The best AI meeting assistants, tested and ranked
We had six AI meeting assistants take notes on the same five scripted video calls and checked every transcript and summary against our own. Here's which one got the most right, which fits how you work, and what each costs to start.
About this list
An AI meeting assistant joins or captures your video calls, writes a transcript, then turns it into a summary and a list of action items you can search later. Many also send notes to your CRM or Slack and let chat assistants such as Claude and ChatGPT read your meetings. You'll find more tools like these in our AI meeting assistants section and our AI chat tools hub.
They differ most in how accurate the notes are, how well they tell speakers apart, how much the free plan allows, and what happens to your recordings. Some send a visible bot into the call while others capture quietly from your computer, and some use your meetings to train their own AI.
To compare them fairly, every tool took notes on the same five scripted video calls between October 16 and October 22, 2026. The calls included different accents, people talking over each other and background noise, and each had a transcript checked by a person. We counted word errors, checked speaker labels, scored the summaries against a fixed checklist, timed how fast notes arrived and checked how each tool handles your data.
Compare at a glance
| # | Tool | Best for | Cheapest way to start | Score |
|---|---|---|---|---|
| 1 | Otter.ai | Best overall | $16.99 a month (Pro) | 4.0 |
| 2 | tl;dv | Best for European data hosting | 29 EUR a month (Pro) | 3.0 |
| 3 | Fathom | Best free plan | $19 a month (Team) | 2.5 |
| 4 | Fireflies.ai | Best for integrations and languages | $18 a month (Pro) | 2.5 |
| 5 | Read.ai | Best for meeting analytics | $19.75 a month (Pro) | 2.5 |
| 6 | Tactiq | Best bot-free option on a budget | $8 a month, billed yearly (Pro) | 2.0 |
Features and terms compared
These facts come from each company's own pages and our research, checked on November 2 and 3, 2026.
| Tool | Free plan | Cheapest monthly start | Bot-free option | Languages | Trains on your data |
|---|---|---|---|---|---|
| Otter.ai | 300 minutes a month, 30 minutes per call | $16.99 a month (Pro) | Yes, desktop app | Not listed | Own models, de-identified; no opt-out described |
| tl;dv | Unlimited recordings; AI notes capped | 29 EUR a month (Pro) | Yes | 30+ | No, per its site |
| Fathom | Unlimited recordings; some AI features limited | $19 a user a month (Team, two-user minimum) | Yes, in beta | Not stated | Own models, de-identified; opt-out available |
| Fireflies.ai | 400 minutes of storage per team | $18 a month (Pro) | Not listed | 100+ | No, per its site |
| Read.ai | 5 transcripts a month, 1-hour meetings | $19.75 a month (Pro) | Not stated | 20+ | No by default, per its site |
| Tactiq | 10 transcripts and 5 AI credits a month | $8 a month, billed yearly (Pro) | Yes, never uses a bot | 60+ | Not stated |
Test results side by side
Otter.ai was clearly ahead. It got 291 of 5,000 words wrong (a 5.8% error rate), labelled 87 of 100 speakers correctly, caught 42 of 50 points on our summary checklist and had notes ready 10.5 minutes after each hour of audio. Every other tool made at least 500 word errors, got 70 or fewer speaker labels right and took between 15.6 and 17.9 minutes per audio hour.
Behind Otter the field was close and the order was decided by small margins. tl;dv came second with 500 word errors and 36 checklist points. Fireflies.ai wrote the second-best summaries with 37 points but came fourth, held back by ease of use and privacy. Fathom and Read.ai each caught only 30 points, and Tactiq came last with 705 word errors and only 54 of 100 speaker labels right. The lesson: even good tools miss things, so check names and action items before you send a recap.
Every tool got the same test, between October 16, 2026 to October 22, 2026.
