How-To Guide

How do you get more impressions on X for a small brand account?

How a small brand gets more impressions on X: what X's published code says decides reach, which vendor numbers lack data, and a 48-hour log to test changes.

9 min read

By the Superdirector editorial team. X's repository and vendor pages read November 7, 2026.

At a glance

How do you get more impressions on X for a small brand account?

To get more impressions on X as a small brand, make each post easy to reach, worth scoring and unlikely to be muted. Stay on one topic so X can match your posts to the right people, write for replies, quotes and shares, reply to the larger accounts your buyers read, keep off-topic bait out of your feed, and judge every post at 48 hours, because X's For You feed drops posts older than that. Compare each post with the median of your last ten, not with a big account. X's published code gives new and small authors a cold-start lift, but it shows no multiplier for Premium, images or links.

  • X's For You feed drops posts older than 48 hours, so read each post at 48 hours against the median of your last ten.
  • Shares by copied link weigh most in X's published file, then replies, quotes and direct-message shares; replies from accounts a viewer does not follow are dropped from their feed.
  • The file shows a cold-start lift for small authors and no multiplier for Premium, images or links.
A four-step numbered sequence of X's For You path: sources, filters, score and visibility, each with a one-word cue.

The four stages follow X's published README, and the cue under each is our reading. Original diagram from X's published README, read November 7, 2026.

Superdirector

Make it work for your brand

Find it. Remake it. Post it. Superdirector turns the viral formats in your niche into your brand's carousels, X posts and video scripts, and keeps the calendar full.

The short answer: reach, score, stay unmuted

To get more impressions on X as a small brand, make each post easy to reach, worth scoring and unlikely to be muted. Stay on one topic so X can match your posts to the right people, write for replies, quotes and shares, reply to the larger accounts your buyers read, keep off-topic bait out of your feed, and judge every post at 48 hours, because X's For You feed drops posts older than that. Compare each post with the median of your last ten, not with a big account. X's published code gives new and small authors a cold-start lift, but it shows no multiplier for Premium, images or links.

Those three parts follow the order in which X assembles its For You feed, now published as code. A post enters the pool because the viewer follows you or because X recommends it. It is then scored on the actions that viewer is predicted to take. Finally it has to survive the filters and the negative reactions that remove posts. A small account leaks impressions at each stage, and each leak has its own repair. This guide goes through them in order, tests the popular tactics against evidence, and ends with a log and a worked example you can copy this week.

Count impressions the way X does

An impression is one appearance of your post on someone's screen. It does not need a click, a like or a reply. Practitioner guides such as Sprout Social's describe it as the total number of times users saw a post, and several add that repeat views by one person count again, which is why impressions can exceed your follower count and exceed reach, the number of different people. X's own Help pages could not be read for this guide, so treat the repeat-view detail as the practitioners' description. The reach versus impressions glossary entry covers the difference.

Impressions and engagement answer different questions, so keep them apart. Impressions say how far a post traveled. Engagement rate says how many of those viewers did something. Ordinal's engagement data study divides likes, replies, reposts and quotes by impressions, leaves clicks out, and uses the median instead of the mean, because a few viral posts drag an average up. The same study reports that average impressions per post fell 5 percent year over year in its dataset, which skews toward B2B and startup accounts. A flat line on your own chart may therefore describe the platform as much as your posts.

Two habits make the number useful. Read every post at the same age, and 48 hours is the natural choice because For You stops showing a post after that. And compare with your own median, since Metricool's 2024 study of 23,561 accounts and about 2.1 million posts found that the largest accounts average the most impressions. A 400-follower account reaching 1,500 impressions is not behind a brand with 40,000 followers reaching 30,000; the comparison that matters is your last ten against this one.

Where impressions go missing for a small account

Before changing tactics, match your own chart to one of these patterns. Each row names a likely cause, which is a hypothesis, and the first thing to check.

What you seeLikely causeFirst check
Most posts land near or below your follower countPosts mostly reach followers and are rarely recommended to anyone else.Compare your best post of the month with your median, and look at how many topics you cover.
One post spiked and the next ten did notA single post cleared the bar once and its trait was not repeated.Write down what the spike post did differently, then copy one trait in the next post.
Impressions dropped suddenly and your habits did not changeA visibility label on the account, or a platform-wide dip.Open X's Under the Hood report if your account has it, then compare with your last three weeks.
Your replies get seen and your original posts do notReplies travel through the thread they sit under, not through feeds.Post an original on the topic of the thread that worked.
Impressions are healthy and followers are flatThe problem is the profile, not the reach.Divide new followers by profile visits in the log, and check your bio and pinned post.
Three posts in an hour each did worse than usualPosts bunched together compete for the same feed slot.Space posts at least a few hours apart for two weeks and compare medians.

