Otter.ai Review (2026): Good Software, and a Judge Just Named the Catch
Short answer - should you buy Otter?
- Buy it if you live in a CRM. The Otter Meeting Agent, the SDR and Recruiting agents, and 20+ integrations (Salesforce, HubSpot, Slack, Jira, Notion) are the actual product, and nobody local touches that.
- Buy it if you need Windows, iOS, Android, and Mac in one team. Otter runs everywhere; most of the private alternatives, mine included, are Mac-only.
- Skip it if the meeting contains someone else's confidential material - client work, candidates, patients, legal strategy - and you'd struggle to explain to that person where the audio went.
- Skip it if per-seat pricing is the problem. Otter Business is $30/user/month monthly, $19.99/user/month annual (otter.ai/pricing, 25 August 2026).
- Skip it if you specifically wanted "no bot." That reason expired - see below.
Bias, up front and unmissable: I make Humla, a Mac notetaker that competes with Otter, and I would love it if you downloaded it. So read everything below as coming from an interested party. What I can offer against that is receipts - every price, quote, and date here comes from Otter's own live pages or from published legal coverage, checked on 25 August 2026, with links, and you're very welcome to go read them and decide I've read them wrong.
First, kill the premise everybody still repeats
Half the "Otter alternatives" posts on the internet - including, embarrassingly, an earlier one of mine that at least had the decency to correct itself - are still fighting the 2024 version of Otter, where a bot called "Otter.ai" barged into your call and sat in the participant list like an uninvited cousin.
That's no longer the only mode. Otter shipped a desktop app (announced October 2025) for Mac and Windows that records the meeting audio directly, bot-free, alongside the classic Notetaker bot. So if your one reason for shopping around was the bot, that reason is gone, and you can stop reading the posts that say otherwise.
What hasn't changed is everything downstream of the microphone. Bot or no bot, the audio goes to Otter's cloud, gets transcribed there, gets stored there, and gets used there. Which turns out to be the whole review.
What it costs
Straight off otter.ai/pricing, 25 August 2026:
| Plan | Monthly | Annual | What you actually get |
|---|---|---|---|
| Basic | Free | Free | 300 monthly transcription minutes |
| Pro | $16.99 / user / mo | $8.33 / user / mo | 1,200 in-app recording minutes, up to 90 mins/meeting, 10 monthly file imports |
| Business | $30 / user / mo | $19.99 / user / mo | Unlimited meetings + in-app recordings, up to 4 hours/meeting, unlimited imports, join 3 concurrent meetings |
| Enterprise | Custom | Custom | Everything in Business, plus unlimited custom AI workflows |
Two honest notes on that table. One, the annual prices are real prices, not a trick - if you're going to use it for a year, use the annual column. Two, the free tier is a demo, not a plan: 300 minutes is roughly five hour-long meetings a month, and then you're done until the calendar flips.
What Otter is actually good at
I want to be fair here, because there's a version of this post that's just a competitor sneering, and that post is boring and nobody links to it.
The transcription is good. Not "good for free" - good. English conversational audio with crosstalk is a hard problem and Otter has been solving it commercially longer than most of the companies now shipping notetakers have existed.
The integration catalogue is the moat. Salesforce, HubSpot, Slack, Notion, Jira, Asana, Google Docs, Dropbox, 20+ of them. If your meeting is supposed to end with a CRM record and a ticket, Otter closes that loop and Humla flatly does not - my app pushes action items precisely nowhere.
It runs everywhere. Web, Chrome extension, iOS, Android, Mac, Windows. That sounds mundane until you're the one telling a colleague on a ThinkPad that your lovely local Mac app has nothing for them.
The agents are a real product, not a demo. The Otter Meeting Agent, the SDR Agent that handles live demos and books meetings, the Recruiting Agent - I have opinions about whether an AI should be running your discovery call, but this is clearly built by people who talked to sales teams, at length, and shipped what they asked for.