| Tool | Score | Word error rate against the human transcript | Speaker labels correct | Summary and action items: points captured out of a fixed checklist | Turnaround time per hour of audio | Data handling: storage location, retention, opt-out of model training |
|---|---|---|---|---|---|---|
| Otter.ai | 4/5 | 291/5000 word errors (5.8%) | Speaker labels 87/100 correct | Checklist captured 42/50 | 10.5 min/audio hour | Retention/deletion/training checks 5/6 |
| tl;dv | 3/5 | 500/5000 word errors (10.0%) | Speaker labels 69/100 correct | Checklist captured 36/50 | 15.6 min/audio hour | Retention/deletion/training checks 4/6 |
| Fathom | 2.5/5 | 633/5000 word errors (12.7%) | Speaker labels 70/100 correct | Checklist captured 30/50 | 16.6 min/audio hour | Retention/deletion/training checks 3/6 |
| Fireflies.ai | 2.5/5 | 600/5000 word errors (12.0%) | Speaker labels 66/100 correct | Checklist captured 37/50 | 17.9 min/audio hour | Retention/deletion/training checks 3/6 |
| Read.ai | 2.5/5 | 653/5000 word errors (13.1%) | Speaker labels 65/100 correct | Checklist captured 30/50 | 16.9 min/audio hour | Retention/deletion/training checks 4/6 |
| Tactiq | 2/5 | 705/5000 word errors (14.1%) | Speaker labels 54/100 correct | Checklist captured 27/50 | 16.4 min/audio hour | Retention/deletion/training checks 3/6 |
Otter.ai
Best overallOtter.ai came first of six with 4 out of 5, the only tool above 3. It made 291 word errors in 5,000 (5.8%), labelled 87 of 100 speakers correctly, caught 42 of 50 summary points and was the fastest, at 10.5 minutes per hour of audio. It passed 5 of our 6 data checks, the best result here, and scored 4.5 for value. Our Otter review has the full results.
Otter records Zoom, Google Meet and Teams calls with a bot, or bot-free through its desktop app, and gives you a live transcript, a summary and action items. Its AI Chat answers questions about past meetings, and notes can go to Salesforce or HubSpot. It has apps for iOS, Android, Mac and Windows, rated 4.7 on the App Store from 50,142 ratings and 4.4 on Trustpilot from 830 reviews (checked November 2026).
The catches: Otter trains its own models on de-identified data with no opt-out described, fees are non-refundable, and a class action filed in August 2025 claims its notetaker recorded people without consent. The free plan is tight at 300 minutes a month and 30 minutes per conversation.
Who it's for
Individuals and small teams who want accurate everyday meeting notes with good phone and desktop apps.
Skip it if
You want a training opt-out or EU data hosting. tl;dv hosts data in Europe and says it doesn't train on your meetings.
What we liked
- 5.8% word error rate, the lowest in our test
- 42 of 50 summary points caught
- Notes ready 10.5 minutes after each audio hour
What we didn't
- Trains its own models, no opt-out described
- Free plan capped at 30 minutes per call
- Subscriptions are non-refundable
Key features
- Live transcripts with speaker labels
- AI Chat across past meetings
- Bot-free recording in the desktop app
- Salesforce and HubSpot sync from Pro
- MCP server for Claude and ChatGPT
Plans and prices
| Basic 300 monthly minutes; 30 minute meetings. | Free |
| Pro 1,200 monthly minutes; 90 minutes per meeting. | $16.99 a month, or $8.33 a month billed yearly |
| Business Unlimited meetings; 4 hour meetings. | $30 a month, or $19.99 a month billed yearly |
| Enterprise SSO and SCIM (100-user minimum); API and webhooks. | Custom contact sales |
tl;dv
Best for European data hostingtl;dv came second with 3 out of 5. It made 500 word errors in 5,000 (10.0%), caught 36 of 50 summary points and passed 4 of our 6 data checks. Speaker labels were its weak spot at 69 of 100, and notes took 15.6 minutes per audio hour. Value was its lowest part at 2.5. Read our tl;dv review for more.
tl;dv is a German company that records, transcribes and summarises Zoom, Meet and Teams calls in 30+ languages and can run without a bot. It hosts data in Europe, lets you choose EU or US AI processing and says no customer data is used to train AI. Sales teams get notes in formats such as MEDDIC, CRM logging and reports across meetings.
There's no native mobile app, AI notes are limited on the free plan, prices are in euros and fees are non-refundable. Its Trustpilot profile is tiny and negative, at 2.1 out of 5 from 10 reviews (checked November 2026).
Who it's for
European companies and sales teams that care where their meeting data lives.
Skip it if
You want the most accurate notes. Otter.ai did better on every measure in our test.
What we liked
- EU hosting and a choice of EU AI processing
- Free plan with unlimited recordings
- Says it doesn't train on customer data
What we didn't
- Only 69 of 100 speaker labels correct
- No native mobile app
- Fees are non-refundable
Key features
- Transcripts in 30+ languages
- AI notes in custom formats such as MEDDIC
- Reports across many meetings
- CRM logging and follow-up drafts
Plans and prices
| Free Unlimited recordings. | Free |
| Pro | 29 EUR a month, or 18 EUR a month billed yearly |
| Business | 39 EUR a month, or 29 EUR a month billed yearly |
Fathom
Best free planFathom came third with 2.5 out of 5. It made 633 word errors in 5,000 (12.7%) and labelled 70 of 100 speakers correctly, the second-best after Otter. But its summaries caught only 30 of 50 points, notes took 16.6 minutes per audio hour, and it passed 3 of our 6 data checks. Ease of use and privacy were its weakest parts at 2. Our Fathom review covers the full test.
