83% of the interactions arrive in 10 days: how long a YouTube video really keeps earning

Every creator asks how long a video keeps earning, and the honest answer is that the calendar has nothing to do with it. A video ends when its traffic source ends, and there are only two families of source. One is push, the other is pull.

Push is the home feed and the browse features. YouTube shows the video to your subscribers and to lookalike audiences for a window of days, the graph spikes, and then it falls off a cliff. That fall is not punishment, it is the end of a distribution window that was always temporary.

Pull is search and suggested. Somebody types a question, or finishes a related video and is offered yours. Nothing about that depends on the upload date, which is why a three year old tutorial can outperform what you published this morning.

So before asking how long your video lasts, open the traffic sources tab. A video living on browse has weeks. A video living on search and suggested has years, and it is the second kind that turns a channel into an asset instead of a treadmill. How the YouTube algorithm works in 2026 is the same story told from the platform's side.

Metricool published a study in July 2026 covering 799,718 videos across 71,177 accounts, comparing February 2025 with February 2026. It is the largest recent read on how content behaves after publication, and three of its numbers change how you plan a channel.

First, distribution grew. Long form views rose 76% year over year and Shorts views rose 127%, with the Shorts feed alone generating 61% of all views on the platform. More people are watching more video, and the ceiling did not move down.

Second, attention thinned. Engagement fell 45%, and viewers spent three times less time on each Short. Third, and this is the uncomfortable one, ad impressions, monetized playbacks and estimated ad revenue all fell by more than 50% per account in the same window.

Read together they describe one thing: the same view is worth less than it was, so revenue now comes from having more videos working at once rather than from one video going further. The study also found only 11% of accounts under 10,000 subscribers moved into a higher tier in the year, which is a volume problem wearing a talent costume.

The single most misread number in that study is that 83% of interactions happened within the first 10 days of publication. Creators see it and conclude a video is dead after a week and a half. That is not what it measures.

Interaction is a subscriber behaviour. Likes and comments come from people who saw the video in their feed while it was new, and those people arrive early by definition. Views from search arrive from strangers who never comment, so the interaction curve collapses long before the view curve does.

That gap is why comment count is a terrible health metric for an evergreen video and a fine one for a news video. Judge a tutorial by views in month three, not by likes in week one.

It also sets the right expectation for the first days. A video that gets 450 views in its first week and 2,000 in its first year did not fail in week one, it simply lives on pull traffic. Panicking at day three and deleting it is how creators throw away the exact asset they were trying to build. Whether to pivot or persist around video 30 is the same trap at channel level.

The two formats do not age the same way, and treating them as one library is a planning error. Analysts working with large channels report that Shorts older than roughly 30 days receive noticeably less distribution, with the feed favouring fresh clips.

That makes Shorts a flow business. They buy reach and subscribers now, and they stop paying rent almost immediately. If your entire catalogue is Shorts, your income is only as good as the last 30 days of production, every month, forever.

Long form is the opposite. A tutorial, a documentary, a comparison or a how much does it cost video keeps answering a question people will still type in 2028, and it earns from search and suggested with no new work from you.

The practical split is to let Shorts feed the funnel and let long form hold the asset. Evergreen content against news content is the same decision made at the topic level: a news video is worth more in its first 48 hours than it will ever be worth again, and an evergreen one is worth more in year two than in week one.

Here is where the argument stops being philosophical. Say each evergreen video settles at a modest 300 views a month once its launch window closes. That is a small, unremarkable number, and it is the point.

Ten videos like that produce 3,000 views a month. One hundred produce 30,000 views a month with nothing published today. At the long form RPM range of $4 to $12 that we use across this blog, that catalogue pays $120 to $360 a month in ad revenue alone, and it keeps doing it while you sleep.

The same 100 videos are also a watch hours machine. A 12 minute video watched to 40% delivers 4.8 minutes per view, so 30,000 views a month is 2,400 hours a month, purely from the back catalogue.

The cost side is what decides whether the plan is possible. Producing 100 long videos by hand runs 950 to 1,350 hours at 9.5 to 13.5 hours each. Producing them on the economy mode of an automated line costs 137,700 credits, which is $431 at $0.003133 per credit. That is the entire difference between a catalogue you talk about and one you own.

There is a trap in the monetization rules that almost nobody writes about. From February 1, 2027, new entrants to the Partner Program need 1,000 subscribers plus 8,000 qualified public watch hours in 365 days, or 20 million qualified Shorts views in 90 days. Both windows roll.

Rolling means the hours you earned 400 days ago no longer count today. A catalogue that stopped delivering views does not just stop growing, it quietly drains out of the qualifying window while the number on screen falls.

So the asset is not the video count, it is the video count that is still being watched. An old video that pulls 300 views a month is inside the window forever. An old video nobody finds is a file, not an asset.

This is also why the honest advice is not publish once and wait. To keep 8,000 hours alive you need the catalogue delivering roughly 100,000 views a year at that 4.8 minute average, which is the arithmetic behind our number of about 100 evergreen videos. Channels already inside the program are not held to the new bar, but they do have to accept the updated terms in Studio by January 31, 2027.

Never re-upload it. A re-upload throws away the URL, the watch history, the comments, the external links and the ranking the video earned, and it starts a brand new file at zero. The only correct move is to improve the file that is already ranking.

What actually works is unglamorous: a new thumbnail, a title that matches how people search today, an updated description, and adding the video to a playlist so it feeds the next one. End screens on old videos are the cheapest internal traffic you will ever get.

The second move is translation. A video that answers a question in English answers the same question in Spanish and Portuguese, and the catalogue effect multiplies by market rather than by luck. In FalconVid you duplicate the project into another language and pay only the difference, which is why a proven video is the cheapest new video you can make.

The third is knowing which old video deserves the work. That is an analytics job, and it is what the Senior AI Analyst does from the Pro plan up: it reads the account and writes to you every two days, in your language, with the specific move to make inside the product. An AI analyst reading your channel beats guessing which of 100 files is worth a new thumbnail.

Everything above rewards the same behaviour: publishing evergreen videos consistently for long enough that the back catalogue carries the channel. The obstacle was never the idea, it was the 9.5 to 13.5 hours each video costs by hand.

The FalconVid production line researches the topic, writes the script, narrates it with a premium voice, generates every scene, designs the thumbnail, writes the YouTube metadata and publishes to five networks on a calendar you approve. The AI specialists work in parallel, so a finished video takes up to 30 minutes, and the plan decides how many run at once: 2 on Starter, 5 on Pro, 25 on Agency and 50 on Scale.

The quality mode is chosen per video and all three are in every paid plan. The same 12 minute video costs 1,377 credits on economy, 3,550 on balanced and 15,084 on premium, which at $0.003133 per credit is $4.31, $11.12 and $47.26. Starter at $47 covers 10 economy videos a month, or 7 when two of them go balanced.

And a catalogue can be revived as well as built. If you already have an abandoned channel, the videos that still pull search traffic are the foundation for the next 90 days, and reviving a dead channel or starting a new one is a decision you make with the traffic sources tab open, not with your feelings about the old niche.