Why are my X views so low? The one slot you get

#Quick answer

Your views are low because the only mechanism that shows a small account's post to strangers is one slot, once, and it closes fast. X's open-sourced For You code reserves roughly a single position near rank 15 of each reader's feed for an under-exposed post, and the shipped defaults only consider authors with 50,000 followers or fewer, posts under two hours old, and posts still under 200 views on Home. Replies and reposts are excluded outright. After that window, your reach is whatever your own followers happen to open.

Last updated: October 2026

#TL;DR

Most answers to this question are a list of habits: post at the right time, use fewer links, engage more. Those are fine, but they are guesses about a system that is now public. X (formerly Twitter) open-sourced the For You ranker, and the file that decides whether a small account's post gets shown to anyone outside its follower graph is sitting in the repo with its default thresholds in plain text.

Read it and the question changes. It stops being "what did I do wrong" and becomes "which of four gates did this post fail". This page walks the gates, gives you a real baseline to compare against, and separates the causes that are in the code from the ones that are folklore.

One honesty note before the numbers. The repository ships default parameter values, and X can override any of them in production through its feature-switch layer. The defaults tell you the shape of the system and the thresholds X considered sensible. They are not a guarantee of the live value on any given day.

#First, what a "view" even counts

X's own View counts page is short and worth reading before you diagnose anything:

• Anyone logged into X who sees the post counts as a view, wherever they saw it: Home, Search, a profile, a notification.
• Looking at your own post counts. Your first view is you.
• Repeat views count. Reading a post on your laptop and again on your phone is two views.
• Community posts and older posts do not report view counts at all.

That page defines a view by a logged-in reader and does not credit logged-out visitors. If your post got shared into a group chat or picked up by an aggregator, a chunk of the people who actually read it were never counted. It also means the floor is not zero. A post showing 7 views has been seen by roughly six strangers and you.

#The one slot: what author_cold_start.rs actually does

The file is home-mixer/scorers/author_cold_start.rs. Its job is the opposite of a penalty: it exists to give posts that nobody has seen yet a chance to be seen. Here is the mechanism, with the default values from home-mixer/params/param.rs.

Two details do most of the damage to a small account's reach.

It fires once per feed request. The function picks a single best candidate and lifts that one post's score to the target rank. Not a batch. One post, one slot, per reader, per refresh.

Replies and reposts are not eligible. The gate function cold_start_base_eligible requires in_reply_to_tweet_id.is_none() and retweeted_tweet_id.is_none(). So the one promotion path built for accounts your size only applies to original standalone posts. If your last ten posts were replies, none of them were ever candidates for it.

The selection is also deliberately pessimistic. When Thompson sampling is on, each eligible post gets a reward sampled from a Beta distribution seeded with ColdStartBetaAlpha0 0.75 and ColdStartBetaBeta0 49.25, then updated with its real likes and views. A brand-new post with no engagement starts at a prior mean of 0.75 divided by 50, which is 1.5 percent. X samples two candidates and takes the higher-scoring one. You are in a lottery with a small ticket count, and the prior assumes you will lose.

#The second clock: 48 hours, then nothing

Even if a post clears cold start and starts travelling, out-of-network distribution has its own expiry. SimclustersMaxCandidateAgeHours defaults to 48. SimClusters is the system that surfaces posts from accounts a reader does not follow, so after two days your post is no longer retrievable that way regardless of how well it performed.

This is why "it picked up three days later" almost never happens on X the way it does on YouTube or Pinterest. There is no long tail in the feed. The practical consequence: a post that underperforms is not worth reviving, it is worth rewriting. Our breakdown of how the X algorithm works covers the retrieval stages this sits inside.

