The best time to post on X (Twitter), from 10,699 posts

Every guide to posting on X recommends a time. Most recommend midweek mornings and say to avoid weekends and evenings. If you run company or client accounts, you may use that advice to time your best posts. But it comes from adding up engagement across millions of posts, which favors the hours when large, busy accounts post.

I compared recent posts from 236 public X accounts across 85 niches with each account's usual views. A brand with millions of followers and a founder with a few thousand count the same. A post scores high only if it beats that account's usual views. By this measure, the hour barely matters, and weekends do better despite most guides' advice.

At a glance

Weekend posts did better than weekday posts

Saturday and Sunday posts got about 8% more views than weekday posts from the same accounts. They were the two best days, and weekend mornings did best on the heatmap.

I'd trust this result most because I chose the test before seeing the data. Buffer, Sprout Social and SocialPilot all call weekends the worst time to post on X. I compared weekends with weekdays and tested their other claims. I also checked whether the weekend lead held:

• Without viral posts: the lead was about 9%.
• Likes instead of views: the lead was about 7%.
• Accounts with 10,000 to 100,000 followers, and larger accounts: both groups led by about 8%.
• Days with the same number of posts: weekend posts still did better. Accounts post less on weekends, but the lead held when comparing days with the same post count.

Accounts post less on weekends. Fewer than one post in five went out on Saturday or Sunday, compared with closer to three in ten if posts were evenly spread. Weekend posts face less competition, and people have more time to scroll. But accounts may save better posts for weekends, so the day itself might not explain the lead.

If you only post on weekdays, try one of your better posts on Saturday or Sunday morning and compare its views with your usual results.

The hour you post barely matters

No time of day helped enough to plan around. After smoothing across nearby hours, results stay within about 4% of average. The six four-hour blocks are all within about 3%.

Close to half the posts in every time block beat their account's usual views, from overnight to evening.

Early afternoon, noon to 4pm, had the most posts and did worst, about 3% below the rest of the day. I noticed this only after comparing all six blocks, so I wouldn't change your schedule for it.

Tuesday to Thursday trailed the rest of the week

Posts from Tuesday to Thursday did about 4% worse than posts on the other four days. Buffer, Sprout Social and SocialPilot all recommend these days. Thursday did worst and had the most posts.

Thursday did poorly mostly during working hours. About 8am to 5pm ET on Thursday was the worst period on the heatmap, while late evening looked fine.

Sprout Social recommends Tuesday to Thursday from noon to 6pm, which did about 5% worse than all other times. Buffer's top choice, Tuesday at 9am, was exactly average.

Scheduled posts did slightly worse

About one post in eight went out in the first minute of an hour. People rarely publish manually at exactly 2:00, so most were scheduled. They got about 7% fewer views than the same accounts' other posts.

Scheduling may not be the cause. People often schedule planned posts, such as launches, webinars and blog links. Those might do worse whenever they're posted; this data can't tell. The gap in likes was smaller and could be chance. If you schedule everything, try posting off the exact hour to look less scheduled, but don't expect much improvement.

How this compares to what's already published

• Buffer studied millions of posts sent through its tool. It recommends Tuesday at 9am, weekday mornings from 9am to 11am as most reliable, and Wednesday as the best day. Saturday and evenings from 6pm to 11pm rank worst. Times are local to each poster.
• Sprout Social studied engagement across about 307,000 profiles on six networks. It recommends Tuesday to Thursday from noon to 6pm for X, with Saturday and Sunday ranked worst.
• SocialPilot studied about 700,000 posts scheduled through its tool. It recommends weekday mornings and 3pm Eastern, with Tuesday to Thursday ranked best and weekends worst.
• Social Champ studied 3.5 million posts. Unlike the others, it recommends Sunday from 9am to 10am, followed by Monday lunchtime.

These results disagree with the first three on weekdays and weekends, but agree with Social Champ about Sunday. Sunday at 9am also did well here, but with only 42 posts, there's too little data to say more. None of these guides gives a margin of error.

