The 42 (Unwritten) Rules Of YouTube

I led operations for Kwebbelkop, I've had several calls with MrBeast's own ideation team, I lead Chris Bumstead's channel, and somehow ended up talking YouTube strategy with the Dalai Lama.

This is everything I recently learned.

1. YouTube doesn't wait to judge a video by how it performs. It predicts first. If it estimates the audience for a video is small, it limits how many people see it, regardless of how good the video turns out to be.
2. YouTube also predicts your total addressable audience for a video and stops serving it once that pool looks spent, even if the video's per-impression performance is still strong.
3. Don't put your best material first. Viewers judge a video by its trajectory, not its average. Three strong segments followed by two weak ones reads as a video running out of ideas, even when the opening segments were strong.
4. Viewers trust what happened to you more than what you tell them to do. "Here's what happened when I tried this" beats "here's how to do this," every time.
5. Never judge retention by average view duration. Look at the shape of the retention graph. Look for at least five separate lessons in each graph.
6. Data can confirm an idea, but it can never produce one. Building ideas from outlier tools just remakes a generic consolidation of a topic, for an attention market that has long moved past it.
7. Low impressions alongside a strong click-through rate and strong retention don't mean the video is bad. They mean the predicted audience is small. Anchor to higher-interest themes.
8. Appear in the suggested feed of embedded creators, and you inherit part of their audience. YouTube’s discovery algorithm is lazy. It wants to put you in a bucket, so make sure it’s a bucket with existing foot traffic.
9. Keep the topic familiar. Change the angle. Roughly 70 to 80% of what makes a video feel new should come from your perspective, not the topic itself.
10. A niche isn't one audience. It's dozens of separate audiences using the same label. Aim at the whole niche and you reach none of them.
11. AI-written scripts have identifiable patterns: the negative-then-positive turn, the transition that doesn't lead anywhere, the escalation that never escalates. Viewers are subconsciously hyper-sensitive to these patterns.
12. Expect five to twenty videos with flat performance before YouTube understands your channel. That's normal. The algorithm is playing it safe. It wants to map-out how different viewers react to different types of videos you make.
13. A video with a strong click-through rate, strong retention, and low total views usually means YouTube found the right audience and stopped there deliberately.
14. Expert-level audiences are the least served on YouTube. Winning them takes roughly twenty flat videos of patience, which is why most creators give up and start talking to beginners instead.
15. A brand is a small set of things repeated on purpose. Most channels never choose those things. They happen by accident instead.
16. Not all impressions are equal. Around 20% of your impressions come from cold discovery by default. If most of those viewers stay, that's YouTube deliberately routing you to the right people.
17. Don't make a bold claim before you've earned the credibility for it. An unearned claim from an unknown creator doesn't get debated. It gets dismissed, and remembered that way.
18. A hook that promises more than the video delivers loses both audiences. Packaging needs to filter out the wrong viewer before the click, not just attract the right one.
19. Suggested and homepage aren't the same test. Suggested asks what fits after this video. Homepage asks who you already are to this viewer.
20. YouTube's own search hides underperforming videos by default. Most outlier tools only surface videos that already did well. Finding flops to falsify any video ideas takes deliberate digging.
21. Compare how new viewers watch you to how returning viewers watch you. A large gap between the two means your thumbnail promises something the video doesn't deliver.
22. Build a position around what a model can't replicate: your access, your relationships, what you've actually lived through. "I explain things clearly" isn't a differentiator any more. Most tools can now do that too.
23. A large buildup needs a large payoff. Undersell the reveal and it breaks the same trust a misleading thumbnail breaks, just later in the video.
24. Go quiet and YouTube's model of your channel starts to fade. It's a rolling average that decays without new signal.
25. Speak to your audience exactly one level ahead of where they are, never two. An idea that worked a year ago can sound stale now, because the audience has already heard it and moved on.
26. Views are a secondary measure. The goal is narrowing a video down to the exact people it's made for. Everyone else who watches adds cost without adding value.
27. Don't describe what the picture already shows. If the audio adds nothing the image hasn't already communicated, the pacing drops, and the viewer feels it immediately.
28. Study the creator who went from 3,000 subscribers to 300,000, not the one at 5 million. The larger creator wins on reputation you don't have yet. The smaller one is solving the exact problem you're facing right now.
29. A single obscure name or ticker in a video can limit its reach on its own, even when everything else in the video is simple.
30. Existing subscribers can slow down a new direction. YouTube shows new videos to your existing audience first, even when they're the wrong fit for it, and a lukewarm response from them can suppress the video before the right audience arrives.
31. Confirming something a viewer already suspects builds trust faster than teaching them something new.
32. YouTube groups viewers by behaviour, not by topic label. Two viewers who both watch technology videos can otherwise have nothing in common.
33. Every video needs a one-sentence reason to exist. Without one, it's competing against hundreds of similar videos with no advantage.
34. Saying popular advice is wrong filters your audience automatically. Only viewers who already know the popular advice will understand the claim, and they're exactly the audience you want.
35. What you actually say routes your audience as much as your title does. Word choice and pacing can send the video to the wrong crowd without you noticing.
36. "More on that in the next video" loses viewers faster than almost anything else. Pay off the idea now, or don't raise it at all.
37. YouTube doesn't reward being different for its own sake. It rewards what's already proven. A new idea only spreads once it has worked with a known audience first.
38. Watching because something interests you and watching out of boredom look identical on a watch-time chart. They aren't the same signal, even at the exact same duration.
39. A steady click-through rate is a better sign than a high one. Steady means YouTube has identified your audience with confidence. Erratic means it's still guessing.
40. The size of your topic's network matters more than the quality of any single video in it. A well-made video in a small network still stays small. Getting into a larger network matters first.
41. Suggested traffic is the slow way out of your existing audience. It starts as low-cost testing and only scales once it clears a confidence threshold with YouTube.
42. Pacing isn't the speed of cuts or the speed of talking. It's the rate of new information. A fast talker with nothing new to say still feels slow.