AI meeting notetaker

An AI meeting notetaker (also called an AI notetaker or AI meeting assistant) is software that captures a meeting’s audio, transcribes it, and automatically produces notes — typically a summary, key decisions, and action items — so participants don’t have to write them by hand. Otter, Fireflies, Fathom, Granola, and Canary are all AI meeting notetakers; they differ less in what they produce than in how they capture the call and when you get the result.

What every AI notetaker does

Under the hood, nearly every tool in the category runs the same four-stage pipeline: capture the audio, transcribe it with streaming transcription, attribute it to speakers via diarization or platform labels, and summarize the transcript into notes and action items. The full walkthrough is in how AI meeting notetakers work.

Because the pipeline is shared, the outputs look similar across vendors. The real differences are on two axes.

Axis 1: how it captures the call

There are four capture methods, and every notetaker uses one of them:

• A meeting bot — an automated participant that joins through the meeting link and sits in the attendee list (Otter, Fireflies, Read AI, tl;dv, Circleback). Works from any device, but everyone sees it, and hosts and admins can block it.
• Platform-native AI — the meeting platform’s own feature (Zoom AI Companion, Teams intelligent recap, Slack’s huddle notes). No extra guest, but it only works on that platform and is usually gated by plan, admin, and host settings.
• A caption-scraping browser extension — reads the platform’s live captions inside your browser tab (Tactiq is the best-known). Lightweight, but tied to the browser and to platforms that expose captions.
• System audio capture — a desktop app reads the sound your computer is already playing, so it works across apps with no bot, no plugin, and no virtual device (Granola, Jamie, Canary). This is the bot-free branch.

Axis 2: when you get the notes

Most notetakers are post-meeting tools: they record during the call and deliver the document after it ends. A smaller group works during the meeting — a live transcript in a few cases, and in Canary’s case real-time summarization: a live, multi-resolution rolling summary you can glance at mid-call. The distinction is covered in post-meeting vs in-meeting AI notes, and it decides what the tool is for — a record to read later versus awareness while the conversation is still happening.

What an AI notetaker is not

Two common conflations are worth untangling. A notetaker is not the same thing as its bot — the bot is one capture method, not a synonym for the category, which is why “bot-free AI notetaker” isn’t a contradiction. And it’s more than a transcription app: pure transcription gives you a verbatim record, while a notetaker’s defining output is the condensed note layer on top.

The honest caveats

The category’s limits apply to every capture method. Notes are approximately right, not court-record right — summarization can omit or flatten what mattered. Virtually all production notetakers process audio in the vendor’s cloud, which is the heart of the safety question. And no capture method removes the courtesy of telling participants you’re taking AI notes — recording-consent law varies by region and doesn’t care which architecture you picked.

Canary sits in the corner of the two axes that the rest of the category leaves empty: bot-free and during-the-call. It captures your computer’s system audio (no bot, no plugin) and shows a live, multi-resolution rolling summary — from what’s being said right now to the whole call — so you can catch up the instant your name is called. To see how that differs from a classic bot-based notetaker, start with Canary vs Otter; for the full landscape, see the complete guide to bot-free meeting notes.