How-To Guide
How often should a brand post on X without losing reach?
Guides say 1 to 25 posts a day on X. See Hootsuite, Buffer, Sprout and Metricool beside X's own limits and ranking code, then run a four-week test with a log.
If you post less than once each weekday, begin with five original posts a week and build a sustainable routine. When ready to test, compare two original posts a day with three and count replies separately. Two a day is the low end of what Hootsuite recommends and close to what accounts really post on X: Metricool measured an average of 12 posts a week, Hootsuite 18.1 a week for businesses in Q1 2025, and Socialinsider 70 a month for brand accounts in 2025. On reach, the nearest analysis pools Facebook, Instagram and X: Buffer observed an association between posting frequency and follower growth, while reach per post tended to decline at higher frequencies. The association does not establish cause. We found no X-only test, so the number to watch is your own median impressions per post: step up to 3 a day for two weeks, and keep the step only if that median holds.
- Build a sustainable floor first; two posts a day is a later test starting point. Count replies separately.
- Guides range from 1 to 25 a day, but every figure except one scheduling vendor's sits between 1 and 5, and three of the four measured averages sit between 2 and 3. Buffer's pooled observational analysis found reach per post tended to decline at higher frequencies.
- X publishes limits (50 original posts and 200 replies a day for unverified accounts) and ranking documentation, but no recommended frequency.
How often a brand should post on X
Once you can sustain a routine, use 2 original posts a day, or 10 to 14 a week, as a test starting point; add replies on top. If you currently post less than once each weekday, begin at five per week first. Two a day matches the low end of Hootsuite's 2 to 3 a day and sits in the middle of the measured averages on X: Metricool found 12 posts a week across 23,561 accounts in 2024, Hootsuite found businesses posting 18.1 times a week on X in Q1 2025, and Socialinsider found brand accounts publishing 70 posts a month on X in 2025. Those are averages across other accounts, not a test, so treat the starting point as a place to begin and not a target.
The table gives the routine and what each number rests on. The first three rows are our choices, drawn from the overlap of the sources below. Only the limits row comes from X.
| Part of the routine | Starting point | What it rests on |
|---|---|---|
| Floor | 5 original posts a week, one each weekday | Our choice. Buffer's within-account analysis of 4.8 million channel-weeks on Facebook, Instagram and X together found that accounts that skipped a week underperformed their own baseline growth (Buffer). |
| Starting count | 2 original posts a day | Our choice: Hootsuite's low end, Sprout Social's reported average of two, and the 2.6 a day implied by Hootsuite's measured 18.1 a week. |
| Step to test | 3 original posts a day for two weeks | Our choice: the top of Hootsuite's range and the low end of Buffer's 3 to 4 a day. |
| Spacing | Example: 3 hours between posts | Our untested scheduling example; a three-hour gap has not been shown to improve reach. |
| Replies | Counted on their own line, not in the 2 a day | X's limits page counts them separately: 50 original posts and 200 replies a day for unverified accounts (X Help, archived copy). |
| Ceiling | The count where median impressions per post fall against your previous four weeks | Your own data. Check the current limits before planning a high-volume routine. |
No X-only test of posting frequency turned up. The nearest analysis is Buffer's, which pools three platforms and observes an association between posting and follower growth while reach per post tends to fall. It does not establish causation. The next section sets the figures side by side so you can see why guides range from 1 to 25 a day.
Recommendations and measured averages share one axis, so the overlap is visible. Original diagram, drawn from the source table in this article.
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.
Why guides say anything from 1 to 25 posts a day
The figures differ because they answer different questions. Some are recommendations from expert opinion, some are averages of what accounts post, and one is a scheduling vendor's number. Put the kind of figure next to the number and most of the disagreement goes away.
