16 August 2026
No. A bot only joins the call because the meeting platform itself has no way to feed audio to an AI — so a third-party service has to show up as a guest instead. If the transcription runs inside the platform that's already hosting the call, there's no bot, no separate join, and nothing for anyone to remove.
Zoom, Google Meet and Microsoft Teams don't hand raw audio to outside services. So products like Otter, Fireflies and Fathom solve the access problem the only way available to them: they dial into the meeting as a participant, using a URL you give them, and capture audio the same way any human attendee would. The bot is a workaround for a platform that wasn't built to be transcribed, not a design choice anyone would make if they controlled the meeting software.
That workaround has consequences. The bot has a name, an avatar, and a seat in the participant list, because from the platform's point of view it is a guest. It has to be admitted from the waiting room in some setups. If a host or admin doesn't recognize it, they can remove it — and often do, mid-call, without knowing what they just cut off.
Native transcription means the AI runs as a feature of the meeting product itself — it has access to the audio stream because it's part of the same system, not because it asked to join. AVAY works this way: the AI participant is built into the call, not bolted on afterward, so there's no separate bot to admit, mute, or kick out.
This removes an entire category of failure. There's no waiting-room step to get wrong, no bot that silently fails to join because a firewall blocked it, no moment where someone asks 'wait, who added this to the call?' It also means the transcript can't be cut off by removing a guest, because there's no guest to remove — the recording and note-taking stop only when the meeting itself does, or when whoever is running it turns the feature off.
With a joining bot, consent is usually handled by a visible name and a notification banner — 'Otter Notetaker has joined the call' — which puts the burden on participants to notice and object before they keep talking. In two-party consent jurisdictions this matters: a bot with an obvious name at least gives people something to react to, but it's still consent by inference rather than by explicit confirmation.
Native transcription folds into the same consent flow as recording, because it's the same system doing both. If a platform requires an on-record notice or a host announcement before recording starts, the same notice covers transcription — there's no separate third party whose presence needs disclosing. The trade-off is that this consent is only as good as the platform's own recording disclosure; a platform that recorded silently before would transcribe silently now.
This is the sharpest practical difference. A joining bot's transcript lives entirely inside that bot's session with the call — remove it, and the feed stops instantly. Whatever was said in the twenty minutes after removal simply isn't captured, and there's often no visible sign in the meeting UI that notes have stopped, because the person who removed it assumed it was just another guest.
Native transcription doesn't have an equivalent single point of failure, because it isn't a participant that can be ejected. The failure modes shift instead to the platform level: if recording or transcription is turned off by the host, or the connection drops, everyone loses it — but that's a decision made about the meeting, not about a guest someone didn't recognize.
Being native solves the bot problem but doesn't solve every problem. AVAY's recordings are saved to the device that made them rather than kept centrally in the cloud, so a transcript tied to a recording is only as durable as that machine. And because the AI participant is scoped to the meeting it's running in, someone who wasn't on the call can't retroactively grant it access to a conversation that already happened — it has to be present, in the platform, when the talking happens.
| Joining bot | Native transcription | |
|---|---|---|
| Appears in participant list | Yes, as a named guest | No, it's a platform feature, not an attendee |
| Consent mechanism | Visible bot name and join notification | Same disclosure as call recording |
| Effect of being removed | Transcript stops immediately, often unnoticed | No equivalent — not removable as a guest |
| Why it exists | Platform has no built-in access to audio | Platform's AI already has access to the call |
Only if you switch to a meeting platform where transcription is built in, because Zoom and Meet don't natively expose audio to outside AI services. As long as you're using one of those platforms with a third-party notetaker, a bot join is how that service gets access — there's no bot-free way to do it on a platform that wasn't built for it.
With a joining bot, yes — it shows up as a named guest and usually triggers a join notification. With native transcription like AVAY's, no separate entry appears, because the AI isn't a guest; it's part of the meeting software itself.
The transcript stops the instant the bot's session ends, and nothing said afterward gets captured. This often happens without anyone noticing until the notes are read later and turn out to be incomplete.
It removes one specific risk — a third-party service holding a copy of your audio outside the meeting platform — but it isn't automatically more private overall. Privacy still depends on where recordings and transcripts are stored afterward, and on the platform's own consent disclosures.
Most joining bots capture audio directly through their own session with the call, then transcribe and summarize it on their own servers. That's a separate copy of the conversation living outside the meeting platform, which is worth knowing before you approve one to join.
A bot only joins your meeting because the platform can't transcribe itself — if the AI is native to the call, there's no guest to admit, disclose, or accidentally remove.
Meetings that take their own notes, in the browser: avay.ai.