Overview
Otter joins meetings as a visible participant to transcribe, summarize and extract action items. Its AI Chat searches across your meeting history with citations. The trade to weigh: a bot appears in the attendee list, accuracy degrades on heavy accents and jargon, and auto-joining has gotten it blocked by some IT teams.
A bot joins your meeting, visibly
Otter works by sending a participant into your calls. Connect your calendar, and OtterPilot automatically joins Zoom, Google Meet or Microsoft Teams meetings, appearing in the attendee list as "Otter.ai Notetaker." It transcribes live with a two-second delay, produces a summary with decisions and action items, and emails the result before you are back at your desk. The autonomy is the point — setup is connecting a calendar, and after that the bot just shows up. But the visibility is also the catch: everyone in the call can see it, and in client conversations or sensitive discussions, a visible recorder changes what people say, or gets the bot asked to leave.
What the transcripts are actually good for
The lasting value is search, not the transcript itself. Otter's AI Chat answers questions across your entire meeting history with citations — "what did we agree on pricing with Acme Corp?" returns the relevant quotes with timestamps from past calls, which is what turns months of meetings into a searchable memory instead of files nobody reopens. Action items are extracted and assigned to speakers automatically, and follow-up email drafts cut out the last twenty minutes of recap work.
A realistic exchange shows the shape of it. Ask three weeks after a client call:
What pricing did we discuss with Acme Corp?
and the answer comes back with receipts: a 200-word summary plus the exact transcript passages it came from, each with a timestamp and a link back to the source meeting. Instead of re-listening to a 45-minute recording to find one number, the number finds you in seconds. The catch is the same as any search engine: the answer is only as good as the transcripts it searches, so every transcription error quietly propagates into the answers.
Accuracy is where expectations need adjusting. Independent tests put word accuracy at roughly 85–92 percent on clear American English, dropping to the 70s on heavy accents, overlapping speech and poor microphones, and lower still on dense technical jargon — some tests logged accuracy in the 50–60 percent range on API names and frameworks. Speaker labels occasionally merge soft-spoken voices. The practical rule: reliable for summaries and recall, but not for word-for-word records of price-sensitive or technical discussions, and AI Chat inherits every transcription error — it confidently answers from transcripts, including the mistakes.
The failure modes nobody advertises
Two failure modes matter. First, accuracy collapses exactly where the stakes are highest: a vendor call with a heavy accent produced a transcript that misheard pricing terms — caught only because a human was on the call. Second, the summary occasionally invents action items, assigning a task to someone who never spoke. Both are rare but consequential, and both mean the output needs spot-checking for anything that will be acted on. There is also the IT reputation problem: Otter's aggressive auto-join has led some organizations to block it entirely, so enterprise buyers should check their security policy before standardizing on it.
What the free tier gets you
Free gives 300 transcription minutes a month, but the caps stack: each conversation is limited to 30 minutes, and file imports to three total. Pro removes the worst of them. The table is the fastest way to see whether your meeting pattern fits:
| Free | Pro (yearly) | |
|---|---|---|
| Monthly transcription | 300 minutes | 1,200 minutes |
| Per-conversation cap | 30 minutes | 90 minutes |
| File imports | 3 total | Unlimited |
| AI Chat history search | Limited | Full history with citations |
| Price | $0 | $8.33/mo (or $16.99 monthly) |
The per-meeting cap is what bites first: a one-hour call on free eats the whole allowance in one meeting and stops at the 30-minute mark. Anyone in 5+ meetings a week is effectively on Pro from day one; a student discount ($6.67/month) is why it is also common in lecture capture. For a light load — occasional standups and 1:1s — free is genuinely usable.
Where it fits
- Works for: sales and customer-research teams who want call notes pushed to a CRM and searchable history; anyone who reviews meetings days or weeks later and currently re-listens to recordings; students capturing lectures at the discount price.
- Not a fit for: client-facing conversations where a visible recorder changes the dynamic — the bot-free alternatives record locally instead; technical discussions dense with jargon, where accuracy drops and transcripts need heavy editing; teams whose security policy forbids third-party meeting bots, which is an increasingly common enterprise stance.