AI Note Takers 2026: Bot-Free Capture Collides With Lawsuits

A crowded market where every reviewer picks a different winner

By late 2026, choosing an AI note-taker has become a crowded and contested buying decision. A wave of comparison guides has landed in the last few months. Read side by side, they show a category that has matured on transcription quality and is now competing on three other things: how the tool captures audio, how strict the free tier is, and how much legal risk a buyer takes on.

The reviews do not agree on a winner. TechRepublic names Fathom the best overall meeting note-taker, pointing to its generous free plan, its team features and its support for both bot and bot-free capture8. An evaluation of seven tools by the productivity startup alfred_ also puts Fathom first, citing a 5.0 G2 rating from more than 6,000 reviews and summaries ready about 30 seconds after a call ends2. Independent reviewers who tested the tools at length often choose Granola instead. One writer who spent 90 days with fourteen tools uses Granola for actual notes and says its summaries read as if a capable person wrote them3. Another, who says he earns affiliate commissions, calls Granola the best general assistant for people who don't want an extra participant on the call5.

Much of the rest of the coverage should be read with caution. Many of the most visible rankings come from companies that sell a competing product, and they tend to rank themselves first. Laxis names Laxis the best note-taker after what it describes as more than 200 meetings of testing15. Jamie's own test crowns Jamie16. tl;dv puts itself first among free options20. Meetily makes the case for its own open-source, local-processing tool6. The useful signal in these pieces is not the top spot but the pricing tables and the limits they document.

Accuracy is no longer the difference

The guides largely agree that transcription accuracy has stopped separating the leading products. The 90-day tester says every major tool now claims 95 to 98 percent accuracy on clean English audio, and that the gaps between Granola, Otter, Fireflies, Jamie and Fathom were too small to justify switching3. Simular's hands-on review reaches a similar conclusion: top tools score 90 to 95 percent or better in good conditions but slip to around 80 to 85 percent with heavy accents, background noise or people talking over each other7. Rock.so goes furthest and calls recording quality a commodity4.

With accuracy settled, buyers end up choosing on workflow fit. The editorial picks follow a fairly consistent pattern:

• Fireflies is usually the choice for sales teams that depend on HubSpot and Salesforce integrations12.
• Otter stands out for live captions during the call, which one reviewer describes as an accessibility feature rather than a minor extra5.
• Notion AI Meeting Notes suits teams that already work inside Notion1.
• Fellow fits teams that want agendas and action items built around the notes8.

Some analysts argue the next contest is about what happens after the meeting, not about the transcript. Simular says most tools stop at summaries and that the real gap is turning notes into action7. alfred_ makes the same pitch for drafting follow-up emails from the inbox2.

The bot question has become the main dividing line

The biggest design divide in 2026 is whether a tool joins the call as a visible participant. Meetily describes two architectures: bot-based tools that join the meeting and upload audio to the vendor's cloud, and bot-free tools that record system audio on the host's own computer6. Granola is widely credited with leading the bot-free approach, and alfred_ says it records system audio locally so other people on the call don't know it is there2.

That line is starting to blur. Otter and Fireflies now offer bot-free desktop modes alongside their meeting bots1, and Fathom added an optional bot-free mode during 20262. One reviewer notes that Fathom's version is still rolling out and varies by platform1. Simular's review explains why buyers like invisibility: clients sometimes ask why a robot is in their meeting. The same review also says bot-free capture can mean weaker speaker identification and depends on the quality of the user's microphone7.

Where the facts conflict

On concrete details, the coverage is less consistent than it looks. Granola's price is the clearest example:

• TechRepublic lists a $14-per-user Business plan8.
• Meetily lists individual plans from $18 a month and a free tier capped at 25 meetings for the life of the account6.
• Simular lists $12 a month7.
• The 90-day tester says the Basic plan now keeps only 30 days of notes3.
• Laxis describes a lifetime limit of 25 notes of history15.

Platform support is also disputed. Simular describes Granola as Mac-only7, while other guides say it runs on both Mac and Windows26. The language counts don't match either. Otter's support is listed as three languages in one guide2 and six in another4. Fireflies appears as "60+" in one place2 and over 100 elsewhere4.

The most likely explanation is that prices and tiers have changed several times this year and guides were written at different moments. Buyers should treat any single comparison table as a snapshot and check current terms with the vendor.

Free tiers are where the details matter most. Laxis's breakdown shows each vendor limits something different. Fathom records without limit but caps AI summaries at five a month. Otter allows 300 minutes a month but cuts each conversation off at 30 minutes. Fireflies gives 20 AI credits once rather than every month15. TechRepublic confirms both the Otter per-conversation cap and Fireflies' 400 minutes of shared team storage8.

The legal risk changes the picture

The privacy lawsuits are the development that most changes the market. On July 30, 2026, Florida resident Tarra Chamberlain filed a proposed class action against Granola in federal court in Northern California10. The complaint alleges that the company deliberately built its app to record calls without telling everyone on them, in violation of the California Invasion of Privacy Act's all-party consent rule10. It also claims Granola uses transcripts by default to train its models and advertises the hidden nature of its recording as a selling point10. A litigation tracker lists the case number as 3:26-cv-07926, says Granola had not yet responded as of October, and stresses that the allegations are unproven12.

Lifehacker reports that Granola denies the claims. The company says training data is anonymized and that no conversation data goes to third parties9. Computerworld adds that Granola offers two optional transparency features: an automatic chat message when transcription starts and a watermark on the user's video feed10.

Granola is not the only company being sued. Four 2025 class actions against Otter have been combined into In re Otter.AI Privacy Litigation before Judge Eumi K. Lee11. Otter's defense has three parts: its bot is visible in the participant list, its terms make account holders responsible for getting consent, and the plaintiffs can't show concrete harm11. Computerworld, citing MLex, reports that the judge was skeptical of Otter's motion to dismiss at a hearing in August but did not rule from the bench10. Fireflies faces several suits under Illinois' biometric privacy law over voiceprints used for speaker recognition9. Forbes reports that Microsoft Teams faces a similar biometric claim in Illinois14.

The risk does not stop with vendors. Lifehacker notes that note-taking companies argue the user is responsible for getting consent, so the person running the app could also be sued9. Commentary on the Otter case warns that employers who deploy these tools may share the liability11. Forbes adds that AI transcription turns casual meetings into permanent records that can be demanded as evidence in lawsuits14.

The takeaway: invisible capture is now a liability

The coverage points to an uncomfortable irony. The feature reviewers praise most in 2026 is invisible, bot-free capture, and that is the same feature plaintiffs are attacking. One tracker argues that the convenience is the liability: removing the bot also removes the notice that consent law relies on12. Otter's case shows that a visible bot is not a complete defense either11.

The safest approach for buyers is to choose a tool based on how it handles consent and data rather than on its rank in a guide. That means checking whether others on the call are told about the recording, whether training is on by default, and what is kept and for how long. Sellers of local-processing tools such as Meetily and Basil are already marketing on-device transcription as a way around these legal theories611. Those claims come from interested parties.

The lawsuits will not end the category. Google is expanding Gemini note-taking beyond Meet to Zoom and in-person conversations, according to Forbes14. What is likely to change is the default setup. Visible notice and opt-in training look set to become standard, and it will probably take the first court ruling on whether wiretap law applies to AI agents to force that change.