Otter joins Zoom, Microsoft Teams and Google Meet calls to transcribe them live, then turns the transcripts into summaries, action items and a searchable knowledge base you can query. It serves individuals on a free tier and teams that want meeting content pushed into Salesforce, HubSpot, Notion and other systems.
Used this tool? Rate it
Used this tool? Rate it
Otter is a meeting transcription service that has spent the last two years trying to become something larger. The original product did one thing: it listened to a meeting and produced a searchable transcript. The current product still does that, and it remains the reason most people sign up, but the company now describes it in broader terms. The homepage leads with the claim that your AI notetaker is now also your Conversational Knowledge Engine, and that Otter transcribes every meeting, turning it into searchable knowledge that powers your workflow.
That reframing matters when you evaluate the tool, because it explains why the pricing sheet looks the way it does and why the feature list has grown well past transcription. Understanding which parts are mature and which parts are recent is most of the work of deciding whether to use it.
The operator is disclosed plainly. The privacy policy states that Otter.ai, Inc. is a company registered in Delaware with its registered address at 800 W El Camino Real, Suite 170, Mountain View, CA 94040. It is a venture-backed company founded in 2016, not an anonymous web utility, and it has been in this category longer than most of its competitors.
This entry is filed under the name Otter Meeting Agent, which is a specific capability rather than the whole company. The Verge reported its launch in March 2025: the most notable of a trio of agents was the aptly named Otter Meeting Agent, a voice-activated meeting assistant capable of using company data to answer questions and complete tasks, initially usable on Zoom calls and rolled out incrementally to all Otter.ai users. Microsoft Teams and Google Meet support followed within weeks.
It helps to hold the hierarchy clearly. Otter.ai is the company and the platform. Otter Meeting Agent is the voice-activated assistant inside meetings. OtterPilot is the bot that joins calls. Otter Sales Agent, SDR Agent and Recruiting Agent are separate role-specific products. When third parties write about "Otter", they usually mean the transcription platform as a whole.
For most users the loop is unglamorous and useful. A meeting appears on your calendar, the notetaker joins it, the transcript scrolls live while people talk, and afterwards you get a summary with decisions and action items. If you need something from a meeting three months ago, you search rather than scrub through a recording. Everything beyond that — the agents, the CRM sync, the knowledge base — is built on top of that loop and is optional.
This is the mature part of the product and the part reviewers consistently rate highest. Transcription runs in real time, with words appearing as they are spoken rather than after the call ends, and the system attributes lines to individual speakers. The privacy policy explains the mechanism: when you use the Services, you may record and annotate spoken conversations, from which Otter generates speaker identification information, used to recognise speakers and automatically tag their name within the transcript.
Language coverage is narrower than the marketing implies and worth checking before committing. The pricing page states that AI transcription is available in English, Spanish, French, German, Japanese, and Chinese. That is six languages, not universal support.
Otter deliberately does not force one capture model. The homepage summarises the options: whether you automatically send your AI Notetaker to meetings, record bot-free on desktop or Chrome, or capture on mobile, Otter gives you flexible ways to record meetings anywhere.
This matters more than it sounds. The bot-joining model is visible to everyone on the call, which some organisations require and others find intrusive. The bot-free desktop path records system audio without announcing itself in the participant list. The mobile path handles in-person conversations that no meeting platform sees at all. Which one you choose has legal implications, covered in the Limitations section.
Every transcript is condensed into a summary. The homepage frames the output as a clear summary with decisions, action items, and insights, with next steps automatically captured and assigned. In practice this is the feature people cite when they say the tool saves them time, and also the feature that draws the most criticism when meetings are complex — reviewers report that summaries of layered, fast-moving discussions tend toward the generic.
Rather than reading transcripts, you can ask questions about them. AI Chat searches across meetings and, more recently, across connected applications. It is metered rather than unlimited, which is easy to miss: the pricing page caps it at 20 queries per user on the free tier with just 3 per conversation, 50 on Pro, and 200 on Business and Enterprise.
The Meeting Agent is the newest interaction model and the narrowest in scope. Computerworld's reporting describes the implementation: the system uses retrieval-augmented generation to analyse a user's request, breaks it down into subtasks and decides which functions to call or which internal system to query. Importantly, it is on-demand rather than autonomous — the agent is only activated on voice requests and doesn't interrupt conversations.
Fast Company documented the permission model, which is the part worth knowing before enabling it in a mixed-company call: the AI will only provide answers based on meeting records that all current participants have permission to view, ensuring confidential information remains protected.