Fathom's hook is its free plan, with unlimited recordings and transcripts, instant summaries, clips and search; Ask Fathom and AI action items are limited. Paid plans add advanced summaries, team search, CRM sync and coaching scorecards, all with a 90-day money-back guarantee. You can opt out of Fathom training its own models on your data, and bot-free capture is in beta.
Data is stored in the US only, and Superhuman announced it is buying Fathom in September 2026. Users rate it highly, at 4.8 out of 5 on Trustpilot from 287 reviews (checked November 2026), though some report a bot that failed to join and slow email-only support.
Who it's for
Freelancers and small teams who want free, unlimited meeting recording and will skim the notes themselves.
Skip it if
You want a recap you can send without checking. Otter.ai caught 42 of 50 points to Fathom's 30.
What we liked
- Unlimited free recordings and transcripts
- 90-day money-back guarantee
- Opt-out from its own model training
What we didn't
- Summaries caught only 30 of 50 points
- Passed just 3 of 6 data checks
- Data stored in the US only
Key features
- Instant AI summaries after each call
- Clips and playlists of call moments
- CRM field sync and AI scorecards on Business
- Bot-free capture in beta
Plans and prices
| Free Unlimited recordings and transcriptions; Instant AI summaries. | Free |
| Premium Advanced summaries; AI action items. | $20 a month, or $16 a month billed yearly |
| Team Global search; Playlists. | $19 a month, or $15 a month billed yearly |
| Business CRM field sync; Deal View. | $34 a month, or $25 a month billed yearly |
Fireflies.ai
Best for integrations and languagesFireflies.ai came fourth with 2.5 out of 5. Its summaries were the second-best in the test at 37 of 50 points, and it made 600 word errors in 5,000 (12.0%). But it labelled only 66 of 100 speakers correctly, was the slowest at 17.9 minutes per audio hour and passed 3 of our 6 data checks. Ease of use and privacy scored 2. More in our Fireflies review.
Fireflies sends a bot into Zoom, Meet and Teams calls and turns them into a searchable archive. It lists 100+ transcription languages and many integrations, including HubSpot, Salesforce, Greenhouse and Slack, plus 200+ AI Skills, talk-time analytics on Business and an API. It says customer data isn't used to train AI models.
Some features run on AI credits charged per request, the free plan caps storage at 400 minutes per team, and its refund guidance conflicts. A proposed class action filed in December 2025 claims its speaker recognition creates voiceprints without consent. It has 4.5 out of 5 on Trustpilot from 885 reviews (checked November 2026).
Who it's for
Sales, recruiting and multilingual teams that want a searchable call archive linked to their other tools.
Skip it if
You need a reliable record of who said what. Otter.ai labelled 87 of 100 speakers correctly to Fireflies' 66.
What we liked
- 37 of 50 summary points, second-best
- 100+ languages and many integrations
- Says it doesn't train on customer data
What we didn't
- Only 66 of 100 speaker labels correct
- Slowest at 17.9 minutes per audio hour
- Conflicting refund guidance
Key features
- Transcripts with speaker recognition
- AskFred questions and 200+ AI Skills
- Talk-time and topic analytics on Business
- API and MCP access
Plans and prices
| Free 400 minutes storage per team; Unlimited transcription. | Free |
| Pro More credits than Free; Advanced exporting. | $18 a month, or $10 a month billed yearly |
| Business Everything in Pro; Video recording. | $29 a month, or $19 a month billed yearly |
| Enterprise SSO and SCIM; Private Storage. | $39 a month, billed yearly |
Read.ai
Best for meeting analyticsRead.ai came fifth with 2.5 out of 5, scoring 2.5 on every part. It made 653 word errors in 5,000 (13.1%), labelled 65 of 100 speakers correctly and caught only 30 of 50 summary points, with notes taking 16.9 minutes per audio hour. Its data result was fair: 4 of our 6 checks.
Read records and summarises meetings and adds engagement readouts and coaching. Ask Read answers questions with sources across your meetings, email and chat, and Ada is an email assistant that handles scheduling. It has apps for Windows, Mac, iOS and Android and says it doesn't train on customer data by default.
The free plan allows only 5 transcripts a month and 1-hour meetings, and its terms say no refunds. Users are unhappy: 1.3 out of 5 on Trustpilot from 111 reviews (checked November 2026), with complaints about broad permissions and the app joining meetings without approval.
