tags speakers where it can, and offers an AI summary layer on top.
The free plan gives you 300 minutes of transcription per month and some basic features. Paid plans (roughly $16 to $30 per month at current pricing) unlock more minutes, longer recording limits, and collaboration features that matter more for teams than for solo practices.
The collaboration and team angle is where Otter has put a lot of its development energy. If you're running a consulting firm or a team coaching practice, the shared workspace and team folder features make more sense. For a solo coach with 12 clients, you're mostly paying for features you'll never use.
The transcription accuracy question
This is where the honest answer is: it depends, and for coaching specifically, the "depends" matters.
Otter's accuracy on clean speech in a good audio environment is solid. Two native English speakers on a decent internet connection, quiet rooms, no significant background noise, you'll get something usable. Not perfect, but usable.
Where it degrades: accents, speech patterns that don't follow standard interview rhythms, multiple speakers who interrupt each other, anyone who thinks in long run-on sentences with lots of ums and restarted thoughts. Coaching conversations, real ones, tend to have a lot of the last two. Clients trail off. They circle back. They start saying something and then pivot. The moments of silence before someone says the thing they've been avoiding saying for forty minutes don't transcribe well.
Compared to Fathom, which I wrote about previously, Otter has more accuracy issues with conversational coaching speech. Not dramatically more, but enough that I've caught more transcript errors in Otter that would have changed meaning if I'd used them uncritically. Fathom benefits from newer underlying models. Otter has improved, but it started in an older generation of transcription technology and some of that shows.
What this means practically: if you're using Otter transcripts as raw material to feed into Margaret or another AI for processing, you need to spot-check before you process. Errors in the transcript produce errors downstream. A client says "I'm not sure I want this anymore" and it transcribes as "I'm not sure I want this more" and suddenly your AI-generated prep note is framing something differently than what was actually said. That kind of thing happens less than you'd think, but it happens.
I caught a more significant one last year. A client had said "I keep waiting for permission" and Otter rendered it as "I keep waiting for promotion." Two different conversations, two different sets of questions you'd want to bring into the next session. If I hadn't been reading the transcript carefully before loading it as context, I'd have prepared for the wrong thing entirely. Nothing catastrophic, but it made me more careful about treating the transcript as a first draft that needs a read, not a final record I can process without checking.
Privacy and client sessions
I want to say the same thing here that I said in the Fathom review: the standard terms-of-service response is not the same as thinking carefully about where your clients' content is living.
Otter.ai stores data on their servers. Their enterprise tier has more specific data handling options, but most coaches aren't paying enterprise rates. The standard plan means your recordings and transcripts are sitting in their infrastructure, subject to their policies and whatever changes those policies might go through.
Three years ago when I started using Otter, I wasn't thinking about this carefully enough. I was excited about not having to type up notes. That's not a good enough reason to make a decision with someone else's confidential information.
What I landed on: I don't run Otter (or Fathom, or any cloud transcription tool) on coaching sessions as a default. When I do use recording for a session, it's after explicit conversation with the client, documented consent, and a clear explanation of what tool I'm using and where the data goes. Most clients are fine with this once it's explained. The point is that they've chosen, not that I've quietly decided the risk is acceptable on their behalf.
The coaching-specific complication is that our conversations often reach material that a business meeting never would. Someone in a C-suite role might say things in a coaching session that they'd never say in a recorded work context. They've come to you specifically because there's no HR department listening. Treating that content with the same casualness as a project update call isn't consistent with the trust they're extending.
If you're in a jurisdiction with specific data handling requirements (GDPR, HIPAA-adjacent situations, anything involving healthcare or financial services clients), get proper advice on this before you start recording sessions with any cloud tool. The terms of service are not a compliance strategy.
Where Otter actually fits
Here's where I've settled after three years of testing, using, and scaling back: Otter is genuinely useful for a specific slice of coaching work. That slice is probably smaller than the "AI notetaker for coaches" marketing suggests.
