YouTube recommendations: what really drives your Recommended page

If your views feel random lately, here's the uncomfortable truth: You're not "being punished." You're being profiled - by the same people you're trying to reach.

Because YouTube's Recommended page isn't some editorial desk with a mood. It's a mirror. And mirrors don't care about your upload schedule.

What happened

YouTube clarified (again, but more plainly) what drives recommendations: viewer behavior. Watch history. What people click. What they ignore. How long they stick around. And the engagement patterns around all of that. ([blog.youtube](https://blog.youtube/inside-youtube/on-youtubes-recommendation-system/))

Under the hood, YouTube describes its system as constantly learning from an ocean of "signals" (they've used the phrase over 80 billion). The short version: it uses clicks, watch time, and even satisfaction-style feedback (surveys) to estimate whether a video felt "worth it," not just whether it got a click. ([blog.youtube](https://blog.youtube/inside-youtube/on-youtubes-recommendation-system/))

Also: your audience can actively reshape what they see. If they delete videos from their watch history, those views stop influencing their future recommendations. That's not creator folklore - it's stated as a product behavior. ([blog.youtube](https://blog.youtube/news-and-events/safer-internet-day-2022-tips/))

And yes, recommendations aren't a side quest on YouTube. Google has acknowledged that the often-cited "70%" recommendation-driven figure was calculated in 2018 (and that the exact percentage can change), but the point stands: recommendations are a majority driver of watch time. ([committees.parliament.uk](https://committees.parliament.uk/writtenevidence/40785/pdf/))

Why creators should care

Attention: clicks get you a first date. Watch behavior decides if you ever see them again. YouTube literally built its modern discovery stack around the idea that clicks alone were too easy to game, so they leaned hard into watch time - and then into "was it satisfying?" signals. ([blog.youtube](https://blog.youtube/inside-youtube/on-youtubes-recommendation-system/))

Distribution: "Recommended" is personal. That means two people can watch the same video and walk away into totally different YouTube neighborhoods. You're not just competing with creators in your niche - you're competing with whatever the viewer did in the last hour. ([blog.youtube](https://blog.youtube/inside-youtube/on-youtubes-recommendation-system/))

Monetization: creators love to talk CPMs. Fine. But the real lever is repeatable discovery. If YouTube decides your video causes regret (clickbait-y, misleading, bounce-heavy), it has every incentive to stop handing you new people. They've been explicit that "responsible" goals can override raw engagement signals. ([blog.youtube](https://blog.youtube/inside-youtube/on-youtubes-recommendation-system/))

Workflow: the industry's moving toward more "why am I seeing this?" style transparency. TikTok rolled out "Why this video" explanations inside the For You feed, spelling out activity-based reasons like watches, likes, shares, searches, and follows. Different app, same philosophy: behavior drives distribution. ([newsroom.tiktok.com](https://newsroom.tiktok.com/en-us/learn-why-a-video-is-recommended-for-you/))

And the competition angle: Instagram's been tightening screws around recommendations too - especially on original content. They've said they'll replace reposts with the original in recommendations, and they've threatened to stop recommending Reels from accounts that post unoriginal content repeatedly (the widely reported threshold: 10+ in 30 days). If you're building a "clip page empire," that era's getting colder. ([engadget.com](https://www.engadget.com/instagrams-algorithm-overhaul-will-reward-original-content-and-penalize-aggregators-130018977.html))

What to do next

• Audit your "first 30 seconds" like a villain. Not for vibes. For exits. Open your retention graph, find the first cliff, and rewrite that moment until it stops bleeding.

• Design for the next click, not the current view. Build obvious sequel paths: a follow-up video, a tight end-screen promise, a pinned comment that actually earns the next watch (not "watch next!!").

• Stop optimizing for "engagement," start optimizing for "no regret." If your thumbnail/title sells one thing and the first minute delivers another, you'll get the worst kind of signal: fast clicks, faster bounces.

• Plan for "history-off" viewers. Some audiences keep watch history paused, which can thin out personalization on Home. That makes search, subscriptions, shares, and external entry points more valuable than you think - so repurpose with intent, not as an afterthought. ([androidheadlines.com](https://www.androidheadlines.com/2023/08/disable-youtube-watch-history-home-feed.html))

• Run one clean experiment per week. Change one variable (topic framing, intro style, length, structure), keep the rest steady, and let the system learn what your channel actually is. Chaos uploads create chaos recommendations.