Who it's for
Managers who want meeting analytics and search across meetings, email and chat.
Skip it if
You mainly want accurate notes. Otter.ai made far fewer errors and caught 42 of 50 summary points.
What we liked
- Passed 4 of 6 data checks
- Ask Read searches meetings, email and chat
- Engagement and coaching readouts
What we didn't
- 13.1% word error rate
- Free plan capped at 5 transcripts a month
- Very poor Trustpilot rating
Key features
- Transcripts, recaps and action items
- Ask Read with cited answers
- Meeting coach and engagement readouts
- Ada email assistant
Plans and prices
| Free 5 transcripts a month; 1 hour meetings. | Free |
| Pro Unlimited reports; 4 hour meetings. | $19.75 a month, or $15 a month billed yearly |
| Enterprise Video playback and highlights; 200 upload minutes. | $29.75 a month, or $22.50 a month billed yearly |
| Enterprise+ HIPAA; SAML SSO. | $39.75 a month, or $29.75 a month billed yearly |
Tactiq
Best bot-free option on a budgetTactiq came last of six with 2 out of 5. It made 705 word errors in 5,000 (14.1%), the most in the test, labelled only 54 of 100 speakers correctly and caught 27 of 50 summary points. Notes took 16.4 minutes per audio hour, and it passed 3 of our 6 data checks. More in our Tactiq review.
Tactiq is a Chrome extension that transcribes Meet, Zoom and Teams calls live from your browser. No bot joins the call and no audio is stored, only text. It adds AI summaries and action items on credits and sends notes to Slack, Notion, HubSpot and AI chat tools in 60+ languages. When you hit the credit limit, transcription continues but AI pauses.
It's rated 4.8 on the Chrome Web Store from 3,000 ratings and 3.2 on Trustpilot from 15 reviews (checked November 2026), where some users report being billed after downgrading.
Who it's for
Individuals who want cheap, rough notes without a bot in the room.
Skip it if
You need notes you can rely on. Otter.ai also records bot-free from its desktop app and was far more accurate.
What we liked
- No bot joins your meetings
- Free plan with 10 transcripts a month
- Pro is $8 a month, billed yearly
What we didn't
- 14.1% word error rate, the highest
- Only 54 of 100 speaker labels correct
- Users report billing after downgrading
Key features
- Live transcripts in the browser
- AI summaries and Ask AI on credits
- Sharing to Slack, Notion and HubSpot
- MCP connector for AI chat tools
Plans and prices
| Free 10 transcripts; 5 AI credits. | Free |
| Pro Unlimited transcripts; 10 AI credits. | $8 a month, billed yearly |
| Team Unlimited transcripts and AI credits. | $16.67 a month, billed yearly |
| Business Unlimited. | $29.17 a month, billed yearly |
What it costs to get started
Every tool here has a free plan, but the limits differ a lot. Fathom and tl;dv allow unlimited recordings, while Otter.ai caps you at 300 minutes a month, Fireflies.ai at 400 minutes of storage per team, Tactiq at 10 transcripts and Read.ai at 5 transcripts a month.
Billed monthly, Otter Pro is $16.99 a month, Fireflies Pro $18, Fathom Team $19 a user (two-user minimum) or Premium $20, Read.ai Pro $19.75 and tl;dv Pro 29 EUR. Paying yearly is cheaper: Otter Pro drops to $8.33 a month and Fireflies Pro to $10. Tactiq Pro is only sold yearly, at $8 a month. Fathom offers a 90-day money-back guarantee, while Otter, tl;dv and Read.ai say fees are non-refundable.
| Tool | Cheapest way to start | Free plan | Free trial |
|---|---|---|---|
| Otter.ai | $16.99 a month (Pro) | Yes | Not stated |
| tl;dv | 29 EUR a month (Pro) | Yes | Not stated |
| Fathom | $19 a month (Team) | Yes | Yes |
| Fireflies.ai | $18 a month (Pro) | Yes | No |
| Read.ai | $19.75 a month (Pro) | Yes | No |
| Tactiq | $8 a month, billed yearly (Pro) | Yes | No |
Prices from each tool's own pricing page. Prices can change, so check before you buy.
Also worth knowing about
We didn't test these meeting assistants, but they come up often in roundups.
Granola
Described in roundups as bot-free with clean notes. We haven't tested it.
Avoma
A meeting assistant aimed at sales and revenue teams. We haven't tested it.
Zoom AI Companion, Microsoft Copilot, Google Gemini
Assistants built into the meeting platforms themselves, worth a look if your team uses only one. We haven't tested them.