| Source | Posts a day | Kind of figure | What it rests on |
|---|---|---|---|
| Posthell | 1 to 3, plus replies | Recommendation | No data cited; written for founders |
| Hootsuite | 2 to 3 | Recommendation | A theoretical approach: best practices, platform executives and its network of social media managers |
| BulkPublish | 2 to 5 | Recommendation | No test cited; the page says X publishes no recommended frequency |
| Buffer | 3 to 4 | Recommendation | Cites a study of 30 top X accounts that found posting varies widely; the same section later says 3 to 5 |
| Reachmore | 3 to 5 | Recommendation | No test cited for its tiers; its example account is unsourced |
| Circleboom | 15 to 25 | Recommendation from a scheduling vendor | No data cited; it also credits the 2,400 a day limit to Premium accounts, which the X page we read does not say |
| Sprout Social | 2 (average) | Measured industry average | Sprout says brands averaged three to four posts a day on X in 2022 and now average two |
| Socialinsider | About 2.3 (70 a month) | Measured average | Brand accounts on X in 2025, from a dataset of 198 million posts from 1.16 million brand accounts across four platforms |
| Hootsuite | About 2.6 (18.1 a week) | Measured average | Businesses on X in Q1 2025; by industry on X, retail averaged 7.2 a week and technology 7.6 |
| Metricool | About 1.7 (12 a week) | Measured average | 23,561 accounts and 2,144,853 posts, 2024; the largest accounts averaged almost 95 a week |
Three things stand out. First, every figure except one lands between 1 and 5 a day, and three of the four measured averages sit between 2 and 3. Only a scheduling vendor reaches 15 to 25, with no data behind it.
Second, bigger accounts post more, and that does not show that posting more makes an account bigger. Metricool says the largest accounts post the most and that increasing frequency "could positively impact" follower growth, but a study of different accounts cannot separate cause from size.
Third, volume and engagement do not move together at the platform level. Rival IQ found brands cut their tweet frequency by about 15% in the year behind its 2024 report while engagement per post fell about 20%. Socialinsider found X posting rose from 50 to 70 a month between 2024 and 2025 while the average engagement rate slipped from 0.15% to 0.12%. Engagement per post fell in a year of less posting and in a year of more, so neither period shows frequency as the cause.
| Source and period | Posting volume | Engagement per post |
|---|---|---|
| Rival IQ, report published February 2024 | Down about 15% | Down about 20% |
| Socialinsider, 2024 to 2025 | 50 to 70 posts a month | Average rate 0.15% to 0.12% |
The one test that compares accounts with themselves is Buffer's, and it is not X-only. Buffer modelled 4.8 million channel-weeks from about 161,000 profiles on Facebook, Instagram and X, and found that weeks with no posts underperformed an account's own baseline growth. Accounts posting 10 or more times a week gained 32 more followers a week than in silent weeks. Buffer also reports that its engagement-rate analysis of 15.7 million posts suggests reach per post tends to decline at higher frequencies. In this pooled observational analysis, higher posting frequency coincided with higher follower growth and lower reach per post. The association does not establish that posting more caused either result.
That tension is why the routine below tracks median reach per post and not totals.
What X itself publishes about posting frequency
X publishes limits and ranking rules, and no recommended number of posts. BulkPublish says the same, and we found no X page that gives one. What X does publish sorts into three groups, and each one bears on cadence in a different way.
| Group | What X's pages say | What it means for cadence (our reading) |
|---|---|---|
| Ceiling | Unverified accounts get 50 original posts and 200 replies a day, and the page says the daily limit is broken into smaller limits for half-hour intervals (X Help, archived copy of 26 September 2026; the live page refused our fetch). Lower down, where it explains what happens once you hit a limit, the page names 2,400 updates a day instead and never says which figure applies to whom. | The cited page also describes smaller half-hour limits, so a high-volume schedule should check them. It does not reconcile the other limit figures on this page. |
| Counting rule | The limits include actions from the web, mobile and the API (X Help). | A scheduler's posts use the same allowance as posts typed in the app. |
| Ranking | The For You README describes an author-diversity discount, a decaying factor on each post after an author's first, down to a floor, and says posts older than 48 hours are dropped before scoring (xai-org/x-algorithm). | The inspected code excludes posts older than 48 hours before scoring; younger posts must also pass other filters. In the current code the author discount is a switchable setting whose live values are not published, so we do not build a spacing rule on it. |
| Negative signals | Mute, block, report and not interested are predicted and carry negative weights (xai-org/x-algorithm). | A post that makes people mute you counts against that post, so padding to reach a count has a cost. |
| Reach by follow status | Posts from accounts the viewer does not follow are multiplied by a factor below 1, and recent posts from followed accounts come from a separate store (xai-org/x-algorithm). | Your followers see your posts by one route and everyone else by another, so a count that suits your followers does not fix reach beyond them. |
The README places the author discount among the adjustments made while one viewer's candidates are ranked. It says nothing about posting times, and the age filter removes posts older than 48 hours; a younger post must still pass other eligibility filters. The code we read in November 2026 also treats the discount as a setting that can be switched off, and X does not publish the values it runs with. So the discount tells you little about cadence. Spacing rests on a plainer reason, which step two covers: your followers are online at different hours.
