How Long Should a YouTube Video Be? The Question Is Wrong

Search this question and you get numbers: seven to fifteen minutes, over ten for mid-rolls, under six for retention, twenty-two because that is what the big channels do. All of them are answers to a question that does not have one, and following any of them will make a specific video worse.

Length is not an input. It is an output — of how much genuinely interesting material the topic contains and how densely you are willing to cut it.

Why there is no optimal number

The reasoning behind "longer videos rank better" runs: YouTube optimises for watch time, longer videos produce more watch time per view, therefore make longer videos. Every step is true and the conclusion is still wrong, because it ignores what happens to the rate.

A twenty-minute video that people leave at four minutes produces less watch time than a six-minute video people finish, and it teaches the system something worse: that your videos are not worth staying for. That signal follows the channel, not the video.

The evidence people cite for long videos is selection bias in plain sight. Large channels make long videos, and large channels do well. But they earned the licence to be long by first proving they were worth watching. Copying the runtime without the licence copies the wrong half.

Where the eight-minute rule actually came from

Not the algorithm. Ad inventory.

Videos past a certain length can carry mid-roll ads, which raises revenue per view. That is a real economic effect and it is why the number became folklore — creators noticed longer videos earning more and concluded the system was rewarding length.

The trap is that padding a six-minute idea to clear the bar costs you retention across the whole video, and the retention damage suppresses impressions. You gain an ad slot on a video shown to fewer people. That trade is usually negative, and it gets worse as your channel grows.

If a topic naturally runs long, take the mid-rolls. Do not manufacture length to reach them.

What actually decides length

How much the topic contains. Some questions are answered in four minutes. Stretching one of those produces the most common bad video on YouTube: a good idea surrounded by filler. Others cannot be done justice in thirty. The material decides.

Your density. Two creators can cover the same subject in nine minutes and twenty-five, and both be right, because they cut differently. Density is a style choice, and it should be a consistent one — it is a large part of what an audience is actually subscribing to.

Where the audience is. A viewer who arrived from search on a specific question wants the answer, not an essay. A viewer who arrived from browse, from a channel they already like, has different patience. Search-driven topics run shorter than they feel like they should.

Format expectations. A tutorial, a video essay, a news explainer and a review carry different implicit contracts about length. Breaking that contract needs a reason.

The test that replaces the number

Instead of asking how long the video should be, ask this about every section:

If I cut this entirely, does the video get worse?

Not "is this section fine". Not "did I work hard on it". Does removing it damage the result. Anything that survives that question belongs. Anything that does not is padding, regardless of how long the video ends up.

Run it honestly and length answers itself. It also tends to produce videos shorter than you planned and better than you expected, which is uncomfortable the first few times.

Reading your own retention graph

Your analytics already contain the answer for your channel, and it is more reliable than any benchmark.

• A cliff in the first thirty seconds is a packaging problem, not a length problem. The title and thumbnail promised something the opening did not deliver.
• A steady, gentle decline is normal and healthy. Do not try to fix it.
• A steep drop at a specific timestamp is a section that should have been cut. Go and watch it. That is the single most useful thing in YouTube Studio.
• A drop that arrives consistently at the same proportion across videos — say, two-thirds of the way through, regardless of runtime — usually means your structure runs out of turns before it runs out of runtime, and making videos shorter will help more than making them denser.
• Videos that hold to the end are permission to try something longer next time.

Compare your videos against each other, not against a benchmark. Absolute average view duration is nearly meaningless across channels; the shape of your own curve over your last ten uploads is not.

Practical starting points

If you want numbers to start from — knowing they are starting points and not targets:

• Answering a specific search query: as short as the answer honestly takes. Four to eight minutes covers most.
• Tutorial or walkthrough: as long as the task, with chapters so people can skip to their part. Chapters matter more than length here.
• Explainer or video essay: ten to twenty, if the research supports it. Under-researched long videos are the worst failure mode on YouTube.
• Commentary on something current: shorter than you think. The value decays quickly and so does the patience.

Then let the cutting test override every one of these.

The structural fix for videos that need to be long

If a topic genuinely runs long, the problem is not length — it is that attention has to be re-earned repeatedly. What holds a twenty-minute video together is turns: points where the video changes direction, complicates its own claim, or answers a question it raised earlier.

Roughly every two to three minutes, something should change. A new question, a reversal, a consequence you did not see coming. A long video that holds is not one that is interesting for twenty minutes straight; it is one that gives the viewer a reason to stay about eight separate times.

If you cannot find those turns in your outline, the video is not long. It is padded, and you have just discovered it before publishing rather than after.

The short answer

Make the video as long as it is interesting, and not one section longer. Then look at where people actually leave and cut that part next time.

Every real improvement in your average view duration will come from that loop, and none of it will come from picking a better number.