Algorithm updates

What YouTube has actually said about Shorts ranking, 2024 to 2026, and what is folklore

YouTube's own statements on Shorts ranking, 2024 to October 2026, separated from reported claims and rumours like the flattening and retention thresholds.

Updated October 4, 20269 min read12 sources

In short

Trust what YouTube has published: it ranks with "billions of signals" such as watch history, topic interests and engagement, plus survey feedback and channel quality, and since October 1, 2026 it reduces Shorts reach for channels that mainly re-upload other creators' videos. Views have counted from the first play on Shorts since March 31, 2025, and on every format since August 24, 2026. Fixed retention thresholds, the "flattening" of older Shorts and similar claims come from creators and analysts, and YouTube has not confirmed them.

On this page

Three labels decide how much weight a claim deserves
What YouTube says its recommendation system uses
The official Shorts statements we found, 2024 to October 2026
What these statements change for a clipper
Reported but not checkable: the January 2025 Todd Sherman interview
Rumours YouTube has not confirmed
Myths that official statements already disprove
How to test a claim on your own channel, step by step
What a clipper should take from all this
Questions
Sources

Three labels decide how much weight a claim deserves

Most Shorts advice mixes three kinds of claim. We sort every claim on this page into one of them:

Official: YouTube wrote it on its blog, in its Help Center, in a TeamYouTube post or on its How YouTube Works site, and we read the page.
Reported: a news outlet attributes it to a YouTube employee, but we could not read the original video or transcript.
Rumour: a creator, consultant or tool vendor says it, based on their own data, and YouTube has not confirmed it.

A rumour can be true. A good test is whether the claim makes a prediction you could check on your own channel. "The algorithm favours quality" predicts nothing. "Shorts older than 30 days stop getting impressions" predicts something you can check in Analytics. Treat the second kind as a hypothesis to test, not as a rule.

What YouTube says its recommendation system uses

YouTube's How YouTube Works site describes the inputs in general terms. It says understanding "a user's viewing behavior, likes, dislikes, subscriptions, and feedback, including from satisfaction surveys" helps it recommend well. It also says that, as on Google Search, it considers "the reputation and the quality of a channel" and uses external evaluators to judge how an average viewer might see the content.

The October 1, 2026 Shorts post repeats the idea in shorter form: "We evaluate billions of signals", including watch history, topic interests and engagement.

Three things follow, and none needs a rumour. First, the system is personal: it predicts what a given viewer will value, not what is "good" in general. Second, viewer feedback counts, not only watch time: likes, dislikes and surveys are named. Third, the channel matters as well as the single video. The October 2026 update acts on that third point: it targets "channels identified as primarily aggregating or re-uploading", not single clips.

The official Shorts statements we found, 2024 to October 2026

These are the statements we found from YouTube that affect how Shorts are measured or recommended, newest first. Dates are publication dates.

Date	What YouTube said	Where
October 1, 2026	Recommendations will prioritize original content and reduce the reach of content re-uploaded from other creators	TeamYouTube, Help Community
September 23, 2026	Dynamic product tag placement "has absolutely no impact on a video's reach"	Feature Experiments thread
August 24, 2026	Views count "the moment a video starts to play" on all formats	Help Center
May 27, 2026	An AI disclosure label alone "does not change how a video is recommended"	YouTube Blog
January 21, 2026	YouTube is "reducing the spread of low quality, repetitive content" with its spam systems	CEO letter, YouTube Blog
December 18, 2025	Test merging "Dislike" and "Not interested" on Shorts into one thumbs-down icon	Feature Experiments thread
December 16, 2025	Test showing image posts in the Shorts feed	Feature Experiments thread
March 31, 2025	Shorts views count each start or replay, with no minimum watch time; the old metric becomes "engaged views"	TeamYouTube, Help Community
From October 15, 2024	Shorts over one minute with an active Content ID claim are blocked and "will not be playable or recommended"	Help Center
October 3, 2024	Shorts can run up to 3 minutes; YouTube will "improve recommendations for longer Shorts in the coming months"	YouTube Blog

The list is short. YouTube rarely publishes ranking details. Most of what it says is about measurement, eligibility and policy, and creators then infer ranking effects from that.

What these statements change for a clipper

Views are not a quality signal any more. Since March 31, 2025, a Short gets a view when it starts to play. A swipe-away after half a second counts. YouTube kept the old metric as "engaged views" in Advanced Mode and pays on it; Partner Program eligibility uses a related count, "qualified views". If you judge clips by views, you judge them by how often the feed showed them, not by whether people stayed. See three kinds of YouTube view.

Negative feedback is a named input. Dislikes and survey feedback are listed on How YouTube Works. In the December 2025 test that made "Dislike" and "Not interested" one button on Shorts, YouTube said it hoped to "make it clearer on how to tune your Shorts recommendations". This is our reading: a clip that makes people tap thumbs-down sends a stronger signal than one they just swipe past.

Originality is now judged at channel level. The October 2026 update speaks of channels that are "primarily" re-uploading. This is our reading: one original clip on an aggregator channel may not change how the channel is classified. TeamYouTube's post does not say how fast reach recovers.

Long Shorts carry a rights risk, not a ranking penalty. YouTube has not said that 2-minute Shorts rank worse. It has said that a claimed Short over one minute is blocked worldwide, which removes it from recommendations entirely. See the 60-second line.

Reported but not checkable: the January 2025 Todd Sherman interview

On January 28, 2025, YouTube's blog summarised a Creator Insider conversation between Shorts product lead Todd Sherman and creator Jenny Hoyos. The blog summary is about hooks, thumbnails and storytelling. It contains no ranking mechanics.

