How-To Guide
What is the best time to post on X, and how do you find yours?
Six studies of the best time to post on X disagree on the hour and the day. See where and why, then run a four-week test of your own slots with a log.
By the Superdirector editorial team. The study pages and X's developer documents cited here were read on November 7, 2026.
At a glance
What is the best time to post on X, and how do you find yours?
On X the studies do not just differ, they invert: Buffer calls 6 to 11 p.m. the worst time to post and SocialBee calls 7 to 11 p.m. the best. Sprout Social favors Tuesday to Thursday afternoons and Eden early morning and late evening US Eastern. Eden puts the whole spread between its best and worst hour at 2.5 percentage points, so the useful move is a test on your own account: score each post's impressions against your median across three slots and four weeks.
- Eden: timing is worth 2.5 points of breakout odds, attaching media 6.7.
- Read each post's impressions after seven days, once counts settle; X's API keeps non-public metrics such as engagements and link clicks for only 30 days.
- Sprout's summary names Tuesday to Thursday as best days while its own day table says Monday to Thursday.
A post's score is its impressions after seven days divided by the median of your previous twenty posts, the measure Eden and SocialBee take against an account's own baseline. Original diagram with made-up numbers, from this page's method and Eden's seven-day read.
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
The studies on the best time to post on X do not just differ, they invert. Buffer calls 6 to 11 p.m. the worst stretch and puts its peak at 9 a.m. on Tuesday. SocialBee calls 7 to 11 p.m. the best and finds the worst hour near 1 p.m. Sprout Social favors Tuesday to Thursday afternoons, and Eden favors early morning and late evening US Eastern, with Sunday the best day. They cannot all be right for your account, so test it on your own.
Eden's data also shows how little is at stake. Between its best and worst posting hours the odds of a post doing twice as well as its author's median move by 2.5 percentage points, and attaching media (an image, video or GIF) moves them by 6.7. The test below asks you to score each post's impressions against your own median, put twelve comparable posts into three slots over four weeks, and expect a small answer.
The X pages we could read do not name a best hour. The component list in the open-source repository for X's recommendation system says about half of the posts in For You come from the in-network source, meaning accounts you follow, and that a neural-network ranker picks timeline posts from the candidates. It names no posting-time setting. If timing matters, it is likely because of which of your followers are online when a post first goes out, which is a reading, not an X statement.
The steps run in this order: your baseline, the six studies, why they conflict, the four-week test, a worked example with made-up numbers, and the decision rule.
Set your own baseline on X
Your own data is the only evidence about your own followers, so set it up first. You need one number per post and a way to keep it. Where to read it:
- Post analytics. Roboin's March 2026 guide says to open the menu on your own post and choose View post analytics. Fedica lists most active times among the things X Analytics shows, and Buffer says X's own analytics sit behind X Premium. Check what your account shows.
- Impressions. X's developer docs define an impression as the post appearing on a user's screen, counted again if the same person sees it twice. Impressions are listed among the public metrics, visible to anyone through the API.
- The 30-day limit. The same docs say non-public, organic and promoted metrics are only available for posts created within the last 30 days. Record each post's numbers in a sheet before then. The docs attach the limit to those three sets and not to public metrics.
What X reports for each post, from its developer docs, splits into public and non-public, and the split decides what you can log later:
| Metric | Public or non-public | What it counts |
|---|---|---|
| Impressions | Public | Times the post appeared on a screen, repeat views included |
| Likes, reposts, quotes, replies | Public | Interactions from anyone; quote posts are always organic |
| Bookmarks | Public | Times the post was bookmarked |
| URL clicks, profile clicks, engagements | Non-public | Only for your own posts, and only for 30 days after posting |
Pick impressions as the timing measure and engagements as a second column if you can see it. A worked example, with made-up numbers: if your last twenty posts have a median of 800 impressions and a Tuesday 9 a.m. post reached 1,040, its ratio is 1.3; a Wednesday 2 p.m. post that reached 640 scores 0.8.
Read each post after seven days, not the same day. Eden reads every post at least seven days after it went out so counts have settled; we copy that rule. Impressions are the right measure for timing because they ask who saw it. Likes and replies mix in how good the post was. For the difference, see reach versus impressions.
Then score each post as its impressions divided by the median impressions of your previous twenty posts. A ratio of 1.0 is your normal and 1.3 is 30% above it. SocialBee and Eden also measure against an account's own baseline, which is why their numbers can be compared across big and small accounts.