The most recent expansion inverts the data flow. TechCrunch reported in April 2026 that with this launch, users can connect their Gmail, Google Drive, Notion, Jira, and Salesforce accounts and query that data along with existing meeting data, and that users can not only search across these tools but also push meeting summaries to Notion or draft a Gmail message. Previously Otter pushed meeting data outward; now it also pulls other systems inward.
Transcripts are working documents rather than archives. The pricing page lists editable text and speaker tags, real-time annotation with highlights and comments, a takeaways panel that captures annotations and assigns action items, automated slide capture for virtual meetings, folders, sharing via private channels and public links, and export to mp3, txt, pdf, docx and srt depending on tier.
The most common real use is asynchronous catch-up. Someone who could not attend reads the summary, skims the transcript for the section that concerns them, and skips the recording entirely. This works well precisely because the transcript is searchable and attributed, and it works badly when the meeting was a fast four-way discussion whose summary flattens into generalities.
Sales is the first of Otter's five named industry lines. The homepage describes it as capturing sales call notes, getting live coaching tips, automating follow-ups, and driving sales efficiency with CRM. The practical value is eliminating post-call data entry, though note that CRM sync is tiered — see Pricing.
Education and media are two more named lines: automatically capturing lectures, discussions and study sessions and turning them into searchable notes, and capturing interviews and production meetings to produce searchable transcripts, key quotes and draft-ready notes. Journalists in particular use transcription tools for interview turnaround, where an 85%-accurate transcript that can be corrected beats typing from audio.
The mobile app covers what no meeting platform captures: conference hallway conversations, site visits, in-person interviews. Reviewers on third-party platforms single this out as a genuine strength. It also carries the highest consent risk, because in-person recording rules vary sharply by jurisdiction.
The enterprise pitch is that meetings stop evaporating. Instead of decisions living in individual memories, they live in a queryable corpus. This is the ambition behind the Conversational Knowledge Engine framing, and it is the part most worth scrutinising — an outside analyst quoted by Computerworld pushed back directly on that positioning, as covered in Limitations.
Live captioning is a use case the marketing underplays. Real-time on-screen transcription serves participants who are deaf or hard of hearing, and this is one of the few scenarios where live transcription is strictly better than an accurate post-hoc transcript, because the value is in the moment.
Create a free account before evaluating anything else. Accuracy is highly dependent on your specific conditions — accents, microphone quality, how much people talk over each other — and no review can tell you how it will perform on your calls. Run it on three real meetings before deciding.
Choose between the bot joining calls, the bot-free desktop app, the Chrome extension, and the mobile app. This is a policy decision, not a preference. TechCrunch quoted Otter's CEO saying enterprise customers mostly prefer the notetaker that joins the Zoom meeting because it provides the transparency, and that they also prefer the meeting notes to be shared with all the meeting attendees, so that the note is not limited to one person.
Calendar integration is what makes the tool effortless and also what makes it risky: it will join everything unless told otherwise. Reviewers commonly report that auto-join was the first thing they had to learn to switch off. Decide explicitly which meeting types should be recorded before you connect anything.
Do this regardless of what the tool does automatically. Otter's own privacy policy places the obligation on you: if you provide an Audio Recording, this may contain the Personal Information of third parties, and before you do so, please make sure you have the necessary permissions from your co-workers, friends or other third parties. This is not boilerplate — it is the exact issue under litigation.
Custom vocabulary is the cheapest accuracy improvement available. Add names of colleagues, product names, acronyms and industry jargon. The free tier allows only 5 terms, so this is realistically a paid-tier optimisation.
Treat the generated summary as a first draft. Correct action items and misattributed speakers before circulating, because errors in a shared summary propagate into other people's task lists. The takeaways panel exists for exactly this editing pass.
Once you have a few weeks of history, change your habit: search the corpus rather than hunting for a recording. This is where the tool's compounding value actually lives, and it is the step most users never reach.
The single most common complaint in third-party reviews is running into minute caps mid-workflow. The numbers are specific: 300 minutes per user per month on Basic, 1,200 on Pro, unlimited on Business and Enterprise, and the pricing page notes these do not roll over. Count your actual weekly meeting hours before picking a plan.
This is the limit that catches people out. The cap on a single conversation is 30 minutes on Basic, 90 minutes on Pro, and 4 hours on Business and Enterprise. A free-tier user with a 45-minute meeting gets cut off mid-conversation even with monthly minutes to spare. Match the per-meeting cap to your longest routine meeting, not your average one.