#The six real causes, in the order to check them

1. The post was a reply or a repost. Excluded from cold start by default. Expected reach is your own followers plus whatever the conversation gives you, and nothing more.
2. It crossed 200 views before anyone engaged. The threshold is a gate, not a floor. A post that drifts past 200 views with no likes has used its eligibility and gets no second pass.
3. You posted when your own followers were offline. The two-hour clock runs on wall time. If your first hundred readers arrive at hour four, the cold-start window closed while they were asleep. This is the one real reason timing matters, and it is a narrower reason than most timing advice implies. See best time to post on X for the window that fits your audience.
4. You posted in a burst. Stacking three posts in ten minutes means they compete with each other for the same single reserved slot in the same reader's refresh. How often to post on X has the per-request arithmetic.
5. The platform got more crowded. Socialinsider's 2026 benchmark report, built on 70 million posts from January 2024 through December 2025, has X posting volume rising from 50 to 70 posts a month per account while engagement rate fell from 0.15 to 0.12 percent. More supply chasing the same attention. Your flat view count may be a flat share of a bigger pile.
6. Actual visibility limiting. Rare, and checkable rather than inferable. X documents reduced reach and interstitials as enforcement actions, and you can read what applies to your account instead of guessing. The X shadowban check page covers what X's own tooling shows and what it does not.

#What is not causing it

Links are not penalised in the open code. There is no link demotion parameter. What exists is OpenLinkWeight at 0.2 and ClickWeight at 0.3, against ReplyWeight 5.0, QuoteWeight 5.0, ShareWeight 2.0 and FavoriteWeight 0.5. A link post is not pushed down, it is competing with a weaker signal. That distinction matters, because the usual workaround of putting the link in a reply does not escape a penalty that was never there. It moves the burden to a post that is itself ineligible for cold start.

Deleting and reposting does not reset anything useful. A repost of your own content is a new post ID with a fresh two-hour clock, which is the one thing that genuinely resets. But the audience that already scrolled past it has not forgotten, and NotInterestedWeight sits at -47.52 while MuteAuthorWeight is -58.8. Repetition has real downside.

"Shadowban" is not a platform state. The thing people describe by that word is visibility limiting, which X names in its own enforcement documentation. Using the right word matters here because one of them is checkable and the other invites you to fix a problem you do not have.

#How to diagnose yours in ten minutes

1. Open your own analytics and pull the last 30 posts. Our guide to reading X analytics covers which numbers are real and which are rolled up.
2. Split them into originals and replies. Compare medians, not averages, because one lucky post ruins an average.
3. For your originals, note how many views each had at the one-hour mark if you have it. Posts that were still in single digits at an hour never cleared the gate.
4. Compute replies plus quotes divided by likes on your best and worst. If the good posts have a healthier conversation rate, your problem is the opening line, not distribution.
5. Check your X engagement rate against your own history rather than a published benchmark.

#What actually raises it

Given the mechanism, three things move the number.

Front-load the first two hours. Not because early engagement is magically weighted, but because the eligibility window is literally two hours and 200 views wide. Post when your people are awake, then be present in the replies while the clock is running.

Make originals the thing you optimise. Replies grow you by putting you inside someone else's audience, and they are still the best use of a small account's time. But they can never win the cold-start slot. If every post you make is a reply, you have opted out of the only distribution channel built for your follower count.

Write for the reply, not the like. A reply is predicted at 5.0 against a like at 0.5, and the bidirectional-follow reply boost sits at 15.0 as of the July 2026 change. A post that gets 12 replies and 9 likes outranks one with 40 likes and silence. That is a writing problem, which is the good news, because it is the part you control. If you want the scoring weights in full, what the open-source X algorithm shows walks the file.

If the bottleneck is that you cannot be present for the two-hour window on the days that matter, that is what X-Autopilot is for: it runs from your own Mac, in your own browser session, so the posting and replying happen on your schedule rather than a cloud queue's. It lowers the detection surface compared with cloud tools because your session cookie never leaves your machine, though automating engagement through a browser is a gray area under X's rules and no tool makes that risk disappear.

#The honest summary

Low views on a small account are usually not a punishment. They are the default outcome of a system that gives you one reserved slot, for two hours, while you are under 200 views, and only if you posted something original. The fix is not a trick. It is making sure you are actually eligible, and then giving that single slot a post worth replying to.