How to Learn From YouTube: Study With Videos Worth Your Time

In a 2018 Pew survey of US adults, half of YouTube’s users said it was very important for figuring out how to do things they had never done before ¹. For a lot of people YouTube is already a school, and learning from YouTube rewards the people who treat a video as material to work with. Pressing play and watching to the end is the easiest way to use it, and in the research it is also the least effective of the ways that have been compared.

This guide is the method that research supports, in the order you would use it, with the studies behind each step. The studies are mostly on lecture video, in labs and classrooms, not on YouTube itself. Where we add numbers about YouTube, they come from Gistil’s own measurement of explanatory videos.

Why watching feels like learning when it is not

In one experiment, people watched one of two short videos of the same lecture. In one the instructor spoke fluently, without notes, and looked at the camera. In the other the same instructor read from notes and hesitated. Viewers of the fluent version rated the instructor as better and predicted they would remember more. On the test, the two groups remembered the same amount ².

The same trap shows up everywhere in learning research. Rereading feels better than testing yourself, and testing yourself works better: in the classic study, people who reread a passage did better five minutes later, while people who were tested on it did far better a week later ³. Passive formats feel smooth because nothing asks anything of you, and that smoothness is what you mistake for understanding.

So the method below has one theme. Every step makes you do something with the video other than watch it. A widely used model of learning, ICAP, ranks engagement from passive to active to constructive to interactive, and learning rises at each step ¹⁶.

Step 1. Start with a question, not with YouTube

Open YouTube with a broad topic and the recommendations decide what you learn. Open it with a question, and you can judge each video against it.

A useful question is specific and checkable:

• not “machine learning” but “how does gradient descent decide the step size?”
• not “investing” but “what does an index fund’s expense ratio cost me over 20 years?”
• not “photography” but “when do I change aperture instead of shutter speed?”

Write the question down. You will use it in step 5, when you check whether you can answer it without the video. Goals work the same way at a larger scale: decades of research find that specific, hard goals beat “do your best” ¹⁴. “Learn statistics” is not a goal. “Be able to explain a confidence interval to a friend by Friday” is.

Step 2. Pick a video that is worth the time

Most of the cost of learning from YouTube is spent before any learning happens, on videos that turn out to be weak. Gistil’s measurement of explanatory YouTube videos gives a few rules that hold across the library (the full study):

• Subject matters more than length. 67% of science explainers reach gold, a score of 70 or more, against 29% in the weakest subject we report (design & ux).
• The title is evidence. Videos whose titles overpromise (honesty below 60) were judged not worth their runtime 49% of the time. Videos with honest titles (90 and above), 2%.
• Length is a weak signal. Videos over two hours reach gold most often (61%). Under ten minutes, 18% are not worth the time they ask for.
• Advanced beats beginner for depth, if you can follow it: 70% of advanced videos reach gold, against 37% of beginner ones.

A good channel helps, but less than people think. Even the best channels have weak videos, so pick the video, not only the channel. Our ranking of learning channels by subject is a good place to start.

Step 3. Skip to the part that answers your question

The useful part of a video is rarely at the start. Across the videos we measured, only 24% of first tenths fall in the part worth playing, against 68% to 70% in the middle.

The median video reaches its first segment worth playing at about two minutes. Chapters, if the creator added them, tell you what each section is about. They do not tell you which one answers your question, so read them against the question from step 1.

Step 4. Watch in segments, and pause on purpose

People learn better when a video comes in short, meaningful segments they move through at their own pace, rather than as one continuous run. A meta-analysis of 56 studies found that effect on both memory and the ability to apply what was learned, with lower mental load ⁶.

You can create the segments yourself. Pause at the end of each idea, and do one of these before you continue. As a rule of thumb, which is our suggestion rather than a tested number, every five to ten minutes is enough; the lecture study in step 5 used segments of about five minutes ⁵.

• Predict what the next step will be.
• Explain the last step out loud in one or two sentences, as if to a friend.
• Ask what would break if one assumption changed.