Line up six studies by sample, metric and clock
Before you trust a headline hour, check three things: what was counted, over what window, and on which clock. The first table sets them out page by page, and the studies sheet holds the same rows plus the day each page was read.
| Study | Sample and window | Metric and clock |
|---|---|---|
| Sprout Social | About 307,000 profiles on six networks, X among them, and nearly 2 billion engagements; window 27 November 2025 to 27 February 2026 | Engagements; local time of each audience |
| Buffer | More than 8.7 million tweets sent through Buffer, page of Mar 13, 2026 | Engagement rate; local time |
| SocialBee | 707,338 posts sent through SocialBee in the first half of 2026 | Impressions against each account's own typical post; local time |
| Eden | 7,627,422 tweets from 14,671 tracked creators, Aug 2025 to Aug 2026 | Odds of 2x the author's median likes; US Eastern |
| Brandwatch | No sample stated; published Nov 7, 2024 | Engagement guidance; no clock stated |
| Metricool | 23,561 accounts and 2,144,853 posts, 2024 study | When the most users are connected; global |
Two of the six are old. Brandwatch's page dates from November 2024 and Metricool's study from 2024, and Metricool's best time is the hour with the most users connected, not how posts performed at that hour. We show them because they are still ranking pages, and we weight the four 2026 studies more.
Three more sources are left out of the tables because they give no hour to compare. Hootsuite's calculator builds a schedule for your industry and time zone from its own data on an Eastern clock. OneUp's heatmap says only that it analyzed millions of posts. Metricool's 2026 study of 1,123,528 X posts from 15,116 accounts reports no best hour on the page we read.
| Study | Best days | Worst days | Best hours | Worst hours |
|---|---|---|---|---|
| Sprout Social | Tuesday to Thursday (its day table says Monday to Thursday) | Saturday and Sunday (its day table says Saturday) | 12 to 6 p.m. | Not stated |
| Buffer | Wednesday, Tuesday, Thursday | Saturday, then Friday | 9 a.m., with 8 to 11 a.m. strong | 6 to 11 p.m. |
| SocialBee | Sunday, then Saturday | Thursday | 7 to 11 p.m.; Sunday 7 p.m. best | About 1 p.m.; Thursday 1 p.m. worst |
| Eden | Sunday, then Saturday | Wednesday and Thursday | 5 a.m. best; 4 to 8 a.m. and 8 p.m. to midnight ET | 3 p.m. worst; 10 a.m. to 6 p.m. ET weak |
| Brandwatch (2024) | Wednesday and Friday | Not stated | 8 a.m. to noon and 5 to 6 p.m. | Not stated |
| Metricool (2024) | Wednesday, then Tuesday and Thursday | Not stated | 9 p.m. peak of connected users | Not stated |
See where the studies conflict and why
Three clashes stand out. Evenings: Buffer's worst and SocialBee's best. Afternoons: Sprout's best window, and the weakest hours in both SocialBee and Eden. Weekends: Saturday is the worst day for Sprout and Buffer and the second-best day for SocialBee and Eden. They cannot all be describing the same account.
The likeliest reasons are measurement choices, not a changed platform:
- Metric. Buffer scores engagement rate. SocialBee scores impressions against each account's own average. Eden scores likes against each author's own median, and argues that guides averaging raw engagement mostly measure when big accounts post. Sprout's page reports engagements and does not say it adjusts for account size.
- Sample. Buffer and SocialBee only see posts sent through their own tools. Eden follows creators it chose to track, and says its sample skews to established accounts.
- Clock. Three studies use each audience's local time. Eden uses US Eastern, which is the wrong clock for an audience elsewhere.
- Why. SocialBee says its data cannot explain why Sunday performs better, and Eden says its data is observational, so a 5 a.m. post may differ from a 3 p.m. post in ways other than timing.
Eden reads its own curve as a crowding effect: posting volume piles up at midday, so a post sent in a quiet hour faces less competition. That is a hypothesis, not a finding. Eden also found the overnight edge reverses for accounts under 1,000 followers, whose readers are mostly asleep then.
Five studies' named best and worst hours, each on its own clock, with the two tinted evening cells shown because one study calls them best and another worst. Original diagram, from the study pages read on November 7, 2026.
Test three slots over four weeks
Check first that you can run it. You need three posts a week of similar weight for four weeks, a way to read each post's impressions, and the time zone most of your audience lives in. If you cannot keep three posts a week going, settle the pace first: with fewer posts, each slot gets too few tries to compare.
Twelve posts is a small test, but it gives each slot four tries. Pick three slots that put the studies' disagreement into your own account, in your audience's local time:
| Slot | When | Whose claim it tests |
|---|---|---|
| A | Tuesday, 9 a.m. | Buffer's best slot |
| B | Wednesday, 2 p.m. | Sprout's afternoon window, and Eden's worst pocket |
| C | Sunday, 7 p.m. | SocialBee's best slot |
- Post four times in each slot over four weeks, rotating the order each week so no slot always follows a weekend.