Where to start for your account
The routine has one shape, but the place to start depends on where your account is today. Find the row that sounds like you, start there, and then follow the five steps below in order.
| Your situation | Where to start | What to check first |
|---|---|---|
| You post less than once a weekday now | The floor: 5 original posts a week, held for eight weeks | Whether you can keep it without a silent week |
| Your posts usually get under 100 impressions each | 2 a day at most, and no step-up test yet | What the posts say and who they reach, before how many there are |
| One founder writes every post | 2 a day, and step up only with two weeks of extra posts already drafted | The stock check in step four |
| You already post 3 or more a day | Your current count as the baseline, then two weeks at one post a day fewer | Whether median impressions per post rise when you post less |
| A launch, a news event or a holiday falls in the next four weeks | Wait and start the test after it | Nothing yet: those weeks move reach for reasons that have nothing to do with the count |
The rows are our routing, not a study's tiers. The under-100 row comes from the Buffer figure in step one; the rest follow from the test design below.
Pick a weekly floor and hold it for eight weeks
Begin with 5 original posts a week, one each weekday. If that is more than you post now, keep that pace for eight weeks before you test a higher count. If you already post that often, treat it as the line you never drop below during the test. Quiet weeks are the one pattern Buffer found across platforms: in its within-account analysis of 4.8 million channel-weeks, accounts that did not post in a given week underperformed their own baseline growth rates, and any posting beat none. Buffer pooled Facebook, Instagram and X and does not break the result out for X. It also says no single number of posts a week works across platforms, niches, account sizes or teams.
Check where your account stands before you add volume. In Buffer's analysis of 18.8 million X posts, regular accounts typically got under 100 impressions per post and the median engagement rate was 0%, so at least half had no visible interaction on a typical post. If your posts sit there, a third daily post is not the first thing to change. Look at what the posts say and who they reach first.
Spread posts across the day
Three hours between original posts is an example schedule, not a tested reach rule. Choose practical times for your audience and keep the existing slots during your comparison. The inspected X ranking code excludes posts older than 48 hours before scoring; a younger post still has to pass other filters, and current production settings are unknown.
For example, keep 9:00 and 14:00 in both periods and add an 18:00 slot in the three-post period. Choose hours from your own analytics. The added slot and ordinary changes in topic or audience are possible confounders, so this is a practical comparison rather than proof of cause.
If you schedule posts, remember that scheduled posts count against the same X limits as posts typed in the app. A queue is a way to keep the spacing, not a way around the limits.
The first two slots stay put and only 18:00 is added; the gaps are an example schedule, not a tested rule. Original diagram, based on the schedule example in this article.
Keep replies in a separate count
Keep replies out of your original-post count. X's limits page counts them separately, and they serve a different job: answering people who talk to you or to your niche. Mixing them into one number hides whether your original posts are working.
Replies to comments on your own posts are worth doing for another reason. Buffer estimated an engagement lift of 8% on X when the account replied to comments, the second smallest of six platforms. Buffer compared each account with itself and says it cannot prove that replying causes the lift, because good posts attract comments and the replies follow. Treat it as a reason to reply, not as a number to plan on.
| Count | Where to log it | Why separate |
|---|---|---|
| Original posts | Test log, column D | The thing you are testing |
| Replies | Test log, column E | Reply volume depends on who writes to you, not on your plan |
| Threads | Count each thread once, in the original-post column | Keep the number of threads the same in both arms of the test |
Test one step up for two weeks
Post at two a day for two weeks, then three a day for two weeks. Keep the two existing slots and use similar formats and topics; the extra post requires an additional slot. This example is a descriptive comparison, not a controlled causal test.
