Data study · 7 min read
What made X posts go viral in 2026
We compared 297,373 original posts from 4,439 creators on X, posted January 1 to October 31, 2026, with each creator's own usual reach. Posts that led with a number (+17%), carried a video (+19%) or two to four images (+16%), or were 500 to 999 characters long (+20%) went viral more often than their creators' usual; a single image (−13%), a link (−23%) and an opening question (−8%) went viral less often.
Joschua Sünneke·Key findings
- Lead with a number. Posts whose opening line held a figure, an amount or a percentage went viral 17% more often than their creators' usual (95% range +15% to +20%). Lists did about as well (+16%), and announcing something you made did best of the openings (+20%, from fewer posts).
- Don't open with a question. Openings that asked a question went viral 8% less often (95% range −12% to −5%), and "here's how" or "why" openings 9% less often.
- Say more. Posts of 500 to 999 characters went viral 20% more often than usual; posts under 100 characters 9% less often. Posts spread over 8 or more lines did +16%; one-liners −7%.
- Video or several images, not one. Video +19%, two to four images +16%, a single image −13%, a GIF −33%. Text-only posts were close to usual (−3%).
- Links cost reach. Posts with a link went viral 23% less often (95% range −28% to −20%).
- Thread openers did best of any group (+61%, 95% range +41% to +86%), but they were only 1,672 posts.
These are relative changes against each creator's own rate, not percentage points: at the sample's 10.4% viral rate, +17% means about 12.2% of posts going viral instead of 10.4%.
Openings: how the first line starts
Openings are sorted by plain rules on the first line (a teaser shorter than 40 characters is read with the line after it). A list is three or more lines starting with a number, dash, bullet or arrow.
Show the numbers
| Opening | Viral posts vs usual | 95% range | Posts |
|---|---|---|---|
| Announces something | +20.4%+8.8% to +31.0% | +8.8% to +31.0% | 3,519 |
| Leads with a number | +17.5%+14.9% to +20.5% | +14.9% to +20.5% | 55,282 |
| A list | +16.3%+11.3% to +20.7% | +11.3% to +20.7% | 18,034 |
| Takes a side | −3.5%−9.9% to +3.2% | −9.9% to +3.2% | 6,161 |
| Starts with I or we | −4.1%−7.8% to −1.3% | −7.8% to −1.3% | 22,580 |
| No clear hook | −5.8%−6.9% to −5.0% | −6.9% to −5.0% | 166,830 |
| Asks a question | −8.4%−11.9% to −4.7% | −11.9% to −4.7% | 21,531 |
| Promises a how or why | −9.5%−18.6% to −1.2% | −18.6% to −1.2% | 3,436 |
We sorted every post by plain rules on its opening line, the same rules the weekly "what's going viral" page uses: a list (three or more lines that start with a number, dash, bullet or arrow), a figure in the opening line, a question, an opening that starts with how, why, what or here's, an opening that takes a side (unpopular opinion, stop, nobody, most people, overrated, is dead), an announcement (launches, introduces, ships), and openings that start with I or we. 56% of posts matched none of these; they went viral 6% less often than usual.
A number does two jobs at once: it promises something specific and it's easy to see in a feed. A question asks the reader to do the work before they know why they should.