Learn what decides whether your post is shown

Everything marked as X's file comes from the X algorithm repository, read on 7 November 2026. The weights file says it mirrors production defaults and was last synced on 6 November 2026. X also says it runs experiments and that a few files are unpublished, so read the numbers as the current published defaults, not as a guarantee about any one post.

X's README describes the For You feed as assembled for each request from two kinds of source. In-network posts are recent posts from accounts the viewer follows. Out-of-network posts are found by retrieval models, by a cluster-similarity source called SimClusters, and by a shared list of popular recent posts. Both kinds are scored by the same model. Before scoring, a list of filters removes candidates. The ones that matter for a small account are in this table, with the settings that go with them.

Rule in X's filesWhereWhat it means for a small brand
Any post older than 48 hours is removed before scoring.README, age filterA post has about two days to earn its reach, so read it at 48 hours and then move on.
Anything the viewer was already shown is removed.README, previously seen filtersFor You does not keep re-serving your post to the same viewer. Repeat views, which practitioners count, would come from other surfaces such as your profile or search.
Replies and reposts by accounts the viewer does not follow are dropped.README, out-of-network filterA reply reaches strangers through the thread it sits under, not through their feed. Only original posts can be recommended to non-followers.
Recommended posts are scaled down by a factor under 1 against posts from followed accounts.README, scoring adjustmentsTo reach non-followers, a post needs a stronger predicted reaction than one from an account the viewer already follows.
A second or third post by the same author in one feed is scaled by a shrinking factor.README, author diversityPosts bunched together compete with each other. Spread them out.
An author whose impressions sit under a threshold gets lifted toward a target slot.README, new-author boostSmall authors get a lift. The settings are 200 impressions, a follower cap of 50,000 and a post age of 7,200 seconds.
Opening a post straight from a group chat has no effect on ranking.param.rs commentAsking friends to like a post from a chat does nothing for ranking.

Read the cold-start line with care. The values sit in the weights file as configuration defaults, and the README describes the new-author boost only as lifting posts from authors below an impressions threshold toward a target position. We read the numbers as a short trial for posts from accounts under 50,000 followers and under 200 impressions, in roughly a post's first two hours. That is an interpretation. It argues for being present and ready when you post, not for any promised reach.

Ranking sets the order, and separate visibility rules decide whether a post can be shown at all, using labels that other systems place on posts and accounts. The X repository describes an Under the Hood transparency tool that shows an account owner aggregate statistics about the labels on their account and posts that can limit visibility. If your impressions fall suddenly while your habits did not change, that tool is the first place to look before blaming content.

Write for replies, quotes and shares

The score is a weighted sum of predicted actions. The weights multiply the predicted probability that a viewer takes an action, not raw counts, and X says so in the file itself. In the current file replies, quotes and shares by direct message weigh 5.0 each and a like 0.708, while a share by copying the link weighs most, at 20.0. A post that makes one person want to answer, add a view or forward it to a colleague has more to gain than a post built for likes.

Two dated updates in the README point to short, front-loaded writing. On 8 October 2026 X said its ranking model now estimates time spent per post better, and on 13 October it said retrieval is also trained to surface posts people read all the way through. Ordinal's data shows 100 to 200 characters earning the highest engagement rate, which fits. Lead with the point, ask one question, and cut the warm-up sentence. The guide to scroll-stopping hooks has patterns to adapt.

  • State a position someone can disagree with. A clear claim earns a reply or a quote.
  • Make one post useful to one person. A checklist or a measured number is what gets sent by direct message.
  • Ask one real question. Pick one your buyers have an opinion about.
  • Stay on one topic. Off-topic posts reach people who then mute or ignore you.
Two columns of X's published action weights: share link 20.0, reply 5.0 and like 0.7 raise the score, while report -234, mute -58.8 and block -31.2 lower it.

The weights come from X's param.rs and multiply a predicted probability, not a count. Original diagram from X's param.rs, read November 7, 2026.

