How TikTok, Instagram Reels, Facebook Reels and YouTube Shorts rank videos, in their own words

TikTok

• What it uses: your interactions (videos you like or share, accounts you follow, comments you post), information about the video (captions, sounds and hashtags), and your device and account settings, such as language and country, which get lower weight.
• A strong signal is whether someone finishes watching a longer video from beginning to end. A weak one is whether the viewer and the creator are in the same country.
• Neither follower count nor an account's earlier high-performing videos are direct factors.
• Not recommended: duplicated content, spam, videos the viewer has already seen, and videos just uploaded or under review.

From TikTok's explanation of the For You feed (TikTok Newsroom, June 18, 2020). It is TikTok's most detailed public description, and it is several years old.

Instagram Reels

Adam Mosseri, who runs Instagram, wrote that "the most important predictions we make are how likely you are to reshare a reel, watch a reel all the way through, like it", and whether you go to its audio page. The signals, in rough order of importance, are your activity, your history with the person who posted, information about the reel, and information about the person who posted. Low-resolution or watermarked reels, muted reels, reels with borders, reels that are mostly text, and reels already posted on Instagram are shown less (Instagram, May 31, 2023).

Meta's system card for Reels lists more of what it predicts: how likely you are to watch less than three seconds, to comment, to reshare, to share a reel off Instagram, to follow its author, to use its audio, and to watch past the 95th percentile of reels the same length. Its inputs include how often a reel was dismissed, how often it was viewed with the sound on, and the reel's length (Meta Transparency Center, updated November 11, 2025).

In July 2026 Meta said its largest Reels ranking change to date raised Instagram sessions, with particular strength in reshares and time spent, that its models now identify high-quality new reels when they are created, and that over half of the content Instagram Feed recommends is less than a day old (Meta earnings call, July 29, 2026).

Facebook Reels

Meta's system card says the Facebook Reels system predicts how likely you are to like a reel, how long you'll watch it, whether you'll watch it to completion, and whether you'll hide it or ask to see less like it. Its inputs include the reels you recently liked, watched, shared or commented on, the reel's author, and how many times others watched the reel to completion (Meta Transparency Center, updated June 19, 2026).

YouTube Shorts

YouTube says its recommendations draw on "viewing behavior, likes, dislikes, subscriptions, and feedback, including from satisfaction surveys" (How YouTube Works). Shorts revenue counts engaged views only, and non-original Shorts, such as unedited clips from others' films or TV shows, don't count (YouTube Help). In January 2026 YouTube said it is building on its spam and clickbait systems to reduce the spread of low-quality AI content (YouTube Official Blog).

What the four have in common

• The opening seconds count everywhere: Instagram predicts whether you'll leave within three seconds, and TikTok and Facebook weigh watching to the end.
• On Instagram, completion is judged against reels of the same length, so a longer reel is compared with other long reels.
• Sharing matters: Instagram predicts reshares and shares off the app, and Meta's 2026 ranking change was strongest in reshares and time spent.
• Sound matters on Instagram, which counts views with the sound on.
• Copies lose out: Instagram shows reposted and watermarked reels less, TikTok doesn't recommend duplicated content, and YouTube doesn't count non-original Shorts for revenue.
• New videos get a chance: TikTok says follower count isn't a direct factor, and Meta's models look for high-quality reels as soon as they are posted.