And on security specifically, Otter's privacy and security page is better than most of its competitors': SOC 2 Type 2, GDPR compliance, "HIPAA requirements in how we collect, store, and process health-related information", server-side AES-256, deleted conversations purged from trash after 30 days, and a commitment that "Before audio recordings are ever reviewed manually by a human (Otter personnel or its third parties), we require explicit consent from customers."
That last sentence is doing something specific, though. Read it again. It's about humans reviewing your recordings.
The Fourth Attendee
There's a participant in your Otter meetings who isn't in the participant list, doesn't show up in the bot-free desktop mode either, and doesn't need consent because the sentence above was carefully about people.
From Otter's privacy policy, effective 16 June 2026, listing what it does with your data: "training our proprietary AI technology on de-identified audio recordings and on transcriptions (which may contain Personal Information)."
The privacy and security page describes the safety net: "Otter uses a proprietary method to de-identify user data before training our models so that an individual user cannot be identified."
Okay. So the shape of the deal is: your meeting is de-identified by a method Otter doesn't describe, and then it teaches a commercial model. Which is not a scandal - it is, roughly, the deal on offer from most cloud AI products, and Otter at least writes it down in plain language instead of burying it under "to improve our services." Credit where it's due. (Compare OpenAI's version and Deepgram's, which I've read end to end and which take genuinely different positions.)
Here's my problem with it, and it isn't a legal one. Nobody in the meeting agreed to it except you. Your client didn't. The candidate you interviewed didn't. The person who called you crying about their contractor didn't. You clicked accept on a privacy policy on their behalf, and the thing they said out loud went off to be de-identified by an unspecified process and turned into model weights. I'm not describing a crime. I'm describing a conversation you'd rather not have with them.
I looked for an off switch. I could not find a documented opt-out for model training on either the privacy policy or the privacy and security page - Otter documents opt-outs for AI Chat and for Speaker Learning as separate help-centre articles, and various third-party reviews assert that a general training opt-out exists in account settings, but I'm not going to pass along a claim I couldn't verify at the source. If you're an Otter admin, go look, and I'd genuinely like to be corrected.
Then a judge read the same clause
This is the part that makes an August 2026 Otter review different from a July 2026 one.
Four class actions filed in 2025 - the lead one is Brewer v. Otter.ai - were consolidated into In re Otter.AI Privacy Litigation, No. 5:25-cv-06911-EKL, in the Northern District of California, before Judge Eumi K. Lee. The claim, broadly: OtterPilot recorded and transcribed people who never consented, in violation of the federal Wiretap Act, the California Invasion of Privacy Act, and a pile of state statutes.
Otter moved to dismiss. Its arguments were reasonable ones: the bot is visible in the participant list, so there's notice; the account holder, not Otter, is responsible for getting consent under the terms of service; the plaintiffs didn't plausibly allege they'd disclosed anything private.
On 13 August 2026, the court granted that motion in part and denied it in part. Surviving: the CIPA claims, the federal wiretap claims, Illinois BIPA, unjust enrichment, and unfair competition. Dismissed: the CFAA and California computer-fraud claims, the Washington Privacy Act claim, and most of the common-law privacy claims (National Law Review, 17 August 2026; VoIP Review, 19 August 2026).
But the reasoning is the thing. The court rejected the "we're just the host's tool" defence, and the reason it gave was that Otter "independently collects, retains, and uses communications for its own commercial purposes" - including, explicitly, using conversations to improve its products and its machine-learning models. That's what makes it a third party under CIPA section 631 rather than an extension of the person who pressed record.
Read those two sentences next to each other. The privacy policy's training clause and the judge's third-party reasoning are describing the same fact. Otter wrote it down honestly, and the honest version turned out to be the legally interesting one.
Now the disclaimers, and I mean them, because this is exactly where a competitor's blog post gets tempted to overreach:
- Nothing has been proven. Surviving a motion to dismiss means the allegations are plausible enough to proceed. It is not a finding that Otter broke any law. Otter denies the claims.
- No class has been certified, there's no settlement, and there's no claim form. Anyone telling you otherwise is running an ad.
