What isn’t true
Every one of these is contradicted by YouTube’s own documentation. Most of them are still in somebody’s onboarding deck.
- Nothing in the plan depends on an upload time, a posting schedule or an unbroken streak.
- A video’s chances are not set by anything you do to the upload. Not publishing it unlisted first, not holding it back a day, not renaming it a week later.
- A video on limited ads or no ads is not being held back in reach. The search and recommendation system does not know which videos are being monetised.
- One video underperforming does not put a mark against the channel. What does that is viewers passing on most of your videos when the videos are recommended to them.
- Trying a new format does not confuse anything. The system has no opinion about what type of video you make.
- Made for Kids is a distribution setting as much as a legal one. Ticking it across a back catalogue changes what those videos get recommended alongside.
- Deleting an underperformer is neither a documented fix nor a documented penalty. The rule everybody repeats traces back to one post by a YouTube product lead, and what he actually recommended was unlisting.
Sources for points 1-7
Points 1, 2, 3, 4 and 6 are all on one page. On publishing: “Publish time is not known to impact a video’s long-term performance”, growth in views “is not correlated with time between uploads”, and YouTube studied thousands of channels that took a break and “found no correlation between break length and changes in views”. On uploading unlisted and flipping it public later: “No, what matters is how viewers respond after it’s been published.” On retitling after launch: “Maybe, but it’s because our systems are responding to how viewers are interacting with your video differently, rather than the act of changing the video title or thumbnail.” On monetisation: “No, our search and recommendation system doesn’t know which videos are being monetised and which are not.” On the kids setting: “Videos that are set as ‘Made for Kids’ are more likely to be recommended alongside other children’s videos.” And on point 4, what a decline actually looks like: “What can lead to a decline in overall channel views is when viewers stop watching most of your videos when they’re recommended to them.” Performance FAQ.
On point 4, in YouTube’s own italics: “An individual video’s underperformance does not penalize a channel overall.” The documented failure mode is a pattern, not a flop. On point 5: “Experimenting with new content formats, such as Shorts, VODs, or livestreams, will not inherently confuse the algorithm or negatively impact a channel’s overall performance”, and elsewhere, “Our system has no opinion about what type of video you make and doesn’t favour any particular format.” How recommendations work and Search and discovery tips.
Point 7 is the one with no YouTube document behind it, and that is the point. There is no help page anywhere saying a deletion costs you anything. What exists is a single post on X from Todd Beaupre, a YouTube product lead: “When you delete a video, you delete your channel’s connection to the audience that watched that video.” He told people to unlist instead. Everything built on top of that quote since is guesswork, including the version you have probably been told. Search Engine Land’s report.
What changed, and what didn’t
Some of this is recent enough that most of the advice written about it is still wrong.
- Any view comparison that spans the change to what counts as a view is not a comparison. The rate metrics did not move with it, whatever you have read.
- A Short longer than a minute carrying any active copyright claim is blocked worldwide, which makes the music bed on a 90 second cut a publish or do not publish decision.
- The A/B test is settled on watch time share, and a test with no clear winner falls back to whichever title and thumbnail went up first.
- You cannot A/B test a Short, a scheduled live or a Premiere, and changing the title or thumbnail during a test ends the test.
- Automatic dubbing is on by default for eligible creators, so a second audio language is now something you opt out of rather than commission.
- A series with one script skeleton and a swapped product is the shape the inauthentic content policy describes. Sharing an intro and an outro is fine. Sharing the substance is not.
Sources for points 8-13
Point 8, in YouTube’s words: “Starting August 24, 2026, views across YouTube are counted from the first frame going forward” across every format. The part the write-ups got backwards is in the same post: “the vast majority of analytics remain anchored on engaged views, so the way metrics like CTR, AVD, and retention are calculated hasn’t changed”, and there is “no change to recommendations ... and no change to the YouTube Partner Program”. Nothing about the ranking system moved. Your year on year view chart did. Engaged views explained.
Point 9: “Any Short that is over one minute in duration with an active copyright claim of any type, including manual claims, will be blocked globally on YouTube.” It will not play, will not be recommended and cannot be monetised, though “there will be no penalties to your channel”. Worth knowing alongside it: “Any videos uploaded on or after this date with a square or vertical aspect ratio up to three minutes in length will be categorised as Shorts” (the date is 15 October 2024), so the aspect ratio decides the format whether you meant it to or not. Three minute Shorts.
Points 10 and 11: “At the end of the experiment, you will see one of following results based on watch time share”, and where nothing wins, “the first title and thumbnail that you uploaded will be the default title and thumbnail for your video”. Also: “You are not able to A/B test Shorts, scheduled lives and Premiere videos at this time”, and “If a video’s title or thumbnail is changed during the test, the test will automatically stop.” A/B test titles and thumbnails.
Point 12: “This feature is enabled by default for eligible creators”, and “When you upload a new video, dubs will be generated automatically”. The reason to care, from YouTube’s own reporting: creators uploading multi-language audio tracks “saw over 25% of their watch time come from views in the video’s non-primary language”. Automatic dubbing and multi-language audio.
Point 13 is the one that catches a normal content calendar. The policy was renamed from repetitious content to inauthentic content, and it says a channel “shouldn’t have content that appears to be produced using a template or where each video doesn’t deliver creative, educational or other value to the viewer”. What it explicitly allows is “the same intro and outro for your videos, but the bulk of your content is different”. The line is the substance, not the format. Channel monetisation policies.
What the edit actually controls
YouTube publishes editing instructions in its own help pages. I have almost never seen anyone quote them.
- Mid-roll slots serve where the cut leaves a pause. A timeline with every breath tightened out of it has nowhere to put one.
- A sponsor read is cut so it can be lifted out. Clean in, clean out, no music bed running across the boundary and no callback to it later in the video.