Length and shape
Length in characters
Lines with text
Show the numbers
| Length / Lines | Viral posts vs usual | 95% range | Posts |
|---|---|---|---|
| Under 100 | −9.1%−10.8% to −7.3% | −10.8% to −7.3% | 80,903 |
| 100–139 | −7.3%−10.0% to −4.8% | −10.0% to −4.8% | 41,209 |
| 140–199 | −5.4%−8.5% to −2.8% | −8.5% to −2.8% | 40,577 |
| 200–279 | +0.5%−1.9% to +3.3% | −1.9% to +3.3% | 47,254 |
| 280–499 | +9.4%+6.5% to +12.9% | +6.5% to +12.9% | 43,598 |
| 500–999 | +19.8%+16.0% to +23.7% | +16.0% to +23.7% | 25,069 |
| 1,000–2,499 | +16.7%+12.3% to +23.4% | +12.3% to +23.4% | 15,436 |
| 2,500 or more | +30.7%+20.5% to +43.5% | +20.5% to +43.5% | 3,327 |
| 1 line | −7.4%−8.7% to −6.0% | −8.7% to −6.0% | 106,744 |
| 2–3 lines | −2.8%−4.5% to −1.3% | −4.5% to −1.3% | 97,356 |
| 4–7 lines | +11.3%+9.0% to +13.7% | +9.0% to +13.7% | 65,002 |
| 8+ lines | +16.0%+12.4% to +19.6% | +12.4% to +19.6% | 28,271 |
Because every post is compared with its own creator's usual, this isn't longer posts coming from bigger accounts: creators' own longer posts beat their own short ones. The median viral post was 191 characters long, against 176 for all posts. Posts over 2,500 characters did best (+31%), but there were only 3,327 of them.
Media, links and threads
Media
Links
Threads
A thread opener has a 🧵, starts with “1/”, “1.” or “1)”, or says “thread” followed by a colon or a down arrow. Replies aren't in the dataset.
Show the numbers
| Post type | Viral posts vs usual | 95% range | Posts |
|---|---|---|---|
| Text only | −3.1%−4.6% to −1.4% | −4.6% to −1.4% | 116,786 |
| 1 image | −13.3%−15.2% to −11.8% | −15.2% to −11.8% | 98,562 |
| 2–4 images | +16.1%+11.3% to +20.9% | +11.3% to +20.9% | 20,047 |
| Video | +18.9%+16.2% to +21.8% | +16.2% to +21.8% | 60,130 |
| GIF | −32.9%−47.2% to −18.7% | −47.2% to −18.7% | 1,848 |
| No link | +1.9%+1.6% to +2.4% | +1.6% to +2.4% | 270,656 |
| Has a link | −23.4%−27.9% to −19.9% | −27.9% to −19.9% | 26,717 |
| Single post | −0.4%−0.5% to −0.2% | −0.5% to −0.2% | 295,701 |
| Thread opener | +61.5%+40.7% to +85.7% | +40.7% to +85.7% | 1,672 |
Over the year, a single image did worse than no image at all (from June to October the two were about even). Two to four images and video both did better than usual. If a link matters, put it in a reply under the post rather than in the post.
Day of the week
Show the numbers
| Day (ET) | Viral posts vs usual | 95% range | Posts |
|---|---|---|---|
| Monday | +2.2%−0.7% to +4.5% | −0.7% to +4.5% | 41,965 |
| Tuesday | −2.5%−5.0% to 0% | −5.0% to 0% | 45,082 |
| Wednesday | −6.6%−8.7% to −4.0% | −8.7% to −4.0% | 46,520 |
| Thursday | −4.0%−6.0% to −1.6% | −6.0% to −1.6% | 47,794 |
| Friday | −3.1%−5.6% to −0.9% | −5.6% to −0.9% | 46,606 |
| Saturday | +6.5%+3.6% to +9.6% | +3.6% to +9.6% | 36,333 |
| Sunday | +12.6%+9.8% to +15.6% | +9.8% to +15.6% | 33,073 |
Sunday posts went viral 13% more often than usual and Wednesday posts 7% less often, the same pattern as our best time to post on X study, which covers hours and day-and-time blocks in detail. That study ends on October 26, so its figures differ by a point or so.
The formats behind the viral posts
Every day, Postel's format check reads the posts from the last two weeks by creators with at least 1,000 followers that have reached 5 times their follower count in views, and tags each one when it clearly follows one of Postel's post formats. Of the 8,309 posts checked so far, 473 (5.7%) did; most viral posts follow no named format.