Video length is part of this. In 6.9 million viewing sessions on edX, median engagement topped out at about six minutes, however long the video was ⁸. That is a reason to break a long video up, not a reason to avoid it: in our measurement the longer explainers carry more useful minutes, not fewer. Attention is also less fragile than the popular “8-second attention span” suggests; that figure has no research behind it. Reviews of lecture research find no fixed 10-to-15-minute limit either, and who is teaching, and how, matters more ¹³.

What appears on screen matters too. Videos that highlight what is important, with arrows, color or headings, help learners remember more (g = 0.53 for retention across 103 studies) ¹¹. When the video does not do this, your pause does it for you.

Step 5. Test yourself, then check

This is the step that does the most and is skipped most often.

When you finish a segment or a video, close it and try to answer your question from step 1 from memory. Write the answer down or say it. Then open the video again and check what you missed. Across the testing literature, self-testing beats rereading with a medium effect, and the effect is larger when you check your answers ⁴. It holds in real classrooms too, not only in labs ¹⁵.

The most striking result is specific to video. In an online statistics lecture, one group answered a few short questions after each segment. Another group reviewed the material, and a third simply kept watching. The tested group’s minds wandered 19% of the time, against 39% and 41%. They took more notes, and they scored 86% on the final segment against 66% to 72% ⁵.

Videos with built-in questions that pause and wait for an answer show the same advantage, g = 0.52 across 22 comparisons. The pause is part of it: a question on screen while the video keeps playing splits attention and hurts ¹².

Step 6. Come back a few days later

Learning one thing twice in one evening is less effective than learning it once today and once a few days from now. The question is how long to wait. In a study of more than thirteen hundred learners, the best gap between two study sessions was about 20% to 40% of the time until the test, when the test was a week away, and closer to 5% to 10% when it was a year away ⁷.

In practice:

• for an exam next week, review after one or two days;
• for something you want to know in a year, review after three to five weeks;
• each time, test first (step 5), then rewatch only the parts you got wrong.

Rewatching is where speed helps. In one study, rewatching a lecture at 2x right before the test beat watching it once at normal speed a week earlier ⁹.

Step 7. Turn videos into progress, not a pile

The opposite of learning from YouTube is a Watch Later list that only grows. Two habits keep it under control:

• Save with a question attached. If you cannot say which question a video answers, do not save it.
• Count what you finish against a goal, in hours or in questions answered, not in videos opened.

Self-regulated learners do better in online courses, and the habits that matter most in the research are time management, monitoring your own understanding and keeping up effort, not rehearsing material over and over ¹⁷.

Three myths worth dropping

“Take notes by hand.” The famous study showing longhand beat laptop notes did not hold up when repeated directly: the laptop group did as well on the quiz ¹⁰. What research on generative learning does support is the kind of note: summarize ideas in your own words and test yourself, rather than copying what is said word for word ¹⁹.

“Watch at 2x to learn faster.” Up to 2x, younger adults lose little comprehension; past it, they lose more ⁹. Older adults lose more at faster speeds ¹⁸. And speed compresses everything evenly, useful and padding alike. Skipping to the useful part saves more time than speeding through the rest. We looked at this in detail in is 2x speed helping you learn?

“Short videos are better for learning.” Short videos hold attention more easily ⁸, but in our measurement they are the least likely to be worth their time as lessons. Watch the longer, better video in segments.

Where Gistil fits

Gistil helps with the parts of this method that are about choosing and finding. It does not quiz you or schedule your reviews. That part stays yours, and it is the part that does the most.

• In Chrome and Firefox, next to a YouTube video it has measured, or one you ask it to analyze, Gistil shows whether it is worth watching (gold, silver or copper), the key moments, and a map of where the useful minutes are. One button, Play the useful part, plays only those. That helps with steps 2 and 3.
• In the Gistil app for iPhone and Android, you set a learning goal in hours and get feeds of measured videos on your topics, with no copper videos in them. Each video can start at a useful moment, and what you watch counts toward the goal. That helps with steps 2, 3 and 7.

Videos already measured open free, with no account. To try the method, pick one question this week: write it down, choose one video, pause three times, and answer the question from memory at the end.