The Hidden Cost of Free Notetakers

Updated October 3, 2026: we re-checked every policy and free-tier limit against the vendors' live pages, corrected Fathom's training policy, Otter's retention terms, Fireflies' deletion window and storage cap, and Granola's free tier, removed a claim that we tested the tools, and added sources. Details are on our corrections page.

The pitch is compelling: free meeting recording with AI summaries.

Five AI meeting notetakers offer that deal right now. Fathom, Otter, tl;dv, Fireflies, and Granola all let you record, transcribe, and summarize meetings without paying a cent. But the product economics don't work unless you're the product, or the free tier is a funnel designed to push you toward paid. Both models carry real costs. We read each tool's privacy policy, pricing page and help docs (checked October 3, 2026) to map exactly what "free" means for each one. This is desk research: we did not test the tools.

The business models behind free

These five tools monetize through three distinct models. The differences matter because each model creates a different kind of cost for the user.

Model 1: The freemium funnel. Fathom gives individual users unlimited recordings and transcripts forever. The product stays free. Revenue comes from teams: when a company wants shared meeting libraries, CRM field sync, and team analytics, Fathom charges per seat. The individual user is the foot in the door. Your data is the bait that pulls your company into a paid contract. Fathom's privacy policy does let it train its in-house AI models on de-identified meeting data, but you can opt out in account settings, and it bars third parties such as OpenAI, Anthropic and Google from training on your meetings. You're mainly a lead, and only secondarily a data source.

Model 2: Data-subsidized. Otter's privacy policy (effective June 16, 2026) lists "training our proprietary AI technology on de-identified audio recordings and on transcriptions" among its uses of your data, and we found no training opt-out in it. Enterprise customers can negotiate contract terms; free users accept the standard policy. Granola follows a softer version of this pattern: it trains on de-identified data by default, and every user can switch that off in settings.

Model 3: Feature-gated. tl;dv and Fireflies offer generous free recording but lock the features that make recordings useful behind paid walls. tl;dv's free plan gives you unlimited recordings and transcription (up to 40 web recordings a week) but includes AI notes for only 10 meetings and keeps data for up to 3 months. Fireflies' free plan gives you 400 minutes of storage per team and 20 AI credits. Both tools bet that once your archive grows, you'll pay to unlock it.

What the privacy policies say

We read the current privacy policies of all five tools (each policy's own "last updated" date appears below). Here is what each one says about data training, third-party sharing, retention, and deletion.

Based on our review of publicly available privacy policies and pricing pages, checked October 3, 2026. Policies change frequently. Verify directly with each vendor.

Fathom has one of the stronger privacy stances of the group. Its policy (last updated August 16, 2026) says it does not authorize third parties "(e.g., OpenAI, Anthropic, Google, etc)" to train on your meeting content. Fathom does use de-identified meeting data to train its own in-house models, but you can opt out in account settings. After an account deletion request, Fathom says it uses "commercially reasonable efforts" to delete recordings and personal information within 30 days. It offers a signed HIPAA BAA on its Enterprise plan, per its pricing page. One change since we first published: Superhuman acquired Fathom on September 14, 2026. Fathom told users that "nothing changes today", but future policy changes will come from a new owner.

Otter tells a different story. Its privacy policy (effective June 16, 2026) lets Otter train its proprietary AI on "de-identified audio recordings and on transcriptions (which may contain Personal Information)." It discloses data to data labeling service providers and AI service providers. On retention, Otter keeps personal information "for as long as necessary" for the purposes in the policy or as required by law, and says it renders deleted data "irrecoverable"; it gives no fixed deletion timeline. For free tier users, there's no way to negotiate stricter terms.

Otter also faces a consolidated class action, In re Otter.AI Privacy Litigation, in federal court in California. On August 13, 2026, the court granted Otter's motion to dismiss only in part, letting core wiretap and privacy claims proceed. That ruling decides only that the case continues; no court has found that Otter broke the law.

tl;dv takes a GDPR-first approach. Its privacy policy says personal data is "primarily processed and stored" in the European Economic Area, with part of its AI processing depending on your hosting choice. The policy states that tl;dv does not use customer content, including recordings and transcripts, to train "foundation models, large language models, or other generative AI models," and that its AI providers (it names Anthropic models via Google Cloud Vertex AI) do not train on it either. Its pricing page lists GDPR and SOC 2 compliance.

Fireflies' privacy policy (last updated September 28, 2026) applies a "Zero Data Retention policy" to meeting content: third-party vendors do not store it after processing, and it is not "used for training internal or external AI models." The policy also discloses identifiers and site-activity data to "marketing and advertising partners." If you close your account, Fireflies says it deletes related personal information within 30 days.

Granola is the newest entrant, and two of its defaults deserve attention.

The first is sharing. Granola notes can be shared by link. On Free and Business plans, each user sets their default link sharing to "Anyone with the link", "Only my company" or "Private", and a change applies only to notes created afterward, per Granola's sharing docs. Enterprise admins can enforce a workspace-wide setting that also applies to existing notes. For teams recording sensitive meetings (sales negotiations, HR discussions, strategy sessions), check this setting first.