- Keep the post type the same across slots: all text, or all with an image. Eden found media worth more than timing, so mixing types would hide the answer.
- Use posts of comparable weight: not a launch announcement in one slot and a quick thought in another.
- Publish every test post the same way: all from the same scheduler, or all by hand. Then the slot is the only thing that changes between them.
- Seven days after each post, copy its impressions into the test log.
- Enter each post's ratio from step 1, its impressions divided by your median of the previous twenty, and compare the three slot medians.
If you post from a scheduler, set the slot there. PostPlanify says X has a free built-in scheduler on desktop, and the steps are in our guide to scheduling posts on X.
Enter the slots in your audience's time, then convert them to the clock your scheduler uses. If most of your audience is in New York and you schedule from Berlin, Berlin is usually six hours ahead, so slot A, Tuesday 9 a.m. in New York, goes in as Tuesday 3 p.m. Berlin time. The US and Europe change their clocks on different dates, so for a week or more around each change the gap is five hours; check the slots again in March and in late October. For more on mismatched clocks, see aligning post time with analytics time zones.
Worked example: score a four-week test
Here is the test run through once, with made-up numbers so the arithmetic is easy to follow. A small skincare brand's audience is mostly on US Eastern time, and the median impressions of its previous twenty posts is 800. A real median drifts as new posts come in; to keep the sums visible, this example holds it at 800 for the whole month.
| Slot | Impressions after 7 days | Ratios (impressions / 800) | Slot median | Posts above 1.0 |
|---|---|---|---|---|
| A: Tuesday 9 a.m. | 1,040; 760; 784; 816 | 1.30, 0.95, 0.98, 1.02 | 1.00 | 2 of 4 |
| B: Wednesday 2 p.m. | 640; 720; 880; 704 | 0.80, 0.90, 1.10, 0.88 | 0.89 | 1 of 4 |
| C: Sunday 7 p.m. | 1,120; 960; 1,040; 680 | 1.40, 1.20, 1.30, 0.85 | 1.25 | 3 of 4 |
Slot C, Sunday 7 p.m., has the highest median at 1.25 and beat the brand's normal on three of its four posts, so it passes both parts of the rule in the next section. Slot A sits at its normal: one strong post and three close to 1.0. Slot B fell below normal on three of its four posts.
So the brand moves its most important post of the week to Sunday evening, keeps the other two slots for lighter posts, and runs the check again next month. Four posts per slot is a thin sample, and one unusual post can swing a slot's median, so the result is next month's default, not a permanent answer. Had the three medians come out at 1.02, 0.97 and 1.05, no slot would pass and the brand would keep its current schedule.
Only slot 3, Sunday 7 p.m., clears both parts of the decision rule on the made-up numbers from the table above. Original diagram, from this page's worked example.
Decide and lock in a schedule
After four weeks you have three medians. A rule of thumb that is ours, not any study's:
- Move your default slot only if one slot's median ratio is clearly above the other two and it beat your baseline on at least three of its four posts.
- Treat timing as noise if the slots are close or the order flips by week. That is the likeliest result if Eden's 2.5-point gap holds for you, and it frees you to spend the effort on the post itself.
- Under 1,000 followers, Eden advises small accounts to post when their own readers are online and spend the effort on replies.
Three things can spoil the result:
- A launch, sale or holiday in one week. It lifts whichever slot it lands on. Fix: drop that week from the log and add a fifth week at the end.
- Reading too early. A count taken the same day mixes a post's first hours with its settled total. Fix: read every post at seven days.
- Clock drift. A slot saved in your own time zone moves against your audience's when either side changes its clocks. Fix: recheck the scheduler's slots after each change.
Whatever you choose, hold it for a month before changing it, and keep the pace steady. See posting cadence, and batch a month of content to turn the winning slot into a calendar. For the Instagram studies, see the best time to post on Instagram.
Where Superdirector fits
Superdirector, which publishes this guide, does not run timing tests; the test above is yours to run. Its headline is "Autonomous social media growth.*", and the asterisk matters: AI video and full autopilot are coming soon.
Its main offer is $199 a month for one social account that its AI runs in the direction you set. It finds the posts and formats taking off in your niche, remakes them as X posts and carousels in your brand, and publishes eligible single X text posts and Instagram carousels once you turn posting on for the account, after one warning, with no per-post approval and nothing out in the ten minutes after you turn it on. The pricing page lists two connected account slots on that plan; a slot lets you connect an account, while the service we run is one account, so confirm which account it will be before you pay. The $29 a month plan is for hands-on work from Claude or ChatGPT. See what runs today.