How long does an X post keep getting views?

#Quick answer

How long an X post keeps getting views is two numbers, not one. Half of the engagement is done inside the first hour: the 2026 half-life measurement for X is 52 minutes. The structural end is 48 hours, because the For You candidate pipeline in X's open-sourced ranker drops any post older than a constant called MAX_POST_AGE, set to 48 hours, before scoring begins. Between those two points the curve is already flattening. After the second one, views come from search, your profile and other people's links rather than from recommendation.

Last updated: October 2026

#TL;DR

The reason this question matters is that most posting advice quietly assumes a post accumulates reach for days, the way a blog post does. It does not. On X (formerly Twitter) a post is closer to a conversation than a page: it gets one burst of distribution, and the burst is mostly over before you have finished replying to the first comment.

Three numbers set the shape:

Everything useful about scheduling, replying and reposting falls out of those three.

#What the half-life numbers actually measure

"Half-life" is borrowed from bibliometrics, where it describes how long a journal article takes to lose half its citation value. Applied to a post, it is the time until half of all the engagement that post will ever get has arrived.

For X that figure is now 52 minutes, from the 2026 update of Scott Graffius's half-life report, published 23 January 2026 and built from more than 5,600,000 posts published across 2025. Two things are worth noting about that number. It has been drifting slowly upward, from 43 minutes in 2024 to 49 in 2025 to 52 now. And it is still the second shortest on the list.

A more granular picture comes from an academic study. Jürgen Pfeffer, Daniel Matter and Anahit Sargsyan at TU Munich collected 22,144 posts in January 2023 and then re-measured the impression count on every one of them 99 times, which is how you get a curve rather than a snapshot. Their findings: the peak of impressions per second lands 72 seconds after publishing, the median half-life is about 80 minutes, and after 24 hours there is no relevant number of new impressions for roughly 95 percent of posts. Between the 20-minute mark and the 24-hour mark, the median post's view count grows by a factor of 3.75.

Read those two together and you get the honest answer. Your post is being seen fastest a minute after you hit publish, it is halfway done inside an hour, it roughly triples or quadruples from the 20-minute reading by the next day, and then it stops.

One caveat I will not skip: the TU Munich data is from January 2023, before the Phoenix ranker that runs the feed today. The 2026 figure is newer but measures engagement rather than impressions. They agree on the shape, which is the part to act on, not on a precise minute.

#The 48 hour line in X's own code

The studies describe behaviour. The repository explains the mechanism.

X open-sourced the For You ranker, and the candidate pipeline that assembles posts for scoring includes a filter built from one constant. In home-mixer/params/config.rs:

pub const MAX_POST_AGE: u64 = 48 * 60 * 60;

That value is handed to AgeFilter, which keeps a candidate only if its age since creation is at or under the maximum and drops everything else. The filter runs early, alongside deduplication and hydration, which means an older post is removed before any scoring happens. No engagement score can rescue it, because it never reaches the scorer.

Out-of-network retrieval has the same ceiling from a different direction: SimclustersMaxCandidateAgeHours defaults to 48. So the two routes a stranger could reach your post through both expire on the same day.

Before that, small accounts have a much tighter window. The cold-start scorer that reserves one feed slot for an under-exposed post only considers posts under two hours old, and only from accounts at or below 50,000 followers. That window, and the four gates around it, is the subject of why your X views are so low if that is the question underneath this one.

As always with the open repo: these are shipped default values, and X can override any of them through its feature-switch layer. They tell you the shape of the system rather than the live value on a particular Tuesday.

#Where the late views come from

A post past 48 hours still collects views. They just arrive through doors that have nothing to do with recommendation.

The Following tab. This is the big one, and it surprises people. The candidate source behind the follows-only timeline is a search ranked purely by recency, and its recall window is RECALL_MAX_AGE_SECS = 365 * 24 * 60 * 60, one year. Someone who opens X after a week away and switches to Following is scrolling a chronological list that can reach back far further than the For You feed can. If you want that tab yourself, how to get a chronological feed on X covers the switch and why it keeps resetting.

Search and profile. A post indexed for a phrase people search keeps earning a trickle indefinitely. So does a post pinned to your profile, which is the one piece of real estate on X where an old post gets new eyes every day.

Quote posts and embeds. When someone quotes you, the quote is a brand new post with its own 48-hour clock, and your original rides along inside it. This is the usual explanation for a post that was flat for two days and then moved: you did not get a second burst, someone else got a first one.

Bookmarks and DMs. Quiet, invisible, and the reason your view count sometimes outruns every visible metric on the post.

None of these scale. Together they are a small fraction of a normal post's total. The practical read: do not plan for a long tail on X, because there mostly is not one.

#What this changes about how you post

The first hour is the post. Publishing and walking away wastes most of the distribution you were going to get. If half the views are gone by minute 52, a reply you write at hour three reaches a quarter of your eventual audience. Be there while the clock runs.

Timing matters more than it should. Not because the ranker rewards a magic hour, but because the window is short and absolute. A post published into a sleeping audience burns its half-life on nobody. Best time to post on X has the day-by-day windows from two large studies and, more usefully, how to find your own.

Spacing beats stacking. Since each post has its own short window, the instinct is to publish more of them. That works up to a point and then fights itself, because the ranker applies an author-diversity decay inside a single feed request. How often to post on X has the multipliers.

Measure at one hour and at 24. Two readings tell you more than one. A post that was flat at 60 minutes never cleared its window; a post that was healthy at 60 minutes and barely moved by the next day simply had a normal life. Reading your X analytics covers which numbers in the native dashboard are real.

Recycle by rewriting, not reposting. A fresh post ID genuinely resets both clocks. The same words do not deserve a reset, and the negative signals in the ranker are heavy enough that repetition carries a real cost.

If the honest blocker is that you cannot be present for the hour that matters, that is the gap X-Autopilot fills: it runs on your own Mac, in your own browser session, so posts and replies land inside your windows without you sitting there. Your session cookie never leaves the machine and there is no shared cloud IP pool, which is a lower detection surface than cloud schedulers, not an absence of risk. Automating engagement through a browser remains a gray area under X's automation rules, and no tool changes that.

#What does not extend a post's life

Editing it. An edit creates a new version of the same post, not a new post, so the age the filter reads is unchanged.

Liking or replying to your own post later. Your own view counts, and your own reply is a post with its own clock, but neither resets the parent.

Adding it to a thread two days later. The new entry is new. The old one is still 48 hours old.

Buying engagement. Beyond the account risk, it arrives on a corpse: by the time most purchased engagement lands, the post is no longer a candidate anywhere.

#The honest summary

An X post gets one real burst of distribution. Half of it is over in about 52 minutes, roughly 95 percent of posts are finished inside a day, and at the 48-hour mark the recommendation system stops considering the post at all because a constant in the candidate pipeline says so. What is left is search, your profile, the year-long recall of the Following tab, and whatever other people choose to quote.

That sounds bleak until you invert it. A short lifespan means yesterday's flop costs you almost nothing, and the next post starts from a clean slate in a few hours. The accounts that grow on X are not the ones with the best individual posts. They are the ones that keep showing up for the first hour.