The second is training. Granola's privacy policy (effective September 8, 2026) says it uses only de-identified data to train AI models, "which you can opt-out of within your Granola account settings." Every user on every plan has that per-user switch; Enterprise workspaces are opted out by default and admins can enforce it. Granola deletes audio after transcription, according to its pricing page, which limits exposure. The practical risk sits in the defaults: on Free and Business, training stays on until each person turns it off.

The feature gates that trap you

The privacy policy is only half the story. The other half is what each free tier withholds, because those gates create switching costs that lock you in over time. If you're evaluating these tools for a sales team, our breakdown of why mid-market teams are leaving Gong covers the paid alternatives.

CRM integration is the most common upsell, with one partial exception: Fathom's free plan includes basic CRM sync for up to 3 users per domain and keeps CRM field sync for its Business plan (Fathom pricing, checked October 3, 2026). The other gates vary. Fathom withholds team collaboration, CRM field sync and its advanced AI features. tl;dv withholds global transcript search, AI notes beyond 10 meetings, and retention beyond three months. Fireflies withholds downloads, integrations, and most AI features. Granola withholds CRM integrations and in-app access to meetings older than 30 days.

The trap works like this: you start recording meetings. After three months, you have an archive of 50 or 100 meetings. Your team starts relying on the searchable library. Then you hit a gate. You need CRM sync, or you need to search across all meetings, or the 3-month retention window approaches and your oldest recordings face deletion. At that point, switching to a competitor means losing your archive. The cost of staying (paying for the upgrade) is lower than the cost of leaving (rebuilding your meeting library from scratch). That's the business model.

The data you can't take with you

Export capability determines whether you actually own your meeting data or just rent access to it. Here's what each free tier lets you take with you if you decide to leave.

Fathom lists unlimited call downloads and clips on its pricing page and offers a public API. The page does not spell out which export formats the free plan includes, so check before you build a large archive. For individual meetings, downloading works; for hundreds of meetings, plan on the API.

Otter allows free users to export transcripts as plain text files and audio as MP3, per its pricing page. But Otter only keeps your 25 most recent conversations on the free tier. Older meetings are automatically removed. If you recorded 100 meetings over six months, 75 of those transcripts are already gone by the time you decide to leave. You can only export what still exists.

tl;dv keeps free-plan data for up to three months. Its pricing page does not make clear which recording downloads and transcript exports the free plan includes, so confirm before you depend on it. The 3-month retention window adds pressure: export what you need before it disappears, or upgrade to keep it.

Fireflies lists "Download transcripts, summaries, recordings" as a Pro feature on its pricing page, so the free tier locks both transcript and audio downloads. You can view transcripts and listen to recordings inside the app, but you cannot download any of it. Your data stays in Fireflies. If you leave, you leave empty-handed.

Granola is the friendliest for export on the free tier. Its CSV export includes note titles, summaries and transcripts for your full history, including notes older than 30 days that the free app no longer shows. There's no audio to export, because Granola deletes audio after transcription. The catch: the free plan shows only 30 days of meeting history in the app, so it works more as a trial than a permanent free tier.

Which free tier is actually free

After reviewing the policies, the pricing-page feature gates, and the documented export options, here's how these free tiers rank.

Best free tier overall: Fathom. De-identified in-house model training with an opt-out, no third-party training on your meetings, and a HIPAA BAA on Enterprise. Fireflies is stricter on training: its policy bars training any AI model on meeting content; tl;dv bars training generative models. The free tier is unlimited for individual recording and transcription. If you only need personal meeting notes and don't need team features, Fathom's free tier is a real product, not a trap.

Best for EU compliance: tl;dv. EU data residency, GDPR compliant by design, and no generative-model training on customer data. The free tier is generous for recording and transcription, though data is kept only three months on the free plan and the pricing page does not make clear what it lets you export. If GDPR compliance is non-negotiable and you can live with that window or pay to keep data longer, tl;dv is the strongest choice.

Use with caution: Fireflies. The no-training policy is good. The zero data retention with third-party vendors is good. But the free tier is so restricted (no downloads, no integrations, 400 minutes of storage per team) that it functions as a demo. The policy's disclosure of identifiers to marketing and advertising partners also deserves a look.

Approach carefully: Granola. The 30-day history window, link-based sharing defaults, and AI training that stays on until you switch it off require active management. Check your default link-sharing setting at signup and opt out of training in your account settings if you want to. It's a capable tool, but you have to set the defaults yourself.

Most risk: Otter Free. Otter's free tier combines training on de-identified recordings with no opt-out in the policy, the most limited history (25 conversations), and the least user control over data practices. The class action, whose core claims survived a motion to dismiss in August 2026, adds uncertainty. If you use Otter Free, assume your recordings will be used for model training, and read the policy's retention section. For sensitive meetings (sales calls, HR conversations, legal discussions), that trade-off may be unacceptable.

The bottom line: "free" is never free. Every tool in this list extracts value from free users, whether through data access, feature gates, or upsell pressure. The question is which trade-off you can live with. For most individual users, Fathom offers the best balance of privacy and functionality. For teams in Europe, tl;dv's GDPR-first approach earns the recommendation. For everyone else, read the privacy policy before you click "join meeting."