Accuracy degradation is overwhelmingly an input problem. A single speaker on a decent microphone in a quiet room produces dramatically better results than four people on laptop mics with crosstalk. If transcripts are disappointing, the highest-leverage fix is usually a meeting-room microphone, not a different vendor.
Legal discussions, HR conversations, performance reviews and anything under privilege should be excluded by rule rather than by remembering in the moment. Configure this once rather than relying on discipline.
Basic retains just the 25 most recent conversations. If you are trialling the tool as a long-term record, that limit will quietly delete the evidence you were trying to accumulate. Trial it for the workflow, not for the archive.
Export is tiered: mp3 and txt only on Basic, adding pdf, docx and srt on paid tiers, with bulk export on Business and Enterprise. If your workflow depends on srt captions or bulk archival, the free tier cannot support it regardless of minutes.
This advice comes from an outside expert rather than the vendor. In Computerworld's coverage, the SANS Technology Institute's dean of research argued that organisations must rigorously test and confirm that Otter's security claims, including HIPAA compliance, hold true in their own environments before broad deployment. That is sound practice for any tool handling recorded conversations.
The clearest fit. Free tier covers light use, Pro covers heavy individual use, and the value is immediate and easy to measure: you stop typing notes. Third-party reviewers generally consider Pro on annual billing the sweet spot for a single user with controlled audio conditions.
The tightest product-market fit among team use cases, because the output has an obvious destination. Call notes land in Salesforce or HubSpot without anyone doing data entry, and coaching features layer on top. Just check the seat limits on CRM sync before assuming it covers the team.
Anyone who converts speech into text as part of their core work. An 85%-accurate transcript that you correct is much faster than transcribing from scratch, and the mobile app handles in-person interviews. Note the six-language limit if you work multilingually.
The newest and most ambitious audience, and the one that should evaluate most carefully. The capability is real — public API, MCP server, enterprise search, admin controls — but so are the governance questions raised by outside analysts, and the pending litigation is squarely about how the product records people.
Teams needing live captions for participants who are deaf or hard of hearing get real value from the live transcription quality, which is the product's strongest dimension.
Anyone whose meetings are routinely in a language outside the supported six will find the output unreliable. Teams handling privileged or highly regulated conversations should weigh the consent questions seriously and consider on-device alternatives that never send audio anywhere. Small teams needing SSO or SCIM will hit a hard wall — those require a minimum 100-user license. And organisations that need certainty about the legal status of AI notetakers should be aware that no court has yet ruled on it.
Otter joins Zoom, MS Teams, and Google Meet to automatically write and share notes, per the pricing page. That cross-platform coverage is the product's structural advantage over Microsoft's and Google's native notetakers, which work best inside their own ecosystems.
iOS and Android apps are listed at every tier including free. Desktop applications for Mac and Windows provide the bot-free recording path, with a Chrome extension covering Google Meet. Web access is the primary interface.
The integration list is long and specific: Slack, Zoom, Salesforce, HubSpot, Notion, Jira, Asana, Google Calendar, Google Docs, Google Meet, Dropbox, Glean and Claude are all named on the homepage. Zapier covers the rest.
Two developer-facing paths exist with very different access levels. The MCP server appears on the free tier's feature list, and the homepage explains that ChatGPT, Claude, and other AI chat tools can securely access your meeting knowledge through Otter's MCP Server for deeper analysis and smarter workflows. The traditional API is far more restricted: the pricing page places Otter API & Webhooks in the Enterprise column only. If you were planning to build on the API without an enterprise contract, that is a blocker.
Six for transcription: English, Spanish, French, German, Japanese and Chinese. The interface is English-first. Multilingual meetings — where speakers switch languages mid-conversation — are not a supported scenario.
The pricing page lists Basic as free, Pro at $16.99 per user per month or $8.33 on annual billing, Business at $30 per user per month or $19.99 on annual billing, and Enterprise as contact-sales. Annual billing roughly halves the effective monthly price, which is a larger discount than most comparable tools offer.
Basic is a genuine free tier rather than a trial, but it is fenced carefully. You get 300 monthly transcription minutes per user, a 30-minute cap per conversation, 3 lifetime audio/video file imports, 1 concurrent meeting, 25 most recent conversations of history, 5 custom vocabulary terms, 20 AI Chat queries per user with 3 per conversation, mp3 and txt export, and 1x playback speed only. It is enough to evaluate the product seriously and enough for genuinely light use.
