- Meeting notes
- Privacy
- Comparison
AI note takers that don't record: transcribing vs recording, and why it matters
People searching for a note taker that doesn't record usually mean one of three things: no audio file kept, no bot in the call, or nothing leaving the device. Each is a different property, most tools have one and not the others, and none of them lowers the consent bar. Here is how to tell which one you need.
Ben Claybrook 5 min read
“AI note taker that doesn’t record” is one of the more revealing searches in this category, because the people typing it mean at least three different things, and the tools that show up in the results each satisfy one of them while implying the others.
Sort out which one you mean and the shortlist gets short.
The three things “doesn’t record” means
No audio file is kept. The tool listens, produces text, and does not retain the sound. The transcript exists; the recording does not. This is what someone means when the worry is “I don’t want a recording of my voice sitting somewhere”.
No bot joins the call. Nothing appears in the participant list and no recording banner lights up in Zoom or Teams. This is what someone means when the worry is how it looks to a client or a candidate. We covered this one on its own in AI note takers that don’t join your meeting.
Nothing leaves the device. No audio, no transcript, no notes go to any server. This is what someone means when the worry is where the content ends up and who can be compelled to hand it over.
These are independent. A tool can keep no audio and still send every word to a cloud model. A tool can have no bot and still store the recording. And a tool can keep everything on your device and still, technically, record.
Recording versus transcribing, as the platforms define it
The meeting platforms draw a clear line between the two, and it is worth knowing because it is the line the consent notices follow.
Teams separates them completely. An organiser can start a live transcription with no recording; all participants see a transcription notice, and the text goes to the organiser’s OneDrive. Recording is a separate switch that adds a video file. Details in how to transcribe a Teams meeting without recording it.
Zoom has both. Its audio transcription is a feature of cloud recordings and needs one. Its in-meeting Meeting transcript runs live with no recording, on paid accounts where an admin has enabled it, and saves to the host’s own computer. Details in how to transcribe a Zoom meeting without recording it.
Google Meet likewise: transcripts do not require a recording, are saved to the organiser’s Drive, and participants get an on-screen warning either way.
The lesson is that “transcribe without recording” is normal, documented behaviour, not a loophole. The platforms still notify everyone, because the text is the thing that matters.
What each kind of tool actually keeps
Sourced from each vendor’s own pages, with the caveat that policies move:
| Audio kept? | Transcript kept? | Where | |
|---|---|---|---|
| Otter, Fireflies, Fathom (bots) | Yes | Yes | Vendor’s servers |
| Granola (device capture) | No, per Granola | Yes | Granola’s cloud; audio processed by third-party providers |
| Teams live transcription | No | Yes | Organiser’s OneDrive |
| Zoom in-meeting transcript | No | Yes | Host’s computer, and Zoom’s Recording and Transcripts tab |
| Private Notetaker | On your disk only | On your disk only | Your computer, nothing elsewhere |
Granola is the interesting row. Its security page states that it does not store your meeting audio, which satisfies the first meaning of “doesn’t record” honestly, while the audio is still processed by transcription and AI providers in the cloud before being discarded. Our write-up of Granola’s own documentation has the specifics. It is a good design; it is just not the third meaning.
Does not recording change the law?
No, and this is the part the search term seems to hope for.
Consent law in the all-party states (California, Florida, Illinois, Maryland, Massachusetts, Pennsylvania, Washington and others) and in most of Europe attaches to capturing the conversation, and a transcript is a capture of the conversation. A tool that produces text without an audio file has not stepped outside the rule; it has produced the same record in a different format. The platforms understand this, which is why Teams and Meet notify participants of transcription exactly as they do of recording.
So whichever meaning you settled on above, the practical rule is unchanged: say it at the start. The state-by-state detail, and where the published lists disagree, is in two-party consent states, and what the lists get wrong.
Which one you need
If the worry is a recording of your voice existing somewhere: Granola-style tools and the platforms’ own transcription satisfy it. The text still lives on a server.
If the worry is how it looks in the call: any device-capture tool. Bot-free is a solved problem and you do not need to pay for on-device processing to get it.
If the worry is the content existing on a server at all: you need the third property, and it is the smallest category. Nothing about “no audio kept” or “no bot” delivers it. The test is the one we use everywhere: turn off wifi and see whether the tool still produces notes. If it does, the processing is on your machine.
Where Private Notetaker fits
Private Notetaker is the third kind. It captures the meeting on your computer, transcribes it there, and writes the notes there, so the recording, the transcript and the summary exist only on your own disk, where you can keep or delete them. No bot, no upload, no vendor copy. In the first sense of the word it does record; in the sense that matters for who else can ever read it, nobody but you.
The limits are the usual ones: English only, no shared team library, and less accuracy on bad audio than the cloud services. If your requirement is the first or second meaning, a cheaper or free tool will do, and we would rather say so.
See how it works with Zoom or with Microsoft Teams, or see pricing and try it.
Platform behaviour is from Microsoft’s, Zoom’s and Google’s own support pages and Granola’s security page, checked 16 September 2026. Nothing here is legal advice.
Frequently asked questions
- Do AI notetakers record meetings?
- Most do, in the ordinary sense: bot notetakers such as Otter, Fireflies and Fathom capture the audio and store the recording along with the transcript on their servers. Some keep the transcript but not the audio; Granola documents that it does not store your meeting audio, only transcripts. Platform features such as Teams live transcription produce text without a recording. Whether the audio file is kept and where it is kept are separate questions, and the vendor's own privacy or security page is where the answer lives.
- Are AI notetakers illegal?
- No, the software is lawful. What can be unlawful is capturing a conversation without the consent the law requires, and that applies to a transcript as much as to a recording. Federal law and most US states allow it when one party to the conversation consents; roughly a dozen states, and most of Europe, require everyone's agreement. A tool that does not record still captures, so it does not change the rule.
- Do you have to record a Teams meeting to get a transcript?
- No. Teams lets an organiser, co-organiser or permitted presenter start a live transcription on its own, with no recording. All participants see a notification, and the transcript is stored in the organiser's OneDrive for Business.
- Can I transcribe a Zoom meeting without recording it?
- Yes. Zoom's in-meeting Meeting transcript runs live without a recording, on Pro, Business and Enterprise accounts where an admin has enabled it, and saves the file to the host's computer. Zoom's other transcript, audio transcription, is a feature of cloud recordings and does require one. A local app on your own computer transcribes any Zoom call you can hear without touching either.