| Weeks | Original posts a day | Spacing | Keep the same |
|---|---|---|---|
| 1 and 2 | 2 | Morning and afternoon, 3 or more hours apart | Similar formats and topics, existing slots, reply habit |
| 3 and 4 | 3 | Morning, afternoon, evening | Similar formats and topics, existing slots, reply habit |
Export your last 90 days from X analytics before week 1 so you have a baseline. Use the median, not the average, because one post that travels would otherwise hide the rest. The test log has a column for the median impressions per post of the previous four weeks and a ratio against it.
Check your stock before week 3 starts. Three a day for two weeks is 14 more original posts than two a day, one extra for each of the 14 days. Draft those 14 in advance, held to the same standard as the rest. If you cannot name them yet, the step-up would measure rushed posts and not the count, so stay at 2 until you can.
Weeks with a news event, a launch or a holiday are poor test weeks. If one falls inside a test, mark it in the notes and extend the arm by a week.
Keep the extra post only if the median held
Keep 3 a day if median impressions per post stayed at or above your previous four weeks and every post met your own quality bar. If the median fell, go back to 2. That rule is ours, not a published threshold.
Compare median impressions per post as well as total impressions and the work needed. The test compares each two-week period with the preceding four weeks, so timing, topics and other changes can affect the result. A stable median does not prove the additional impressions came from new people, and a lower median does not prove the extra post reduced the reach of the others.
The routine does not need X Premium to stay far inside X's limits. Buffer observed that Premium accounts had about 10 times the median reach per post of regular accounts; that comparison does not establish that Premium caused the difference. Test cadence within one account type, and do not switch plans in the middle of a test.
For the definition of cadence, see posting cadence. For turning the routine into a month of posts, the weekly content plan and the content calendar guide show how to keep the schedule going past week three. For what to put in those slots, see a weekly X content mix for brands.
Worked example: a four-week test on one founder's account
The account and every number in this example are invented to show the arithmetic. They are not a result from a real account or from Superdirector.
A founder runs the X account for her budgeting app alone. She has held the floor of one original post each weekday for two months, so she starts the test at 2 a day. Her median impressions per post over the four weeks before the test was 180. Before week 3 she drafts the 14 extra posts the step-up needs.
| Period | Original posts a day | Posts published | Median impressions per post | Against the four weeks before | Total impressions in the period |
|---|---|---|---|---|---|
| Four weeks before the test | 1, weekdays only | 20 | 180 | Baseline | 4,300 |
| Weeks 1 and 2 | 2 | 27 (one missed) | 184 | 184 / 180 = 1.02 | 5,600 |
| Weeks 3 and 4 | 3 | 41 (one missed) | 166 | 166 / 182 = 0.91 | 7,900 |
The 182 in the last row is the median of the four weeks before week 3: the last two weeks of the baseline plus weeks 1 and 2. Her replies sit in their own column of the log and are left out of every number above.
Read the totals first and the step-up looks like a win: total impressions rose from 5,600 to 7,900, about 41% more. The median per post tells a different story. It fell about 9% against the preceding four weeks; that does not show whether the extra post caused the difference or reached new people. By the rule in step five she goes back to 2 a day, which held its median in weeks 1 and 2, and saves the best of the extra posts as ideas for her next test.
The totals rise while the median per post falls, so the test keeps the lower cadence. Original diagram, based on the worked example and test log in this article.
Where Superdirector fits
Superdirector, which publishes this guide, is for founders, app owners and DTC brands who want a channel run in their brand's direction. Its main offer is $199 a month for one AI-run social account. It can prepare X text posts, and an eligible single link-free X post may be published under the account's standing posting permission, subject to account readiness and applicable pricing. Confirm the account and delivery scope before paying. Replies stay with you. The $29 a month plan supports hands-on work from Claude or ChatGPT; check current plan entitlements before choosing.
The site headline is "Autonomous social media growth.*" with the note "* AI video and full autopilot are coming soon," and you remain the publisher of whatever goes out. See what runs today.