YouTube Algorithm: Why Satisfaction Now Beats Watch Time

YouTube's help pages name one objective for the recommendation system: maximize long-term viewer satisfaction. They do not describe a 2025 changeover that demoted watch time. What the documentation does say is narrower and more useful to a creator still choosing a niche: relative watch time carries more weight for short videos, absolute watch time for long ones. Your niche fixes the format length before you record anything, so it also fixes which of those two numbers your videos get judged on.

What does YouTube say the recommendation system optimizes for?

Satisfaction, stated directly and without a date attached. YouTube's performance FAQ says the search and discovery system "helps viewers find the videos they're most likely to watch and maximize long-term viewer satisfaction," and that "Our recommendation system learns from whether or not viewers are choosing to watch a video. The system learns how much of the video the viewer watches and if they're satisfied."

The recommendation system page lists the inputs behind that sentence: what viewers "choose to watch, ignore, or dismiss ('not interested' feedback)", "how long and how much of a video they watch", likes, shares, comments, survey responses, subscriptions, and the languages they consume. The same page prints two system goals — help each viewer find videos they want to watch, and maximize long-term viewer satisfaction.

Read that list again and the popular framing falls apart. Satisfaction is not a layer stacked on top of watch time in some later release. Watch time is one of the signals the system reads to estimate satisfaction, and it sits in the same list as survey answers and "not interested" dismissals.

Is watch time still a ranking input for a small channel?

Yes, and the documentation is specific about how it gets read: "Broadly speaking, relative watch time is more important for short videos and absolute watch time is more important for longer videos."

That one sentence undoes the advice most often built on top of the satisfaction story. Relative watch time is percentage viewed. Absolute watch time is minutes delivered. A 6-minute tutorial competes on how much of it people finish. A 45-minute teardown competes on minutes. Cutting the teardown to 8 minutes to chase satisfaction throws away the measure it was being judged on.

YouTube's position on length sits on the same page: "we want both short and long videos to succeed, so we encourage you to make your videos the appropriate length depending on the content." There is no documented penalty for long videos and no documented bonus for short ones. There is a documented difference in which watch-time measure applies to each.

Did YouTube announce a 2025 shift to satisfaction-weighted discovery?

No YouTube page shows it. The Creator updates changelog — the log where product and policy changes get recorded — carries no 2025 or 2026 entry about recommendations, ranking, satisfaction, or surveys, checked on 5 September 2026. The recommendation system page and the performance FAQ both describe satisfaction as a standing objective, with no changeover date anywhere on them.

A missing changelog entry is not proof that nothing changed inside the ranking systems. It proves something narrower and more practical: the specific story being sold to creators — a named 2025 or February 2026 switch, with surveys formally outweighing watch time — has no page behind it. Nobody outside YouTube can state those weights, so any instruction priced on them was invented by whoever wrote it down.

Two neighbouring claims failed this same check earlier this year. See what the 2026 Browse feed change is documented to be, and what niche commitment actually buys you. In both cases the trail ended at a creator blog quoting another creator blog.

How should you read your own retention numbers?

Against your own channel, not against a benchmark table. YouTube does not publish retention averages by niche. The percentages that circulate — vlogs at 21.5%, how-to at 42.1%, more than half of viewers leaving in the first minute — come from third-party samples of channels that are not yours, and the report page most often cited for that set returns a 404 as of 5 September 2026.

YouTube does publish one range, for impressions click-through rate, and it publishes the caveat alongside it: "Half of all channels and videos on YouTube have an impressions CTR that can range between 2% and 10%," followed immediately by "New videos or channels (like those less than a week old), or videos with fewer than 100 views can see an even wider range." A 1.8% CTR on a three-day-old video with 60 views sits inside that wider range. It says nothing about your niche.

The performance FAQ also names why views fall, and neither reason is a length problem: "Your audience is watching more of other videos and channels on YouTube," and "Your audience is spending less time on YouTube." Both point outward, at what else your audience has to pick from. That is a competition question about the niche, not an editing question about the video.

What does this change about picking a niche?

It moves one decision earlier than most creators make it. The format length a niche runs on is set by the topic, not by preference: unboxing runs short, restoration and long-play commentary run long, tutorials sit in between. That choice fixes which watch-time measure you compete on for every upload after it, so it belongs in the niche decision rather than in the edit.

Your niche's native format	Measure that carries	What to track in Studio	What it does not tell you
Short explainers, 3–8 min	Relative watch time (% viewed)	% viewed trend across your last 10–20 uploads	Nothing about minutes delivered — a long-form channel's numbers are not your bar
Long-form, 20 min and up	Absolute watch time (minutes)	Average view duration in minutes	A 35% retention figure at this length is not a failing grade
Mixed lengths on one channel	Both, per video	The two measures kept separate	A blended channel average hides which format is working
Shorts-led	Relative watch time	Replays and % viewed	The long-form revenue math — that path is priced differently

One length rule still comes from the ad system rather than the ranking system: mid-roll breaks require a video of 8 minutes or longer. Stretching a 6-minute idea across that line is a trade, not an optimization, and YouTube's own guidance runs against it — ad slots placed at natural breakpoints "are more likely to serve ads as our systems generally find higher viewer retention at these breakpoints." Padding lowers the retention that decides whether the break you added ever fills.

Picking the format band is the easy half. The harder half is whether the niche has room left: how much demand is already served, by how many channels, and whether anything in it is still breaking out. That is the part gleam scores from YouTube data before you commit — demand against competition, plus the outlier videos that show a niche is still moving. The low-competition walkthrough covers the search side of that check, and the CPM-by-niche piece covers what each format band pays once you are in it.

The takeaway is smaller than the headlines and more usable: satisfaction is what YouTube says it optimizes for, watch time is one of the readings it takes, and the reading that counts depends on how long your niche's videos are supposed to be. Pick the niche knowing that, and the retention number you chase is the right one from your first upload.