No platform fully discloses how its recommendation systems work, and the rules change regularly. Still, from the documents companies publish, from professionals’ tests and from practices shared across the industry, a fairly stable picture emerges of the signals that matter for short-form video. Knowing them helps you design content with a better chance of being shown.
Note: the exact criteria are proprietary and change over time; what follows summarises general principles, not guaranteed rules.
The basic idea: user interest
Recommendation systems try to show each person the content most likely to keep them on the platform. To do so, they observe how users react to a video and use those reactions to decide whether to show it to more people. That is why quality as perceived by the audience matters more than a profile’s standing.
The most relevant signals
- Watch time and completion: if many people watch the video all the way through, or rewatch it, the signal is positive.
- Replays: a video that is watched again indicates strong interest.
- Shares and saves: actions that demand more effort than a like, interpreted as indicators of value.
- Comments and interactions: especially those that generate conversation.
- Negative signals: drop-offs in the first few seconds, requests to see less similar content, reports.
The hook: the first seconds
The decisive moment is the start. If the audience scrolls away immediately, the system reads the content as uninteresting. The hook can be a question, a surprising image, a clear promise. It must not mislead: a hook that is not delivered on increases later drop-offs. The creative process is described in from brief to reel.
Originality and technical quality
Platforms state that they favour original content and limit the reach of material republished with other platforms’ watermarks. Image and audio quality, suitable length and consistency with the profile’s theme also count, helping the system understand who to show videos to. On the role of audio, see this in-depth piece.
Profile consistency
A profile that posts steadily on the same theme gives the system clearer data about who its audience is. Inconsistency makes recommendation harder. It is one of the reasons why specialisation, already seen in an article on vertical media, works well on social too.
Mistakes to avoid
- Chasing every trend without it fitting your own theme.
- Posting too much and without quality.
- Resorting to tricks to obtain artificial interactions.
- Changing strategy after a few pieces of content, without analysing the data.
Experimenting with method
The best way to understand what works is to test: hook variants, different lengths, alternative formats, posting times. Compare results over a sufficient number of pieces of content, without drawing conclusions from a single video.
In short
Algorithms reward what audiences genuinely appreciate: attention, repetition, sharing. Good content starts there. To choose where to apply these rules, see the comparison of the main platforms.