How we tested
Every meeting assistant took notes on the same five scripted video calls between October 16 and October 22, 2026. The calls had different accents, crosstalk and background noise, and each had a transcript checked by a person, so we could compare every tool against the same answer. You can read the full set-up in how we test.
We counted word errors across 5,000 words, checked 100 speaker labels, scored each summary against a fixed checklist of 50 points such as decisions and action items, timed how many minutes notes took per hour of audio, and ran 6 checks on where data is stored, how long it's kept and whether you can opt out of model training.
Each tool then got a score out of 5, in half points, for accuracy, summaries and insights, ease of use, privacy and security, and value for money. Together they make the overall score, and the ranking follows it.
How to choose
If you send recaps to clients or rely on action items, accuracy saves checking time. Otter.ai made 291 word errors in 5,000 and caught 42 of 50 summary points; every other tool made at least 500 errors.
Fathom and tl;dv allow unlimited free recordings. Otter.ai's free plan stops at 300 minutes a month and 30 minutes per call, and Read.ai's at 5 transcripts, so heavy users will need to pay.
Some hosts and universities don't allow meeting bots. Tactiq never uses one, Otter.ai can record bot-free from its desktop app, and Fathom and tl;dv offer bot-free capture too.
Otter.ai trains its own models on de-identified data with no opt-out described, while Fathom lets you opt out. tl;dv hosts data in Europe; Fathom stores it in the US only.
Recording laws vary, and some US states need everyone's consent. Class actions have been filed against Otter.ai and Fireflies.ai, so turn off calendar auto-join and announce the notetaker.
If you don't need summaries or meeting features, a dedicated transcription tool may suit you better. Our list of AI transcription tools compares them.
Terms you'll see
- Word error rate
- The share of words a transcript gets wrong compared with a checked human transcript.
- Speaker labels
- The names a tool puts next to each part of the transcript to show who was talking.
- Notetaker bot
- An automated attendee that joins your video call to record it.
- All-party consent
- A legal rule in some places that everyone on a call must agree before it is recorded.
- SOC 2 Type 2
- An independent audit showing a company's security controls worked as described over a period of time.
Frequently asked questions
What is the best AI meeting assistant?
In our test, Otter.ai was the best by a clear margin. It scored 4 out of 5, made 291 word errors in 5,000, labelled 87 of 100 speakers correctly and caught 42 of 50 summary points. It was also the fastest. tl;dv came second with 3 out of 5 and suits teams that want EU data hosting.
What is the best free AI meeting assistant?
Fathom and tl;dv both allow unlimited recordings on their free plans, though some AI features are limited. Otter.ai's free plan gives 300 minutes a month with a 30-minute cap per call, Fireflies.ai caps storage at 400 minutes per team, Tactiq allows 10 transcripts and Read.ai 5 transcripts a month.
How much do AI meeting assistants cost?
Billed monthly, Otter.ai Pro is $16.99 a month, Fireflies.ai Pro $18, Read.ai Pro $19.75, Fathom Premium $20 and tl;dv Pro 29 EUR. Yearly billing is cheaper: Otter Pro is $8.33 a month and Tactiq Pro $8 a month, billed yearly. All six also have free plans.
Do AI notetakers train on my meetings?
Some do. Otter.ai trains its own models on de-identified data and describes no opt-out. Fathom also uses de-identified data for its own models but lets you opt out. Fireflies.ai, tl;dv and Read.ai say they don't train on customer data by default.
How did we test these AI meeting assistants?
Each tool took notes on the same five scripted video calls from October 16 to 22, 2026, with accents, crosstalk and background noise. We compared transcripts with a human-checked version, checked 100 speaker labels, scored summaries against a 50-point checklist, timed the notes and ran 6 data-handling checks.
Are AI meeting assistants legal to use?
Generally yes, but recording laws vary and some US states need everyone on the call to agree. Class actions were filed against Otter.ai in August 2025 and Fireflies.ai in December 2025 over consent and voiceprints. Tell participants you're recording and check the rules where you and they are.
The bottom line
Otter.ai takes the most accurate notes we tested; if you mainly want free, unlimited recording, start with Fathom and check its summaries yourself.
Make your case
Think a tool is missing or wrong?
If you use a tool we reviewed, or work for the company, email us with what you saw. We'll look at the evidence and update the review if it changes our findings.
Think your company belongs on this list?
Send us your product, website and the category it fits. We'll test it the same way as every other tool. It's free, and we can't promise a date or a score.
Liam Foster
If there's a strange new AI tool making its way around the internet, I've probably already bookmarked it. My favorite part is figuring out what these platforms can actually do.
Update history
- November 4, 2026: published with the results of our test, run between October 16, 2026 to October 22, 2026.