The methods answer different questions. Total engagement across millions of posts favors hours when more people are online and more accounts post, especially large, busy accounts. Comparing each post with its account's usual views tells you more about when to post. You can't change who else posts at 9am on a Tuesday.

How I measured it

The sample contains 10,699 posts from 236 accounts, between three days and a year old. Collecting them through two AnyAPI X endpoints took 1,576 calls and cost $0.75.

1. Find accounts. I ran one twitter.search_users search per niche across 85 niches, including SaaS founders, growth marketers, nurses, golfers and coffee roasters. Keeping up to five accounts per niche with at least 500 followers and 300 posts gave 423 accounts.
2. Get each account's posts. I fetched up to four pages of twitter.user_tweets per account, newest first, with about 20 posts per page and their times, views, likes and reposts. I stopped at posts older than a year.
3. Keep only the account's own posts. I removed 4,588 replies, 4,161 reposts and 37 pinned posts. One more twitter.user_tweets call per account identified pinned posts using a source that marks them. Every post had a time and view count, so none needed a separate twitter.tweet lookup.
4. Allow time for views. I removed 2,668 posts less than three days old and 1,211 older than a year. I also removed accounts with fewer than eight posts left or identical view counts on every post. That left 236 accounts and 10,699 posts.
5. Compare each post with its account. I divided views by the account's median and capped scores at 6x either way so one viral post counts once. 308 posts reached the upper cap.
6. Group and compare. I converted times to US Eastern, grouped posts by weekday, hour and four-hour block, and took each group's geometric mean score. I resampled each comparison by account 1,000 times to check how much different accounts could change it. I recorded the seven guide claims before calculating any results.

This request gets one account's posts. The response shows only the fields used:

curl -sS -X POST https://api.getanyapi.com/v1/run/twitter.user_tweets \
  -H "Authorization: Bearer $ANYAPI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"handle":"HubSpot"}'

{"output":{"found":true,"data":{"nextCursor":"cur_...","tweets":[{"id":"2114778098956739016","createdUtc":1793037364,"views":2953,"likes":15,"retweets":0,"replies":9,"isReply":false,"authorHandle":"HubSpot"},"..."]}},"costUsd":0.00045}

The sample by follower count: 155 accounts over 100,000, 68 between 10,000 and 100,000, 12 between 1,000 and 10,000, and 1 under 1,000.

The limits of this data

• Total views, not first-day views. I collected view counts three days to a year after posting. They don't show how posts did in their first hour.
• US Eastern for everyone. The account search returned no location, so I used Eastern time throughout. Results for accounts in London or Lagos are offset by their time difference.
• Mostly big accounts. People search finds established accounts. Two thirds of the sample has more than 100,000 followers; only 13 accounts have fewer than 10,000. The weekend lead held in both larger groups, but tells you little about small accounts.
• Better posts may go out on better days. Accounts may save their best posts for weekends, making weekends look better regardless of the day.
• More recent posts. Four pages holds at most about 80 posts. For accounts posting several times daily, that covers only the last few weeks.
• Accounts found through search. The 85 English niche searches sometimes returned fan or news accounts instead of people in that niche. Different searches could change an effect this small.
• Views and a likes check only. I didn't measure replies-per-view, link clicks or follower growth.

The weekend lead is about 8%. Different accounts could put it between about 4% and about 12%. Hourly differences are smaller.

What to do with it

1. Post at an hour that suits you. No hour consistently did better.
2. Save one of your better posts for the weekend. Weekend posts did about 8% better, while most accounts posted less.
3. Don't plan around Tuesday to Thursday mornings. The big guides' recommended times did about average or worse.
4. Post less during Thursday working hours if you can. This period did worst and had the most competition.
5. Keep scheduling if you want, but don't expect it to help. Posts sent exactly on the hour did slightly worse, probably because of their content.
6. Check your own account. Get your posts with twitter.user_tweets. Remove replies, reposts and posts less than three days old. Divide views by your median and compare weekends with weekdays. One page of about 20 posts costs $0.00045.

The Threads study, LinkedIn study and TikTok study use the same method. Use the X endpoints to check your account or a client's.

Use data responsibly and follow AnyAPI's Acceptable Use Policy.