YouTube Shorts algorithm: what YouTube has actually said

Short answer: YouTube says the Shorts Feed is personalised to what it thinks each viewer wants to see next, using signals like watch history, search history, likes, dislikes, "Not interested" and satisfaction surveys. For creators, YouTube's people keep pointing to one thing: whether viewers choose to watch your Short rather than swipe past it. Beyond that, most of what you read online is guesswork.

This post sticks to what YouTube has said in its Help Center, on its official blog and through its own staff. Where we go further, we mark it as speculation.

What has YouTube officially said about the Shorts algorithm?

• YouTube Help, How YouTube recommendations work: "The Shorts Feed is personalized to reflect what we think you want to see next."
• Same page: different YouTube features rely on certain recommendation signals more than others.
• Creator Insider, 2022, as quoted by Search Engine Journal: the performance of a Short is dictated by whether or not people choose to watch it.
• Todd Sherman, then Shorts product lead, 2023, as reported by Search Engine Journal: "The audience is the algorithm."
• Rene Ritchie, YouTube Creator Liaison, 2024, as quoted by Tubefilter: "The algorithm follows the audience, so please the audience."
• YouTube Official Blog, August 2026: the change in how views are counted makes "no change to recommendations".

Put together, the official message is consistent: there is no separate trick for the algorithm. It tries to show each viewer Shorts they will watch and value, and it learns that from how viewers behave.

What signals does the Shorts feed use?

YouTube Help lists the signals behind recommendations in general. It does not give weights, and it does not list Shorts-only signals.

• Watch history: the videos a viewer has watched.
• Search history: what they search for on YouTube.
• Likes: used to predict what else they may like.
• Dislikes: used to avoid similar recommendations.
• "Not interested": used to avoid similar recommendations.
• Satisfaction surveys: viewers rate videos they watched, which helps measure satisfaction and not just watch time.

A 2021 post on the YouTube Official Blog by Cristos Goodrow, VP of Engineering, adds clicks and sharing to that picture, and describes "valued watchtime" measured through those surveys. That post covers recommendations broadly, not Shorts alone.

What does Stayed to watch tell you?

In YouTube Studio, each Short has a metric YouTube Help calls "Stayed to watch": the percentage of times viewers stayed to watch past the initial seconds. It used to appear as "Viewed vs. swiped away". This is the closest public number to "did people choose to watch", which YouTube says drives Short performance.

YouTube does not publish a good or bad value for it. Compare each Short with the median of your own recent Shorts instead. The opening decides most of this number; our post on YouTube Shorts hooks has templates and a way to test them.

Are Shorts and long videos recommended separately?

Partly, by YouTube's own account. In a 2022 Creator Insider video, as quoted by Search Engine Journal, YouTube said "we separate Shorts and long-form content from watch history". In the 2023 interview, Sherman said the core goal is the same for both, getting videos to people who value them, but that Shorts are measured differently because viewers swipe through a feed rather than choosing from a list.

YouTube has also said it does not see Shorts hurting long-form channels. The 2022 video said channels that made Shorts seemed to be growing faster, and Sherman said in 2023 that YouTube does not see Shorts cannibalising long-form, as reported by TechCrunch. Whether a viewer's Shorts habits influence their long-form recommendations today is not something YouTube has spelled out in detail, so treat any firm claim either way as speculation.

Do hashtags help Shorts get recommended?

Not in any special way YouTube has described. YouTube Help's guide to getting started with Shorts lists where viewers find Shorts (the Shorts tab and feed, search, the home page, channel pages, Subscriptions and notifications) and does not mention hashtags. Sherman said in 2023 that hashtags are not required but can be useful, as reported by Search Engine Journal. Use one or two that describe the topic; do not expect them to change reach.

How are Shorts views counted now?

From 31 March 2025, a Shorts view counted each time a Short started to play or replay, according to YouTube's announcement in the YouTube Community. From 24 August 2026, views across all formats count from the first frame, according to the YouTube Official Blog, and the older definition is now called an "engaged view": clicking to watch or watching past the initial seconds. YouTube Help defines engaged views for Shorts as stays past the initial seconds, not including loops.

YouTube said this change does not affect recommendations or the YouTube Partner Program. It does affect your comparisons: public view counts of older videos were counted under older rules.

Why do Shorts get views and then stop?

According to Search Engine Journal's and TechCrunch's reports on the 2023 interview, Sherman said early views are exploratory. YouTube shows a new Short to some viewers, looks at how they respond and widens or narrows from there, so a Short can rise and then taper off. He also advised against deleting and reposting Shorts, which can be seen as spam.

What is speculation, not official?

These claims are common in creator videos and blog posts. YouTube has not confirmed any of them in the sources we checked, so we label them as speculation.

• Fixed test batches, such as "every Short is first shown to exactly N people".
• A specific Stayed to watch or completion rate that triggers more reach.
• Hashtags or keywords that "boost" a Short or get it "shadowbanned".
• A named two-phase "explore then exploit" model. It is a reasonable description of how recommendation systems often work, but it is not YouTube's wording.
• Penalties for posting Shorts and long videos on the same channel. YouTube has said the opposite.

What should I do with this?

1. 1Make people choose to watchPut the promise in the first line and the proof in the first frame. That is what Stayed to watch measures.
2. 2Keep them satisfiedDeliver what the opening promised. Surveys, likes, dislikes and "Not interested" all feed recommendations.
3. 3Study your audience, not the algorithmCheck which of your Shorts beat your own median and make more in that shape.
4. 4Measure fairlyCompare Shorts at the same age, against the median of your last 20 Shorts, and keep in mind the view-count changes.

If you want a quick read on a single video, the free check gives the real multiplier against its channel and three reasons it may have worked. It uses the official YouTube Data API and cannot tell you anything about YouTube's internal ranking; nobody outside YouTube can.

Questions

How does the YouTube Shorts algorithm work, in one sentence?

It shows each viewer the Shorts it predicts they will watch and value, based on their history and on how other viewers responded, according to YouTube Help and YouTube staff.

Is the Shorts algorithm different from the long-form one?

The goal is the same, YouTube says, but the signals are read differently because Shorts are swiped in a feed. YouTube said in 2022 that it separates Shorts and long-form content in watch history.

Does posting Shorts hurt my long videos?

YouTube has said it does not see Shorts cannibalising long-form, and that channels making Shorts seemed to grow faster. It has not published data you can check yourself.

Do I need #Shorts in the title?

YouTube Help does not list it as a requirement, and YouTube's Shorts lead said hashtags are not required. Use them only if they describe the topic.

Did the 2026 view count change affect reach?

YouTube's blog says no: there is no change to recommendations or to the YouTube Partner Program. Only the public count changed.