Pro raises monthly minutes to 1,200 per user, per-conversation cap to 90 minutes, imports to 10 monthly per user, concurrent meetings to 2, and AI Chat to 50 queries. Business removes the meeting minute cap entirely, raises per-conversation to 4 hours, makes imports unlimited, allows 3 concurrent meetings, and raises AI Chat to 200 queries.
The subtler difference is seats and CRM. Workspace membership is capped at 5 members on Basic, 5 billed users on Pro, 25 billed users on Business, and unlimited on Enterprise. Salesforce or HubSpot note sync covers 1 user on Pro, 5 users on Business, and unlimited on Enterprise — so a ten-person sales team on Business will find CRM sync covers only half of them.
SSO and SCIM both require a minimum 100-user license. HIPAA compliance is an add-on rather than included. Otter API and Webhooks are Enterprise-only, as is AI Chat with Salesforce. If any of these are requirements, the effective entry price is an enterprise contract rather than a listed per-seat rate.
Live transcription via RTMP carries additional fees per the pricing page footnote. Salesforce and HubSpot integrations are subject to user limits. And imported-file transcription has its own separate monthly ceiling — 6,000 minutes per user even on unlimited tiers.
Microsoft Teams and Google Workspace both include meeting AI. The trade-off is straightforward: if your organisation lives entirely in one ecosystem, the native option is cheaper and requires no additional vendor review. Otter's counter-argument is cross-platform coverage plus in-person capture via mobile, which native tools cannot match.
TechCrunch's reporting names the field directly: the AI boom that began in 2022 fuelled a surge in startups like Granola and Circleback, while older players such as Fireflies have seen a surge in interest. A separate TechCrunch piece groups Read AI, Fireflies.ai and Fathom as notetakers that moved into enterprise search before Otter did. Several of these differentiate on bot-free capture as a default rather than an option.
For genuinely sensitive conversations, tools that transcribe locally without sending audio to a vendor sidestep both the consent question and the model-training question entirely. They typically lag on features and integrations. If your primary concern is confidentiality rather than workflow, this category deserves a look before any cloud service.
For material where accuracy must be near-perfect — legal proceedings, published interviews, regulatory records — human transcription remains the correct choice. Reviewers who value Otter on cost generally frame it as cheaper than human transcription services rather than as equivalent to them.
Choose Otter if you need cross-platform coverage, strong live transcription, and a mature integration surface, and your meetings are in the supported languages. Choose a native tool if you are single-ecosystem and cost-sensitive. Choose on-device if confidentiality dominates. Choose human transcription if errors are unacceptable.
This is the most important practical caveat. Independent testing gives a consistent picture at odds with the headline figure. One third-party assessment summarised it directly: Otter's marketing says "up to 95%" in clear audio, independent reviewers put it around 85% in good conditions with one clean speaker, a quiet room, a familiar accent and ordinary business words, and introducing accents, crosstalk, background noise or jargon leads reviewers to report accuracy sliding toward 70% or lower.
The same pattern appears across G2, Capterra and Software Advice reviews, where degradation under accents and multi-speaker conditions is the single most repeated complaint. Note that third-party testing methodologies are not standardised, so treat these as directional rather than precise.
Anyone evaluating this tool for organisational use should know this before proceeding, and should also understand exactly what has and has not been decided.
The court docket shows that four putative class actions were consolidated by order of Judge Eumi K. Lee on 22 October 2025 into lead case No. 25-cv-6911, recaptioned In re Otter.AI Privacy Litigation, in the Northern District of California, with a consolidated class action complaint filed on 5 December 2025.
The allegation, as summarised in third-party legal coverage, is that OtterPilot integrates with a user's calendar and auto-joins scheduled calls as a visible participant, capturing audio, transcripts, screenshots and speaker voiceprints, and that Otter never obtains consent from other meeting attendees. Otter denies this. Its reply brief argued that across all claims, plaintiffs do not plausibly allege that they disclosed private or sensitive information (or any information whatsoever), that Otter intercepted communications in transit, or that Otter accessed their computers or data without authorization.
What matters for a reader deciding whether to use the product: as of this writing the motion to dismiss has been argued and no ruling has issued. No court has found Otter liable, no class has been certified, and the allegations remain allegations. The underlying question — whether an AI notetaker is a tool of the meeting host or an independent third party — applies to this entire product category, not to Otter alone.
Whatever the litigation decides, the current contractual position is unambiguous. The privacy policy instructs you to make sure you have the necessary permissions from co-workers, friends or other third parties before providing recordings containing their personal information. In practice this means the compliance burden sits with the person who switched the bot on, and in all-party-consent jurisdictions that burden is substantial.