- I read the coverage, not the order. The docket is on CourtListener and I couldn't retrieve it directly; the details above come from two independent legal-press write-ups that agree with each other on the date, the judge, and which claims survived. Treat it accordingly.
- This isn't an Otter-only problem. The reasoning applies to any notetaker that collects your conversations and uses them for its own purposes, which is most of them. Otter is first because Otter is biggest.
If you want the boring version of the underlying question - who's allowed to record whom, and where consent actually bites - I wrote that post separately, and two-party consent has its own glossary entry.
So who should actually buy it
Sales, RevOps, and recruiting teams. Genuinely. If the meeting's output is a CRM field, buy Otter, buy Business, and don't overthink it. My app cannot do this and pretending otherwise would be a lie.
Mixed-platform teams. Windows and Android exist. Otter is on them.
Anyone whose meetings are internal. Your own standup does not need on-device processing. It needs to be searchable and it needs to stop eating your Tuesday.
Not for meetings where somebody else's confidence is the material. Client calls, interviews, therapy, legal, medical, HR investigations. Not because Otter is careless - the SOC 2 and the HIPAA language and the encryption are all real - but because the training clause and the ruling above are things you'd have to explain, and "it was de-identified by a proprietary method" is a rough sentence to say to a client.
Where Humla is worse
Since I've been ranking a competitor, here's the bias disclosure with teeth. Humla is free and MIT-licensed, records with no bot, transcribes on your Mac or through your own API key, and never uploads anything unless you turn on Cloud sync at $5/user/month - which is a sixth of Otter Business's monthly price. That's the pitch.
The costs of that pitch, plainly:
- Mac only. macOS 13+, Apple Silicon recommended. No Windows, no iOS, no Android, no web. If one person on your team is on a PC, Humla is not your team's tool.
- No integrations. Zero. No Salesforce, no HubSpot, no Jira. There is exactly one integration and it's a read-only MCP server for agent tools. If you want one-click-to-Jira, buy the cloud tool.
- No agents booking your demos. Otter's SDR agent has no counterpart here and won't.
- Much smaller. Otter has years of scale, a support org, and an enterprise sales motion. I have a GitHub issues tab.
- Local Whisper costs you disk and patience. Model downloads are hundreds of megabytes to gigabytes, and the first run is slow.
| Humla | Otter | |
|---|---|---|
| Bot joins the call | Never | Optional - bot or bot-free desktop app |
| Where audio is processed | Your Mac, or your own API key | Otter's cloud |
| Trains a vendor model on your meetings | No | Yes, on de-identified audio + transcripts (privacy policy, 16 June 2026) |
| Platforms | macOS 13+ (Apple Silicon recommended) | Mac, Windows, iOS, Android, web, Chrome |
| Integrations | One, read-only MCP server | 20+ (Salesforce, HubSpot, Slack, Jira, Notion…) |
| Price | Free app; $5/user/mo optional Cloud | Free tier (300 min/mo); $16.99-$30/user/mo |
| Source available | MIT, on GitHub | No |
Longer side-by-side on Humla vs Otter, and the switching guide is at Otter alternatives. If you're an agency or a consultancy specifically, /for/agencies and /for/consultants are the pages that argue this properly.
The verdict
Otter in August 2026 is a mature, well-built, well-secured product that is very good at the job it has decided to do, which is no longer "take notes" but "run the commercial half of your meeting." If that's your job too, it's an easy recommendation and I won't be offended.
But the review can't end on the software, because the software isn't what changed this month. What changed is that the quiet part - your meetings teach the model - stopped being a paragraph in a policy nobody reads and became the load-bearing fact in a federal judge's reasoning. Otter, to its credit, never hid it. It's right there in plain English, effective 16 June 2026.
The Fourth Attendee was always in the room. It's just that now there's a court order describing what it's doing there.
Every Otter price, quote, and policy date above was checked against Otter's own live pages on 25 August 2026. Vendor pricing and policies drift - go and re-check them before you sign anything. The litigation details come from published legal coverage dated 17 and 19 August 2026, not from the court's order itself.