You click "Not interested." The video disappears. A different video from the same category shows up in its place.
You tell YouTube you don't want to see a channel anymore. Three days later it's back. You haven't changed anything. You don't know why this keeps happening.
The algorithm isn't ignoring you. It's working exactly as designed — and once you understand what it's optimizing for, the behavior makes complete sense.
What the algorithm actually optimizes for
YouTube's recommendation system has one job: predict what you'll watch next, and recommend that. Not what you'd like to watch. Not what would be good for you. What it predicts you'll actually click and watch.
It makes those predictions based on your watch history — specifically, what you watched to completion, what you rewatched, and what content users similar to you watch. A single "Not interested" click is a weak negative signal. A week of watching similar content is a strong positive signal. The weak signal often gets overridden by the strong one.
This is why you can tell YouTube "not interested" in gaming content and still see gaming content a week later — because somewhere in your history there's evidence that you watched gaming content, and that evidence outweighs the one signal you sent.
The context problem
The deepest issue is that the algorithm has no concept of context. It doesn't know that you watched cooking videos because you were planning a dinner party and you don't actually care about cooking. It doesn't know that you watched news content during a stressful period and you're done with that now. It doesn't know that the gaming content was from your nephew using your account.
It knows what was watched. It predicts that if you watched it before, you might watch it again. Context is invisible to the system.
What actually changes your recommendations
Consistent positive signals over weeks
2–4 weeksActively seeking out and fully watching the content you actually want changes recommendations more than any negative feedback. The algorithm responds to strong positive signals faster than strong negative ones.
Clearing watch history
Immediate, then 2–3 weeks to rebuildRemoves all historical signal. Your homepage goes generic until you rebuild it. Useful if your history is mostly content you don't want — not useful if it's a mix.
Consistent 'Not interested' and 'Don't recommend channel'
3–6 weeks of consistent effortWorks, but requires sustained effort. Mark every video in the unwanted category every time you see it for several weeks. One click doesn't move the needle.
Filtering what you see by context
ImmediateA browser extension that filters your homepage and sidebar to content matching specific keywords removes the problem from your view — even if the algorithm still wants to show you other things.
Filtering instead of training
ContextTube is a free Chrome extension that adds Work, Hobby, and All modes to YouTube. In Work Mode, your homepage and sidebar filter to tech, programming, science, and business content — regardless of what the algorithm wants to show you. Switch to Hobby Mode for personal content, or All Mode to remove the filter.
You stop fighting the algorithm and start filtering what you see. It takes 30 seconds to set up and you don't think about it again.
Free · Chrome Extension · No account
Try ContextTube
Filter YouTube recommendations by context. Stop fighting the algorithm — filter around it.
See how it works →Frequently asked questions
Why does YouTube keep recommending the same type of video?
YouTube's algorithm learns from your behavior, not your intentions. If you've watched similar content in the past — even once, even reluctantly — the algorithm treats that as a signal of interest. It doesn't know you watched something by accident or out of curiosity rather than genuine preference.
Does clicking 'Not interested' on YouTube actually work?
It works, but slowly. A single 'Not interested' click is a weak signal compared to weeks of watch history showing engagement with that content. You'd need to mark multiple videos in the same category as uninteresting before you see a noticeable change. It works better as a maintenance tool than as a reset.
Why does YouTube recommend videos from channels I don't follow?
YouTube's homepage recommendations are mostly from channels you don't subscribe to. The algorithm recommends based on what users with similar watch histories watched — not just your own subscriptions. This is by design: Udemy wants to introduce you to content you don't know you'd like. The side effect is that it surfaces a lot of things you actively don't want.
Why does YouTube recommend old videos I've already watched?
If a video has high engagement in your watch history (you watched it fully, rewatched it, shared it), YouTube may recommend it again or recommend similar older content. The algorithm optimizes for predicted watch time, not novelty. If it predicts you'll watch something, it recommends it.
Is there a way to tell YouTube what I actually want to see?
Indirectly. Actively seeking out content in the categories you want, watching it to completion, and using 'Not interested' on categories you don't want gradually shifts your recommendations. The process takes weeks of consistent signals. A faster workaround is using a browser extension that filters your homepage and sidebar by content category.