YouTube Killed Watch Time. Here's What Replaced It 👀
For years, the YouTube algorithm had one job. Keep people watching as long as possible. Watch time was the signal that everything else fed into. Build a long video, hold attention, and get recommended. That was the game. That game has changed, and most creators haven't caught up yet. YouTube has been running post-view surveys for years, quietly asking viewers to rate videos after watching. One to five stars. Did this feel worth your time? They collected millions of those responses and trained a machine learning model on them. That model now predicts a satisfaction score for every video on the platform, whether it was ever directly surveyed or not. The shift matters because watch time and satisfaction are not the same thing. A viewer can watch 18 minutes of a 20-minute video out of inertia and register low satisfaction. Another viewer can watch 4 minutes, get exactly what they came for, and immediately search your channel name for another video. That second viewer is worth more to your channel now than the first one. YouTube calls that post-video behaviour return to discovery. When someone finishes your video and searches for more of your content rather than closing the app, your satisfaction score goes up significantly. When someone closes the app, it is neutral to slightly negative. When someone clicks not interested, it is one of the heaviest negative signals in the model. The comment section is also being read. YouTube uses natural language processing to scan comment sentiment. A high ratio of comments expressing genuine value against comments expressing frustration or disappointment feeds directly into the satisfaction calculation. Responding to your first wave of comments matters not just for engagement but because those viewers tend to stay on YouTube longer after watching your video, which strengthens your session signal. Three practical changes worth making now. Deliver on your title promise in the first 30 seconds. Not at the eight-minute mark after a long setup. Viewers who get what they came for quickly are more likely to rewatch, share, and return. All three are positive signals.