Reply where your buyers already read

Replies are the quickest way to be seen by people who do not follow you, and the file explains both why and how far. Because replies from accounts a viewer does not follow are dropped from their feed, a reply works through the post it answers: the people reading that thread see you. Choose threads that are active but not crowded, and add something the original did not say.

X also boosts original posts from people you follow back. Its change note says the bidirectional follow reply boost was set to 15 on 24 July 2026, after tests with values up to 20. For a small brand the practical step is to list 20 to 30 accounts your buyers read, reply to them with substance, and follow the ones who answer, so mutual follows build up. Treat stories of accounts reaching hundreds of thousands of impressions from zero in days, such as one told on Indie Hackers, as one person's report; the page shows no data. The competitor content guide shows how to keep a list of what is working around you.

Keep that list in a sheet, one row per account, and add a line each time you reply: the hour the original went up, its format, how many replies it already had when you arrived, and the impressions your reply showed the next day. After two weeks the sheet shows which accounts post when your buyers are reading and which threads are already too crowded to be seen in. Drop the accounts where your replies keep drawing the fewest impressions and add new ones in their place.

Keep posts out of mute and block territory

Muting, blocking, reporting or marking a post not interesting all carry negative weights in the file, the largest being report at -234.0 and mute at -58.8. They are large because those actions are rare, and they still multiply predicted probabilities, not counts. The README adds that an account can be labelled from how others respond to it, using blocks, reports and spam reports relative to favorites, and that a separate model reads the sequence of an account's actions for signs of inauthentic behavior.

For a small brand that rules out three habits. Do not post engagement bait outside your topic, because the viewers it reaches are the likeliest to mute you. Do not buy impressions or join engagement pods; several vendor pages that rank for this topic sell impressions, and X's published code includes a model that reads an account's action sequence for inauthentic behavior. And do not coordinate likes from a group chat; the file says that path does not count.

Test pace and format on your own account

Posting more is correlated with more impressions in the data we found, but the data cannot say it causes them. Metricool's 2024 study reports an average of 12 posts a week on X and almost 95 a week for the largest accounts, and says content value matters beyond frequency. X's own code decays repeated posts by one author in one feed. A sensible start is one post a day or three to five a week, then testing one step up or down for a month, reading every post at 48 hours.

Format is the second test. Ordinal's dataset has text-only posts averaging 35,975 impressions against 18,334 for media posts, while media posts converted at a higher engagement rate (0.79 percent against 0.68). That contradicts the common claim that images raise impressions, with the caveat that the sample leans B2B and startup. Links are the third. The file gives opening a link a positive weight of 0.2 and shows no negative link weight, yet vendors report lower reach for links. Test it yourself: run ten posts with the link in the post and ten with it in the first reply, and compare medians.

TestChange one thingRead at 48 hoursKeep it if
PaceOne more post a week, or one fewerMedian impressions and the ratio columnThe median holds or rises with no drop in engagement rate
FormatEvery third post gets an imageImpressions and engagement rate by formatEither rises against your text-only median
LinkLink in the post, then in the first replyImpressions for ten posts eachOne version is clearly ahead on the median

Review each post at 48 hours

The 48-hour log has one row per post with impressions at 48 hours, replies, reposts, quotes, likes, profile visits and new followers, and a note for link and media. It computes engagement rate with the formula above and divides each post by the median of the ten rows before it, so 1.0 means your usual. Our rule of thumb, not X's: a post at twice your median or more is a type to repeat, and two posts in a row under half of it say the topic or type is not landing. The weekly content plan guide shows how to fit the posting and replying into a week.

A worked example, with made-up numbers. Ten posts reach 310, 450, 280, 1,900, 520, 390, 610, 340, 470 and 420 impressions at 48 hours. Sorted, the middle two are 420 and 450, so the median is 435. The 1,900 post is 4.4 times the median and the 280 post is 0.64 times. The average is 569, higher than eight of the ten posts, which is why the median is the better yardstick. The next step is to read what the 1,900 post did differently, such as the topic, the first line, the hour and whether it drew replies, and to write one more post that copies one of those traits. One strong post proves little; a second one with the same trait starts to.

Ten made-up posts at 48 hours with the median at 435, twice the median at 870 and half at 218; only the 1,900 post sits above twice the median.

The post numbers are made up for the worked example above, and the twice- and half-median lines are our rule of thumb, not X's. Original diagram from the worked example above.

