Grow your YouTube channel with real viewers.

YouTube’s 2026 ranker isn’t one algorithm — it’s two, and they treat the same channel differently. The long-form feed (Home, Up Next, Subscriptions) and the Shorts feed run on separate ranking stacks with different weightings and different decay windows. Which one your upload lands in determines what kind of supporting signals actually move distribution. Most third-party SEO content collapses these into a single “the YouTube algorithm,” which is why so much of it gives advice that contradicts what you’d see in YouTube Studio if you tested it.

For long-form, the homepage ranker reads two master signals: Click-Through Rate on the thumbnail and Average View Duration as a percentage of total length. CTR alone won’t save a bait thumbnail; AVD alone won’t save an unopened video. The ranker multiplies them — high CTR plus low AVD reads as click-bait and distribution dampens within hours; low CTR plus high AVD reads as a discovery problem (good content, weak packaging) and the ranker waits for impressions to accumulate before scoring. The 50% AVD percentage is the rough breakthrough mark for medium-length uploads (8–15 minutes). Channels consistently above it earn expanded distribution; channels consistently below it get parked in the long-tail bucket regardless of subscriber count.

The first 48 hours after upload are decisive. YouTube seeds the video to a small probe audience — your own subscribers plus a handful of topically similar viewers — and measures CTR against AVD on that probe. Strong probe performance unlocks expanded distribution to the broader Suggested Videos rail and the Home feed; weak probe performance caps the video at the probe audience plus a residual long-tail. After 48 hours the performance profile is largely locked in and active promotion tapers — what creators commonly call watch-time decay, and what the YouTube Creator Insider channel (YouTube Creators) has walked through in its algorithm-update series. Videos can still go viral later, but they almost always need a second-wave catalyst: a Suggested placement off a trending video, an external embed, or a fresh upload that reactivates subscriber attention on the channel.

The Suggested Videos rail — the right-hand sidebar on desktop and the swipe-up surface on mobile — is the single largest source of long-form views for most channels past 50,000 subscribers. Suggested runs on collaborative filtering against viewer watch-history vectors: the ranker matches your video against viewers whose recent watch patterns overlap with viewers who finished similar videos. This is why title and description SEO matter on YouTube in a way they don’t on Instagram or TikTok. The text metadata is what lets the ranker build the topical embedding that anchors your video in the watch-history space; weak metadata leaves the embedding fuzzy and the matching unreliable.

The 30-second view-validation threshold is YouTube’s structural defense against bot view inflation. Views are counted on the public counter at 30 seconds of continuous watch time; views below that threshold tick the counter but get excluded from the validation pass that backs Partner Program watch-hour accumulation, ad-rev attribution, and Suggested-rail signal. This is why bot-tier services priced at $0.99 per 1,000 views can’t actually move any of those signals: their iframe-loops close before the validation gate, so the views show on the counter but never feed the ranker. Real-account services play the video past the threshold every time because the underlying accounts are people actually watching.

The 4,000-hour Partner Program threshold is a separate, non-algorithmic gate. It is a rolling 12-month watch-hour count on public long-form uploads. Shorts don’t count toward the watch-hour total (they have a separate 10M-views-in-90-days path added in 2023), and neither do unlisted or private watches. Hit 4,000 hours plus 1,000 subscribers and the channel becomes eligible for monetization, but only after a manual Community Guidelines review that can take weeks. Watch hours are the bottleneck for most qualifying channels; a 500-hour channel shipping 50-minute long-form closes the gap faster than one stuck at 3-minute content. The official eligibility criteria sit at support.google.com/youtube/answer/72851 if you want the full breakdown.

Likes feed the per-video reputation cohort that the channel-baseline ranker uses to score new uploads. The like-to-view ratio (commonly called the LTV ratio in creator-side analytics) is the single most-watched per-video reputation signal — channels with consistently strong LTV ratios see new uploads enter the probe wave with a higher distribution ceiling than channels with the same subscriber count but weak LTV history. Comments weight heavier than likes in the LTV calculation (because they’re a higher-effort signal), and unique-viewer comments (commenters who haven’t engaged with the channel before) weight heaviest of all. This is why like and comment SKUs aren’t cosmetic on YouTube — they materially shape the probe-distribution ceiling for your next upload.

Embedded views from external sites (blog post embeds, news article inline players, external link previews) DO count toward Partner Program watch-hour accumulation, with a different attribution model than first-party views. The watch session is tagged as “external” in YouTube Studio and weighted into the channel’s watch-hour total at the same rate as direct watches, but the Suggested-rail attribution credits the embed source rather than YouTube itself for the impression. This is why creators with strong inbound external embedding (news features, podcast crossovers, large-blog inclusions) tend to clear the Partner Program threshold faster than channels relying purely on YouTube-native discovery — the watch-hour math compounds differently across both surfaces.

Shorts runs on an almost entirely separate ranker tuned to vertical mobile swipe behavior. Weighting leans on loop count and swipe-away speed rather than CTR — there’s no thumbnail to click. A Short that holds 40%+ rewatches (viewers who tap back to re-watch the loop after it auto-restarts) snowballs into the feed regardless of channel size. That’s also why the subscriber bridge from Shorts to long-form is notoriously weak: the audiences are behaviorally distinct cohorts even when they’re the same accounts. Plan content stacks accordingly — Shorts for top-of-funnel discovery, long-form for the watch-hour accumulation that actually funds the channel. YouTube’s engineering team has been explicit about this separation in the YouTube Official Blog year-end letters, where Shorts and long-form metrics are reported as parallel streams, not unified.

Search-rank inside YouTube itself is a third distribution surface most growth services ignore. YouTube is the second-largest search engine in the world after Google itself; videos that rank for high-intent search queries (“how to fix X,” “product Y review”) earn long-tail watch-time accumulation that compounds for years. Title text, description keyword density, and the first 2-3 seconds of audio (which YouTube transcribes for indexing) drive search-rank placement. This is the surface where text metadata earns the highest leverage — consistent with primary-source guides like Buffer’s YouTube algorithm research, which has tracked the search-rank weighting shift since 2020.