Otter.ai - AI Transcription

Otter defined the automated meeting notes category and still does the core job well: live capture that runs without your attention and produces a searchable archive afterwards. The reason it now requires more careful evaluation than it once did has little to do with transcription quality. It is the unresolved legal question about consent, and buyers should understand that before rolling it out.

What It Actually Does

A bot joins scheduled meetings through calendar integration and transcribes in real time. Afterwards you get a transcript, summary, and action items, all searchable across your meeting history.

Collaboration features sit on top. Transcripts can be edited live, shared, and commented on, with custom vocabulary improving recognition of names and domain terms over time.

Key Features

• Real-time transcription — Live captions during the call, which distinguishes it from tools that only produce output afterwards.
• Searchable archive — Reviewers rate the search across past meetings as one of the strongest parts of the product.
• Custom vocabulary — Teach it names, acronyms, and jargon to reduce recurring transcription errors.
• Speaker separation — Attempts to distinguish voices, though this draws criticism in practice.
• Meeting platform integration — Works across the major conferencing tools via calendar sync.
• Collaborative editing — Shared transcripts with comments and highlights for team follow-up.

Where It Fits Best

English-language teams wanting a reliable meeting archive are the core audience. Individuals with a few meetings weekly can operate within the free tier. Students and researchers use it heavily for lectures and interviews.

It fits poorly for multilingual organisations. Reviewers repeatedly note language support limited to a small number of languages, which is far behind competitors covering dozens.

Tradeoffs Worth Knowing

The consent question is the most important thing here. The company is the named defendant in a consolidated federal class action in the Northern District of California concerning how its bot records participants who did not agree to be recorded. The allegations are unproven and the case is unresolved. No court has ruled the recording unlawful. What it means practically is that if your calls involve external participants in jurisdictions requiring all-party consent, that risk sits with you rather than the vendor. Verify the current status directly, since litigation moves.

Related complaints appear in user reviews independently of the litigation. People describe the bot joining meetings uninvited and transcripts being emailed automatically to all participants including external parties. Whatever the legal outcome, those behaviours create awkward situations.

Accuracy is good in clean conditions and degrades predictably. Reviewers cite figures around 90 to 95 percent with clear audio, dropping with accents, overlapping speakers, and background noise. Speaker identification draws specific criticism, with generic labels rather than names being a common complaint.

Costs escalate for teams. One analysis puts a ten-person deployment in the region of several hundred dollars monthly, with usage caps on higher tiers that heavy users reportedly exhaust.

Practical Notes

Check the data training terms at your tier. Reporting indicates de-identified conversation data may be usable for model training on lower plans, which matters for confidential discussions.

Everything runs in the cloud, so there is no offline capability and connectivity problems interrupt capture directly.

Many published reviews here are written by competing meeting assistants, and several disclose that openly. Their emphasis on the consent litigation is fair reporting, but the recommended alternative is always the publisher’s own product.

Pricing reported for 2026 places the main paid tier near $17 monthly, or roughly half that on annual billing, with a free tier offering a few hundred minutes.

Announce the recorder verbally at the start of external calls. Do not rely on the bot appearing in the participant list.

Review the automatic sharing settings before your first client meeting. Transcripts reaching external parties unprompted is a recoverable mistake only once.

Custom vocabulary is worth setting up properly. Names, products, and acronyms are where errors cluster.

How It Compares

Against newer meeting assistants, Otter trails on language coverage and on privacy posture, while retaining strong search and a mature real-time experience. Against tools that capture audio locally without a visible bot, it carries more consent exposure by design. If you are already using it successfully with internal meetings, there is no urgent reason to move. If you are choosing now, particularly for external calls, evaluate the alternatives properly first.

What to Verify Before Choosing Otter.ai

• Current status of the consent litigation before a team rollout
• Recording consent requirements in every jurisdiction you operate in
• Whether your working languages are supported
• Data training terms at the tier you intend to buy
• Who receives transcripts automatically after each meeting
• Minute caps against your actual meeting volume
• Accuracy tested on your own accents and audio conditions
• Retention and deletion controls for stored recordings
• Export options if you later move to another tool