Numbers you will read, checked against X's file

The pages that rank for this question quote many multipliers. We read them in full and checked each against X's published files or a measured dataset on 7 November 2026. None of the vendor figures below came with data.

Each status says how strong the check is, and the labels are not equal. Absent from the file is the weakest: X says some files are unpublished, so a missing multiplier is not a denial that one exists. Contradicted by a dataset means measured data points the other way, within that dataset's sample. Not in the current or 2023 weights means neither X's current file nor the weights Twitter published in 2023 support the figure.

Claim you will readWho says itStatus
Premium accounts get 20 to 50 percent more impressions, or 2 to 4 times in the first hourNowadays; Indie HackersAbsent from the file. No data shown, and the published weights contain no Premium multiplier; the README lists subscription status among the details loaded for each post.
Images bring about 150 percent more impressions than text-only postsNowadaysContradicted by a dataset. Ordinal's data shows text-only posts averaging more impressions than media posts, with a B2B-leaning sample.
Five posts a day average 4 to 6 times the impressions of oneNowadaysNo data shown. Metricool finds that bigger accounts post more and reach more, which is correlation, not a measured effect of pace.
The first 30 to 60 minutes decide distributionNowadays; BlockAIAbsent from the file. The only time settings found are the 48 hour age filter and a 7,200-second cold-start post age.
Moving links out of the post raises impressions by 30 to 50 percentBlockAIAbsent from the file. Opening a link has a positive weight, 0.2, and no negative link weight appears, so test it on your own account.
Reposts weigh 20 times a like and replies 13.5 timesIndie HackersNot in the current or 2023 weights. The current file gives replies 5.0, reposts 1.0 and likes 0.708. In 2023 Twitter gave a reply 13.5, but a repost was worth 2 likes, not 20.
58 percent of impressions come from non-followersX PromotionNo source, and the page sells X impressions.

Where Superdirector fits

Superdirector publishes this guide, so weigh this section with that in mind. Its main offer is $199 a month for one social account that its AI runs in the direction you set. For X, it tracks the accounts, competitors and keywords you name, finds posts doing unusually well in your niche, writes X posts in your brand's voice and publishes single X text posts once posting is on for that account, which takes one answer to a warning; nothing goes out in the first ten minutes, and after that it posts by itself. The $29 a month plan is the hands-on entry: you work from Claude or ChatGPT with live social data, ideas and drafts, and you post yourself. Replying to other accounts, the tactic this guide leans on most, stays with you. AI video and full autopilot are coming soon. Nothing here promises more impressions; the 48-hour log is how you would judge whether it helps. See what runs today.

Frequently asked questions

How do you get more impressions on X for a small brand account?

Stay on one topic, write for replies, quotes and shares, reply to larger accounts your buyers read, and keep off-topic bait out of your feed so posts avoid mutes and blocks. Read each post at 48 hours, the age at which X's For You feed drops it, and compare it with the median of your last ten. X's published code shows a cold-start lift for new and small authors but no multiplier for Premium, images or links.

Does X Premium increase impressions?

Not according to X's published weights, which contain no Premium multiplier. Subscription status appears only as one of the details loaded with each post, and a filter hides subscriber-only posts from viewers who cannot access them. Claims of 20 to 50 percent more impressions, or 2 to 4 times in the first hour, come from vendor pages that show no data. Compare your own medians before and after paying.

Do links reduce impressions on X?

In the weights file, opening a link scores a positive 0.2 and no link penalty appears. X does not publish every file, and some vendors report lower reach when a post contains a link, so settle it on your own account: ten posts with the link in the body, ten with it in the first reply, each read at 48 hours, then compare the medians.

How often should a small brand post on X to get more impressions?

X sets no minimum. Metricool's 2024 data averaged 12 posts a week, with the largest accounts near 95, and X's code decays repeated posts by one author in a feed. Start with one post a day or three to five a week, change one step, and read the median impressions at 48 hours for a month.

Why are my X impressions so low?

Check five causes: a small following, since posts reach followers first; posts older than 48 hours, which For You drops; replies, which do not reach the feeds of people who do not follow you; off-topic posts that earn mutes or blocks; and a platform trend, since one 2026 dataset shows average impressions per post down 5 percent year over year. Compare each post with your own median.

Related practical guides

Log your last ten posts at 48 hours

Add each post's impressions and reactions to the log, find your median, and pick one of pace, format or link placement to change.