We measured 3,243 YouTube videos. One hundred of them were advanced.

Of the 3,243 videos we have measured, 100 were pitched above intermediate. Not a hundred per subject. A hundred in total, across twenty subjects and roughly a thousand hours of runtime.

If you have ever searched for the next thing after the basics and found forty more introductions, this is the shape of what you were searching. The material is not hidden behind bad search terms. In most subjects it has not been made.

A note on the numbers. Unlike the rest of this site, the figures on this page are frozen at the date above rather than recalculated on each visit. A report people cite should not move under them. When we run this census again it will be published as a new edition with a new date.

The library divides almost in half, and then falls off a cliff

Beginner and intermediate material split the library almost evenly — 1,616 against 1,527, or 49.8% and 47.1%. What is left is the whole of the advanced shelf.

Three levels sounds like three tiers. It is two tiers and a rounding error.

This is not a judgement about the videos. A beginner explanation is the correct thing to make for somebody at the beginning, and half a library of them is what a platform serving newcomers should look like. The finding is about what happens next, to the person who is no longer at the beginning: as far as our measurement goes, the shelf they are standing in front of is nearly empty.

Five subjects have no advanced material at all

Of the twenty subjects we measure, five contain no advanced video whatsoever — productivity, design, history, comedy and travel — and that is across 739 measured videos between them.

Look at the axis before the bars. It runs to 16%, not to 100%, because at true scale every bar on this chart would be a smear.

Gaming sits at the top of the chart on a subject of 64 videos, which is too few to call a rate; read it as ten long technical breakdowns rather than as a property of gaming.

The subjects that do carry advanced material are the ones where the next level is a different body of knowledge rather than a harder version of the same one. Machine learning has 30 of them out of 213, the largest count anywhere in the library. Engineering, science and programming follow. In each of those, the material after the introduction is a genuinely separate subject with its own vocabulary, and a channel that makes it is making something a beginner cannot substitute for.

Productivity, design and history are the other case. The next level in those subjects is not more vocabulary — it is judgement, and judgement is difficult to film. So the videos keep re-explaining the entry point, and 600 measured videos across those three subjects produced no advanced material at all.

Business is the sharpest single figure here: one advanced video in 260 measured. It is the largest subject in our library and the one where the ratio of introductions to everything else is most extreme.

Runtime finds advanced material better than any word in a title

Here is the part that is useful on a Tuesday afternoon. Advanced material is not spread evenly across the shelf — it collects at one end of the runtime scale, and the gradient is steep.

Under ten minutes, 5 videos out of 1,184 were advanced. That is four in a thousand. Over thirty minutes, the share is around one in eleven, and it stays there for every band above it.

The mechanism is not mysterious once you see it. Advanced material assumes context, and assuming context lets a video go deeper rather than faster — so it goes longer. The mean advanced video in our library runs 35.7 minutes against 13.4 for the mean beginner one. Length is not a mark of quality, and we have measured elsewhere that it is a poor filter for most things. For this one question it is the best signal available to a person standing in front of a search results page.

The advanced material is also the least wasteful

Advanced videos were the least likely of the three levels to be a waste of time, by a distance.

6.0% of advanced videos were not worth their runtime, against 28.8% of beginner ones. Padding runs ten points lower, and the useful part starts a minute later — which is the same fact seen from the other side: an advanced video spends its opening establishing where it is picking up from, and that opening is doing work.

We are not claiming that difficulty causes quality. The likelier reading is that both come from the same decision. A channel making something for the fraction of its audience that already knows the basics has given up the easy audience, and it is unlikely to have done that carelessly. The beginner shelf is crowded, and a crowded shelf contains everything, including the things made because the shelf existed.

Why the shelf is shaped like this

Every video on YouTube is competing for the same attention against the same recommendation system, and the recommendation system rewards the video that holds the most people.

An introduction can be watched by everybody who has not learned the thing. The video after it can be watched by the fraction who did. That fraction shrinks at every step, and it shrinks fastest in subjects where the steps are small — which is exactly the pattern in the chart above. The subjects with advanced material are the ones where step two is a big enough jump to be worth somebody’s year.

None of this makes YouTube the wrong place to learn. It makes it a place with a particular shape, and knowing the shape is the difference between concluding that you are bad at searching and concluding that you are looking in a subject where the material stops.

What to do with this

Three things follow from the measurement, in order of how much they help.

Sort by length, not by relevance. Under ten minutes the advanced material is essentially absent. If you already have the basics, a thirty-minute result is roughly twenty times more likely to be pitched at you than a nine-minute one.

Check whether the subject has the material before blaming the search. If you are looking for advanced productivity or advanced design video, our measurement says there was none in 446 videos. That is a fact about the subject, and no search term fixes it.

Read the opening minute as a level test, not as an introduction. The advanced videos in our library take a full 1:43 before the useful part starts, against 0:48 for beginner ones. In an advanced video that opening is usually the author saying which starting point they are assuming. If it names things you know, you are in the right place. If it explains them, you are not.

How this was measured, and what it does not say

We measure explanatory YouTube videos on several axes at once, and one of them is the level the video is pitched at: beginner, intermediate or advanced. The reading is ours. It is not the channel’s own description, not a YouTube category, and not a difficulty rating anyone else publishes. Method and its limits are in what value density means.

The sample is not YouTube. It is 3,243 English-language explanatory videos in twenty subjects, assembled to build a recommendation library rather than to be representative of the platform. It skews towards material somebody thought was worth measuring, which if anything should make the advanced share too high rather than too low.

A hundred is a small number to cut twenty ways. The subject chart is the whole advanced shelf split across twenty columns, so several of its bars rest on one or two videos. Read the top of that chart, treat the middle as an ordering rather than a set of estimates, and treat the five zeros as what they are: no advanced material found in that subject, not proof that none exists.

Padding is measured on explanatory material only. The padding column in the table above rests on 3,002 of the 3,243 videos; the rest are material where a padding share would not mean anything.

Level is not worth. A video being advanced does not make it good, and a beginner video is the right answer for a beginner. What the level measures is who the video assumes it is talking to.

The figures are frozen at 25 August 2026. Scores are recalculated as the underlying measurement improves, so the live numbers on the shelves will already have moved slightly from these. This edition will not be edited; the next census will be published separately.