Discovery calls and intake conversations. Before someone is a client, the privacy stakes are lower (though not zero). A 45-minute discovery call produces a lot of information I want to retain, and Otter handles that well. I can review the transcript before deciding to work with someone, run it through Margaret to pull out what they kept coming back to, and have a more complete picture than my real-time notes alone would give me.
Recordings the client wants. Occasionally a client asks to have a session recorded because they want to review it themselves. If we've agreed to that and I'm using Otter, they get a transcript too. Some clients find it useful. For this specific use case, the tool fits.
Audio files I'm uploading from my own devices. If I record a voice memo after a session, a reflection on what I noticed that I want to process later, Otter handles transcription of that fine. That's my content, not the client's, and the privacy calculus is simpler.
Peer supervision or case consultation sessions. With appropriate consent from any other coaches present, transcribing a peer supervision session can be useful. The notes from those conversations tend to be scattered across whoever was typing, and a full transcript helps.
The AI summary layer
Otter has added AI-generated summaries and Q&A features over the past couple of years, and I want to be honest that I haven't found them very useful.
The summaries have the same problem Fathom's do but somewhat worse: they're optimized for business meeting content. Action items, topics covered, decisions made. For a coaching conversation, that frame misses most of what matters. I've gotten summaries from substantive sessions that read like a particularly dull committee meeting. Technically accurate in the narrow sense, completely wrong in what they emphasize.
The Q&A feature, where you can ask the transcript questions, is more interesting in theory. In practice I find it easier to paste the relevant transcript section into Margaret and ask there, where I have more control over the framing and where I can layer in context that Otter doesn't have. The value proposition of keeping everything inside one tool is real, but not real enough to compensate for the quality difference.
What I actually use now
For most coaching work, I've shifted to a simpler workflow: voice memos on my phone immediately after sessions, uploaded to a transcription step, then processed. That keeps the recording off any call infrastructure and keeps raw session content off third-party servers. The transcript quality is similar, the privacy profile is better, and I'm not dependent on Otter's integrations staying functional through Zoom's API changes.
The voice memo approach took a few weeks to become habitual. I'd finish a session, close my laptop, and walk, which I do between most sessions anyway, talking to my phone like an idiot for about seven minutes, narrating what I noticed. The physical shift out of the room seems to help. By the time I get back to my desk I've already processed the session once verbally, and the transcript gives me a second pass in text. I usually run that through Margaret before the next session: "Here's what I captured right after. What should I be thinking about when we pick this up?"
I keep an Otter account for discovery calls and the occasional specific situation where it's the right tool. I wouldn't cancel it because it does those things fine and the free tier handles my actual usage. But it's not doing the job I imagined it would do when I first set it up.
The bottom line on Otter.ai for coaches
Use it for discovery calls. Use it for audio you're creating yourself. Be thoughtful before you run it on actual coaching sessions, not because the tool is bad but because the decision deserves more thought than most coaches give it.
The transcription is decent but not class-leading for conversational speech. The AI summary layer is weak for coaching-specific content. The privacy question requires an answer specific to your situation, not a default assumption that terms of service settles it.
If you're trying to decide between Otter and Fathom for a solo coaching practice: Fathom's transcription accuracy is modestly better, its privacy situation is roughly equivalent, and its highlight feature is genuinely useful in ways Otter's in-call tools aren't. Neither is a complete solution for session notes. Both are tools that do part of a job.
The thing I keep coming back to when coaches ask me about transcription tools is this: the goal isn't a perfect record of what was said. The goal is arriving at your next session with a clear, felt sense of where this person is and what's alive for them. Transcription can help with that, used carefully, as one input among several. It's not a replacement for the thinking you do in the hours between sessions, and it's not a substitute for the kind of attention that makes coaching worth paying for.
What you're trying to protect is the quality of your presence with the client. The tools are in service of that, or they're not worth the monthly charge.