The trade site ppc.land reported more. It quotes Sherman: "strong signals that predict people enjoying the videos". It also reports that the system looks at view-to-swipe ratios, average view duration, rewatches and retention, and that the feed groups Shorts of similar length. We could not check the video transcript, so treat these as reported, not official.

One common citation is wrong. The figure "99.9% of the views are coming from the Shorts feed" is often credited to Sherman. YouTube's own blog credits it to Jenny Hoyos, in an answer about thumbnails. It is a creator's estimate, not a YouTube statistic.

Rumours YouTube has not confirmed

These claims circulate widely. None has a YouTube source. We list why each is unproven:

"The flattening". Retention strategist Mario Joos said that Shorts older than about 28 to 30 days lost impressions from about September 2025. Search Engine Journal reported it on December 5, 2025. Joos himself called it "a carefully constructed working theory and not a confirmed fact". His evidence is view graphs from large channels. YouTube has not confirmed it. Other causes, such as seasonality or competition, could explain the same drop.
Long-form home page slots cut from about 12 to 2. Dataslayer reported this claim from Joos on December 21, 2025. It is about the desktop home page for long-form, not Shorts. YouTube has not confirmed it, and one analyst's data cannot show whether every user sees the same layout.
Fixed retention thresholds, such as "a Short needs 65% (or 50%) retention to be pushed". We found no YouTube source for any number. YouTube's statements describe many signals, not a pass mark.
"Shorts and long-form were formally decoupled in 2026", often with quotes credited to Rene Ritchie. We found no primary source.
A Sherman quote that Shorts rank "independent of duration". We found no source. The only official length statement is the October 2024 promise to improve recommendations for longer Shorts.

Myths that official statements already disprove

Some advice was true once, or was never true. Here the official record answers it:

Claim	Status	What YouTube said
"A Shorts view only counts after a few seconds"	False since March 31, 2025	Views count each start or replay; engaged views keep the old rule
"An AI label kills your reach"	False according to YouTube	A label alone "does not change how a video is recommended" (May 27, 2026)
"Ask fans to hype your Short"	Not possible	The Hype help page lists YouTube Shorts as not eligible
"Custom Shorts thumbnails drive feed views"	Weak	Hoyos, on YouTube's blog: thumbnails matter little because nearly all views come from the feed
"Product tags hurt reach"	False for the September 2026 placement change	"absolutely no impact on a video's reach"
"A license from the creator protects my clip channel's reach"	Unsupported	The October 2026 post makes no exception for licensed clips
"My views jumped, so the algorithm likes me more"	Check the date first	Counting changed on March 31, 2025 for Shorts and on August 24, 2026 for all formats

How to test a claim on your own channel, step by step

You cannot prove how the algorithm works from outside. You can test whether a claim holds on your channel. Choose a claim that predicts something measurable, and try to prove it wrong.

Write the claim as a prediction. Example: "Shorts older than 30 days get almost no new impressions."
Pick the metric that would show it. Use engaged views, not views, because views count every start since March 31, 2025.
Pull at least 20 Shorts from your own channel, from Analytics in Advanced Mode, with their daily data.
Look for counter-examples. One older Short that keeps gaining engaged views in the feed is evidence against the strong version of the claim.
Change one thing at a time when you test a tactic, such as hook length or clip length. Post comparable clips in the same week, so seasonality affects both.
Keep a log of each test with the date, the change and the result. Recheck it after every official YouTube change, such as the October 1, 2026 originality update.
Drop any tactic that only works by disguising re-uploads. YouTube's Spam Policy now names cropping, heavy filters and sped-up audio as detection evasion; see detection evasion.

What a clipper should take from all this

The official record is short, and it points one way. YouTube measures what viewers do and say, it judges channels as well as videos, and it is reducing reach for channels built on other creators' work. Tactics that try to read hidden thresholds are a weak basis for a channel. Tactics that make a viewer stay, rewatch and not tap thumbs-down are consistent with everything YouTube has published.

For practical tactics, read hooks for clips and clip length. If reach drops and you suspect suppression, shadowbans explains how to tell a policy problem from normal variation.

Questions

Does YouTube use a retention percentage to decide which Shorts to push?

YouTube has never published one. It says it uses billions of signals, including watch history, topic interests and engagement, plus likes, dislikes and survey feedback. Numbers like 50% or 65% come from creators, not from YouTube.

Is "the flattening" of older Shorts real?

It is unconfirmed. Mario Joos described it as a working theory in late 2025, based on graphs from large channels. YouTube has not confirmed any change. Test it on your own engaged-views data before you change your posting plan.

Why did my Shorts views jump after March 2025?

Because YouTube changed the count. From March 31, 2025, a Shorts view counts each time the Short starts to play or replay. The older metric is now called engaged views. Earnings still use engaged views, and Partner Program eligibility uses qualified views.

Does disclosing AI use reduce reach?

YouTube says no. In its May 27, 2026 post on AI labels, it wrote that a disclosure label alone does not change how a video is recommended or whether it can earn money.

Did YouTube say longer Shorts rank worse?

No. In October 2024 it said it would improve recommendations for longer Shorts. The real risk for long clips is rights: a Short over one minute with any active Content ID claim is blocked worldwide.

Where does YouTube post algorithm changes?

Mostly in the YouTube Help Community (TeamYouTube posts and the Feature Experiments thread), the YouTube Blog and the Creator Insider channel. The October 1, 2026 originality update appeared first in the Help Community, not on the blog.

Sources

We checked every claim on this page against these sources on the date above. Platform rules and rates change often. Found an error? Tell us.

In short

Questions

Sources

Related guides