Easy to overlook: the privacy policy states that OtterPilot may take automatic screenshots which are available in meeting transcripts to add value by extracting useful visual information, and that automatic screenshots only take place in virtual meetings. If your screen shares include confidential material, it is being captured as images alongside the audio.
A small but useful detail for procurement review: the privacy policy states that for Pro or Business plans, our payment processing partner Stripe, Inc. may also collect your name, billing address, and payment information, and that payment information is not shared with Otter and is maintained by Stripe. That narrows the payment-data surface Otter itself holds.
The privacy policy sets out that where an enterprise service agreement is in place with a customer asking you to use the service — for example your employer — Otter processes your personal information on behalf of and at the instructions of that customer, and those enterprise customers are the data controllers with their own privacy policies applying. If you use Otter through work, your rights run against your employer, not primarily against Otter.
Reviewers repeatedly report that action item detection and contextual summarisation are shallow relative to expectations, particularly in meetings with layered topics or fast-paced discussion, where summaries turn generic. The transcript is the reliable artefact; the summary is a convenience that needs checking.
Worth weighing against the marketing. In Computerworld's coverage of the enterprise launch, a Gartner analyst argued that no meeting platform should be considered the "single source of truth" for sales or enterprise data, that at best Otter provides a valuable lens onto meeting conversations, and that core systems like CRM or contract management remain authoritative for their respective domains.
The scale figures circulating in coverage originate with the company. TechCrunch reported that last year the company said it had 25 million users and $100 million in annual recurring revenue, and that while it did not provide new financials, it said the platform now has 35 million users. These are vendor statements relayed by journalists, not audited figures.
Yes. Basic is free rather than a time-limited trial, and includes 300 monthly transcription minutes per user, Zoom, Teams and Google Meet support, live transcription, speaker identification, iOS and Android apps, and MCP server access. The meaningful constraints are a 30-minute cap per conversation, 3 lifetime file imports, 25 most recent conversations of history, and 20 AI Chat queries per user.
It depends heavily on your audio. The vendor markets up to 95% in clear conditions, but independent reviewers put real-world accuracy around 85% with one clear speaker in a quiet room, and report it sliding toward 70% or lower once accents, crosstalk or background noise are present. Test it on your own meetings rather than trusting any published figure, including these.
Six for AI transcription: English, Spanish, French, German, Japanese and Chinese, per the official pricing page. Meetings that switch languages mid-conversation are not a supported scenario, and languages outside this set should not be assumed to work.
Yes, and you should treat this as your responsibility rather than the tool's. Otter's own privacy policy instructs users to make sure they have the necessary permissions from third parties before providing recordings containing their personal information. Requirements vary by jurisdiction, and some states and countries require consent from every participant rather than just one.
There is pending litigation. Four putative class actions were consolidated in October 2025 as In re Otter.AI Privacy Litigation in the Northern District of California, alleging that its notetaker recorded participants without all-party consent. Otter denies wrongdoing. As of this writing the motion to dismiss has been argued and no ruling has issued — no court has found the company liable and no class has been certified.
Yes. Alongside the notetaker that joins calls, Otter offers bot-free recording on the Mac and Windows desktop apps and via a Chrome extension, plus mobile recording for in-person conversations. Note that Otter's CEO has said enterprise customers generally prefer the visible bot because it provides transparency to everyone on the call.
Business is $30 per user per month, or $19.99 on annual billing, and removes the meeting minute cap while raising the per-conversation limit to 4 hours and allowing 3 concurrent meetings. Watch two constraints: workspace membership caps at 25 billed users, and Salesforce or HubSpot note sync covers only 5 users on that tier.
Two different answers. The MCP server is listed at every tier including free, letting tools like ChatGPT and Claude access your meeting knowledge. The traditional Otter API and Webhooks are Enterprise-only per the pricing page, so building a custom integration requires an enterprise contract.
Yes. The privacy policy discloses that OtterPilot may take automatic screenshots, which are included in meeting transcripts to extract useful visual information, and that this happens only in virtual meetings. It also generates speaker identification information from recorded conversations to tag names in transcripts.
Your employer, in most respects. The privacy policy states that where an enterprise service agreement is in place with a customer asking you to use the service, Otter processes personal information on behalf of and at that customer's instructions, and those enterprise customers are the data controllers whose own privacy policies apply. Requests about your data generally go to your employer first.
