What the algorithm is actually doing
The For You feed is a ranking problem. Out of the millions of posts X could show you, it picks a few hundred, scores them, and sorts. The score is a prediction: how likely are you to reply, like, repost, or spend time on this post. Every post you publish is being scored the same way for every candidate viewer.
So the useful question is not what does the algorithm like. It is what does it predict people will do with my post. If the honest prediction is they will scroll past, no timing trick changes that. If the prediction is they will stop and reply, you get distribution.
The signals that carry weight
Not all engagement is equal. The ranking model weights actions by how much effort and intent they show. A reply is a stronger vote than a like, because it costs more. Here is the rough hierarchy, from strongest to weakest.
| Signal | Why it is weighted this way | What it means for you |
|---|---|---|
| Reply, then a reply back from you | A real conversation is the hardest signal to fake | Write posts that invite a specific response, then answer them |
| Reposts and quotes | Someone put their own name behind your post | Give people a line worth stealing |
| Dwell time | Reading the whole thing beats a reflex tap | Longer, finishable posts can outperform one-liners |
| Likes | Cheap, but still positive | Nice to have, not the goal |
| Profile clicks and follows | The strongest long-term signal | Consistency in one lane earns these |
Notice that the top of the list is all about conversation, not applause. The model would rather show a post that fifty people argue about than one that five hundred people silently like.
What quietly holds a post back
There is a second set of signals that pulls scores down. These are the ones people trip over without realizing it.
- Negative feedback: people hitting not interested, mute, or block after seeing your post. A little is normal. A pattern teaches the model to show you less.
- Outbound links in the post body. X wants to keep people on X, so a bare link can dampen reach. Leading with the idea and putting the link in a reply is the common workaround.
- Posting into dead air. If your own followers routinely ignore you, the model has less reason to expand your post to strangers.
- Engagement bait that gets reported. Reply if you agree posts can win short term and cost you trust long term.
How a post travels through the system
- 1Candidate generationThe system gathers posts you might see, roughly split between accounts you follow and accounts you do not. Your post enters this pool for your followers first.
- 2ScoringEach candidate gets a predicted engagement score from the ranking model. This is where your first line, your topic, and your track record with that viewer come in.
- 3Early feedbackThe first people who see it react or ignore it. Strong early replies and reposts raise the post's real score and widen its reach. Weak signals cap it.
- 4Expansion or decayA post that keeps earning replies keeps getting shown to new clusters of people. One that stalls fades. This is why the first hour matters more than the first day.
How to work with it, not around it
- Lead with a first line that earns the second, since only that line is guaranteed to show in the feed.
- Write something specific enough that a real reply is easy to give.
- Reply to your own replies. The conversation itself is a ranked signal.
- Keep links out of the main post when reach matters. Put them in a follow-up.
- Post in a consistent lane so your followers actually engage, which feeds every downstream score.