Zoom removed attention tracking in April 2020 over privacy concerns. It only detected whether the Zoom window was in focus for more than 30 seconds — a crude signal that could not tell distraction from note-taking. Current alternatives are post-session engagement analysis of the recording, or behavioural telemetry in the content layer for training delivered in a browser.
What the original feature actually did
Zoom's attention tracking, retired in April 2020, showed the host an indicator when a participant did not have the Zoom window in focus for more than 30 seconds.
That is the whole feature. It was a binary window-focus check with a timer.
It is worth being precise about that, because the feature is often remembered as something much more sophisticated. It did not track eye movement. It did not analyse faces. It could not tell whether you were looking at the screen. It knew one thing: whether Zoom was the frontmost application.
Why that made it bad data
The signal conflated behaviours that mean opposite things:
- Taking notes in another document → flagged as inattentive
- Looking up a reference the presenter mentioned → flagged as inattentive
- Following along in the shared deck locally → flagged as inattentive
- Staring at the Zoom window thinking about lunch → not flagged
- Leaving the room entirely with Zoom in focus → not flagged
The most engaged participant in a technical session — the one with the documentation open alongside — scored worst. The person who had walked away scored perfectly.
So when Zoom removed it under privacy pressure, the measurement community lost very little. The feature had the two failure modes that matter most: it felt like surveillance, and it was wrong.
What exists in Zoom now
No native focus or attention indicator. What you can get:
| Signal | Where | Useful? |
|---|---|---|
| Attendance and duration | Usage reports | Baseline only |
| Chat message count | Post-meeting report | Weak proxy, biased to extroverts |
| Poll responses | Poll reports | Good, but only at poll moments |
| Reaction usage | Meeting report | Weak, culturally variable |
| Camera-on rate | Observable live | Weak, and coercive to mandate |
| Q&A participation | Webinar reports | Decent for webinars |
| Recording + transcript | Cloud recording | The substantive one |
The last row is where the real opportunity is, and it is the one most teams never use.
The three legitimate alternatives
1. Post-session analysis of the recording
If you recorded the session, you already have the raw material for a far better analysis than live focus-tracking ever offered.
A recording lets you measure engagement retrospectively and per moment — who was attending during which segment, where the room collectively checked out, and what was being said at that point. Crucially it is diagnostic rather than supervisory: you are analysing the session to improve the next one, not watching people in real time.
This is what Emotuit Analyse does — per-participant engagement scoring across the timeline, aligned to the transcript, with the drop-off points flagged.
2. Behavioural telemetry in the content layer
If your training is delivered as browser content — an LMS module, a deck, a video — instrument the content rather than the conferencing tool.
This gives you what Zoom's feature could never provide: per-section resolution. Not "Sarah was distracted at some point" but "everyone disengaged during minutes 14–19, which is the escalation procedure."
Signals available: tab-switching, window focus, dwell per section, scroll velocity, interaction latency, replay behaviour, session abandonment. All from ordinary browser APIs, no camera, no biometric data. A 2024 study found tab-switching to be the strongest single predictor of disengagement in online learning — a better signal than Zoom's original feature, collected the same way, just mapped to content.
3. Designed-in interaction
The lowest-tech option and often the most effective: build moments into the session that require a response. Polls at intervals, a chat prompt every ten minutes, breakout tasks with an output. You are not measuring attention passively — you are creating observable evidence of it.
Crude, but it has the advantage of being unambiguous and completely uncontroversial.
The thing to avoid
Facial analysis of meeting participants for corporate training or education, in the EU.
Since 2 February 2025, Article 5(1)(f) of the EU AI Act has prohibited AI systems that infer emotions from biometric data in workplace and education contexts. Guidance reads "workplace" to include virtual workspaces and remote work explicitly, so a video call with your employees is squarely within it. There is no consent exception, and penalties reach 7% of global turnover.
Several tools in this space market on-device processing as the privacy answer. That is a genuine advantage for GDPR — no biometric data is transmitted. It has no bearing on Article 5, which concerns the inference, not where it is computed. A prohibited practice is prohibited on-device.
Choosing between them
| Situation | Best option |
|---|---|
| Live meeting, want real-time signal | Designed-in interaction |
| Session already recorded | Post-session analysis of the recording |
| Training delivered in a browser | Behavioural telemetry in the content |
| Webinar with large audience | Q&A and poll data, plus recording analysis |
| EU employees, any of the above | Anything except facial emotion inference |
The reframe worth taking from all this: Zoom's attention tracking answered "is this person looking at me right now?" — a supervisory question with a poor answer. The more useful question is "which part of this session lost the room?" That one is answerable, uncontroversial, and actually tells you what to change.
Frequently asked questions
Why did Zoom remove attention tracking?
Can I still see if people are paying attention in Zoom?
Is attention tracking in meetings legal?
What is the best replacement for a training use case?
Analyse a session you have already recorded
Upload a Zoom recording and get per-participant engagement, transcript-aligned drop-off points and coaching on specific moments.