How to Use Otter.ai to Search Across a Year of Meetings
Otter’s cross-meeting search returns a timestamped line from any call in your archive, but only if the calls were filed properly. After a year of unfiled meetings, search returns a hundred near-identical results and you give up.
Thirty minutes of cleanup fixes that, and it makes Otter materially more useful than the transcript you never re-read.
How does Otter search actually work?
Search runs full-text over transcripts, summaries, and comments, then groups results by meeting. Each hit shows the sentence, the speaker, and a timestamp you can click to jump into the audio.
What it does not do well out of the box: synonyms. Search “churn” and you get meetings that said churn, not meetings that said “customers leaving in Q3.” Keyword coverage beats semantic understanding here.
That gap is the reason naming conventions matter so much. You are building an index, and the index only sees words.
What naming convention should you use?
One format, every meeting, no exceptions:
[Team] Meeting type — topic
For example: [Sales] QBR — North America pipeline review.
The bracket makes the team filterable. Otter supports folder-style organization on paid tiers but not true nested taxonomy, so the title is the taxonomy.
Three rules:
- Month and year in the title for recurring calls.
[Sales] Weekly — 2026-03 pipelinebeats[Sales] Weeklyrepeated 40 times. - Client or project name spelled the same way every time. If two people type
AcmeandACME Inc, search splits them. - Never leave the auto-generated title. Default titles use the calendar event name, which is often “Meeting” or a person’s first name.
How do you find a decision from months ago?
Work from broad to narrow. In order:
- Search the decision noun, not the verb. “pricing” finds more than “decided.”
- Add the project name as a second term. Most search boxes, including Otter’s, AND the terms by default.
- Filter by date range before scrolling. A year of results in chronological order is unusable; a one-month window is browsable.
- Once you find the meeting, search inside that single transcript for the speaker who raised it.
If you know who said it, search the speaker name plus one keyword. Speaker attribution in Otter is reliable when each person joined from their own device and unreliable in a shared conference room.
What are the plan limits on search and retention?
| Plan | Minutes/month | Retention | Advanced search |
|---|---|---|---|
| Basic (free) | 300 | Standard retention | No |
| Pro (~$17/mo) | 1,200 | Extended, 90-min max per call | Yes |
| Business (~$30/seat/mo) | 6,000 | Extended plus admin controls | Yes |
| Enterprise | Custom | Custom retention, SSO | Yes |
Advanced search — the filters and saved searches — is a paid feature. On free, you get a plain search box. Team libraries, where a shared archive is visible to everyone, also require Business.
Check retention before you rely on it. Deleted or expired recordings drop out of search, and the surrounding summary may stay behind with no audio to verify against.
What breaks search at scale?
Four problems show up once an archive passes a few hundred meetings.
Duplicate uploads. The same call recorded twice (bot plus local recording) produces two transcripts, both searchable. Deduplicate monthly.
Speaker labels across accounts. On a shared Business workspace, the same person may appear under two email addresses, splitting their speech in half.
Board names. If your company calls something “Project Atlas” in January and “Atlas 2.0” in June, search for one misses the other. Keep a glossary of project names.
Non-English calls. Free tier is English only; paid tiers add French, Spanish, and Japanese. Mixed-language calls transcribe poorly and search badly.
Who gets real value from the archive?
Worth it for: account managers reviewing past client commitments, PMs reconstructing why a feature was cut, and anyone in a role where “we agreed X in that call” is a recurring claim.
Not worth it for: teams with fewer than 10 meetings a month (the index never gets dense enough), and anyone whose calls happen over a bad conference-room mic where speaker labels are wrong more than half the time.
One practical note: if a call is confidential, do not record it into a shared workspace. Otter has no per-meeting access control on lower tiers — the workspace sees what the workspace sees.
Quick Summary
- Otter search is keyword full-text, not semantic — naming conventions do the heavy lifting.
- Title format
[Team] Type — topicwith dates on recurring calls. - Search the noun plus the project name, then filter by date before browsing.
- Advanced search and team libraries require paid tiers; Pro runs about $17/month for 1,200 minutes.
- Duplicate uploads and split speaker identities are the two biggest search-quality drains.