- The long-form a Short points at opens on the thing the Short promised. No logo sting, no channel graphics, no housekeeping.
- The picture does not dub. Burned-in English text, English lower thirds and English sight gags cap the audience for every translated track.
- The end screen owns the last 5 to 20 seconds, and a video under 25 seconds cannot carry one at all.
- Chapters start at 00:00, run to at least 3, and none of them is shorter than 10 seconds.
- Loudness is checked against what YouTube actually did to the file, not against a number off a forum.
Sources for points 14-20
Point 14: “Ad slots placed at natural breakpoints, such as a pause in audio or transition in visual, are more likely to serve ads as our systems generally find higher viewer retention at these breakpoints”, while slots “placed mid-sentence or mid-action, are less likely to serve ads”. That is a sentence about where an editor leaves air, printed on a monetisation help page. Mid-roll ad breaks.
Point 15 is the only one here built on an announcement rather than a shipped feature, and I am not going to pretend otherwise. When YouTube announced it, it said: “Instead of permanently burning in a sponsorship segment, creators will soon be able to dynamically insert brand segments directly into swappable slots”, removing the read when the deal ends and reselling the slot. It said testing would start with a small group of creators and gave no general release date. If a read has to lift cleanly later, that is decided in the timeline now, and it costs nothing to cut it that way in the meantime. Brand partnerships announcement.
Point 16, verbatim: “Don’t start your long-form video with slow channel graphics, logos, or unrelated updates. The first 5 to 10 seconds of your long video should directly address the topic, question, or hook promised in the Short.” That is YouTube telling an editor what to do with the head of a timeline. Converting Shorts views into long-form growth.
Point 17 has one half YouTube documents and one half it does not. Documented: a video is ineligible for automatic dubbing if “the speech in the original audio is too fast, which would result in an unlistenable, sped-up dub”, or if it “contains copyrighted content”, or runs over 120 minutes. So a jump-cut VO with every breath removed can disqualify itself. Not documented, because it does not need to be: the audio track is what gets translated, and the picture is not. Every graphic you burn in stays in English in all of them. Automatic dubbing.
Points 18 and 19: “Your video has to be at least 25 seconds long to have an end screen”, end screens go in the last 5 to 20 seconds, and 16:9 takes up to four elements. For chapters, “the first timestamp that you list starts with 00:00”, “your video should have at least three timestamps listed in ascending order”, and “the minimum length for video chapters is 10 seconds”. No chapter under ten seconds is a constraint on the running order, not on the description box. End screens and chapters.
Point 20 has no citation because there is nothing to cite. YouTube has never published a loudness target. The minus 14 LUFS figure everyone repeats has no primary source, and I could not find one. What you can check in ten seconds is what happened to your own file: right click any video, Stats for nerds, and read the content loudness offset. That number is YouTube telling you how far it turned your mix down. It is worth more than the folklore.
What to read after it goes live
All of it is in Studio tonight. Most of it sits two clicks off the front page, which is why nobody looks.
- Audience retention is split by subscribers and non-subscribers before anyone calls the opening a problem.
- A spike is treated as three possible things, because YouTube’s definition of one covers watching, rewatching and sharing, and does not say which.
- Retention is read against the length of the video, because YouTube weights the two watch time measures differently depending on how long the video is.
- Click-through rate is read knowing what it leaves out: end screens, cards, notifications, playlists, embeds and the YouTube mobile website.
- A click-through rate that falls as impressions grow is read as the video reaching new people.
- Monthly audience goes in the report instead of subscriber count, because YouTube says subscriber count is the less accurate of the two.
Sources for points 21-26
Point 21: the audience retention segments report lets you “compare new viewers vs returning viewers, subscribers vs non-subscribers”, and organic against paid. If the opening holds for subscribers and falls off a cliff for everyone else, the hook is fine and the cut is assuming context a new viewer does not have. Different problem, different fix. Point 22: “Spikes appear when more viewers are watching, rewatching or sharing those parts of your video.” Three things, one shape on the chart. A beat people loved and a beat nobody understood look identical. That reading is mine, not YouTube’s. Audience retention.
Point 23, which changes how you rank your own back catalogue: “Broadly speaking, relative watch time is more important for short videos and absolute watch time ... for longer videos.” A 40 minute video at 28% and a 12 minute video at 55% are not being judged on the same measure. Performance FAQ.
Point 24 is the one that quietly breaks reporting. An impression only counts if the thumbnail is “shown for more than 1 second and at least 50% of the thumbnail is visible on the screen”, and YouTube lists what is never counted: external websites and apps, the YouTube mobile website, Kids, Music, playlists, content inside the player such as cards and end screens, email and notifications, videos playing in a background tab, and TrueView discovery ads. A video carried by notifications and end screens has a click-through rate that describes almost none of its traffic. Impressions and click-through rate.
Point 25: “Half of all channels and videos on YouTube have an impressions CTR that can range between 2% and 10%”, so there is no number you are supposed to hit. YouTube’s own worked example takes a video from 9% at 10,000 impressions to 3.5% at 100,000 and calls the drop “a sign of success”. Point 26: “Subscriber count is not the most accurate way to estimate the size of your active audience ... Your monthly audience is a more accurate way.” Monthly audience is unique viewers in the last 28 days, split into new, casual and regular. CTR benchmarks, the worked example and new, casual and regular viewers.
None of these reports covers satisfaction, and satisfaction is one of the three things YouTube says it measures, next to appeal and engagement. It gathers it by asking: “user surveys that ask you to rate videos that you watched help the system understand satisfaction, not just watch time”. No report in Studio shows you the answers. That is usually what is going on when a video holds its retention and still dies, and it is also why no amount of staring at the retention graph explains it. How recommendations work and the three measures.