- The launch21.6%
- The playbook15.9%
- The sharp take9.5%
- Lead with the number6.8%
- The numbers4.7%
- The receipts4.4%
- The role, not the company3.6%
- The milestone3.4%
- A clear position3.2%
- Tool to try3.0%
Show the numbers
| Format | Share | Posts |
|---|---|---|
| The launch | 21.6% | 102 |
| The playbook | 15.9% | 75 |
| The sharp take | 9.5% | 45 |
| Lead with the number | 6.8% | 32 |
| The numbers | 4.7% | 22 |
| The receipts | 4.4% | 21 |
| The role, not the company | 3.6% | 17 |
| The milestone | 3.4% | 16 |
| A clear position | 3.2% | 15 |
| Tool to try | 3.0% | 14 |
| The new chapter | 3.0% | 14 |
| Drop into the moment | 3.0% | 14 |
| Behind the scenes | 2.3% | 11 |
| The mistake and the fix | 2.3% | 11 |
| What people think vs what's true | 2.1% | 10 |
| The list promise | 1.7% | 8 |
| X vs Y | 1.7% | 8 |
| The turning point | 1.7% | 8 |
| The explainer | 1.5% | 7 |
| What I actually do | 1.1% | 5 |
| The question everyone asks | 0.6% | 3 |
| The roundup | 0.6% | 3 |
| Then vs now | 0.4% | 2 |
| Problem, named fix | 0.4% | 2 |
| Quiet wins | 0.4% | 2 |
| What I changed my mind on | 0.2% | 1 |
| The event recap | 0.2% | 1 |
| Flip the script | 0.2% | 1 |
| Go first | 0.2% | 1 |
| Ask the room | 0.2% | 1 |
| The reader's pain | 0.2% | 1 |
The most common formats have their own page with the most viewed real examples and a template: start with the launch, the playbook or the sharp take.
What to do with this
- Put the number in the first line. If the post has a figure, a result or a count, lead with it.
- Cut the opening question. Say the answer you were going to make people wait for.
- Write the longer version. If you have 500 characters worth saying, say them, on several short lines.
- Use video or two to four images, or none. Skip the single stock image and the GIF.
- Move links to a reply. Postel can add the link as a first comment when the post goes out.
- Test it against your own usual. Compare each post with your median views, not with someone else's account.
Write posts that follow what works, in your voice
Postel finds the posts going viral in your niche every week and drafts your own version of each for X and LinkedIn, in the way you write. You edit, approve and schedule.
8,000+ people have signed up
$1 for 7 days · Cancel anytime
Method
- Source. Postel's library of individual creators on X: the creators Postel tracks for its users, with brands, politicians, aggregators and fan accounts removed. It held 508,119 original posts from January 1 to October 31, 2026.
- Filters. English posts, first stored at least 24 hours after they went out (so views had a day to settle), from creators with at least 20 such posts. That left 297,373 posts from 4,439 creators; the median creator had 51,832 followers and the median post 8,709 views.
- Viral. A post went viral when it got at least 5 times its creator's median views in the sample. 30,887 posts (10.4%) did. (The weekly pages use a looser cut, 3 times the creator's usual and 10,000 views, so they have enough posts each week.)
- Compared with each creator's usual. For a group of posts (an opening, a length, a media type) we compare the viral posts it got with what those creators' own rates predict. 0% means as expected; +17% means 17% more viral posts than those creators usually get. A creator who always posts video only lifts the video number if their video posts beat their own other posts.
- Uncertainty. The 95% ranges come from resampling creators 300 times. A group whose range crosses 0% isn't clearly different from usual.
- Checks. We reran every group with "viral" at 3 times and 10 times the creator's median instead of 5, and for January to May and June to October separately. Number, list and announcement openings, 500 to 999 characters, 8 or more lines, video, two to four images, thread openers and Sunday stayed above usual in every version; question openings, posts under 100 characters, one-liners, a single image, GIFs, links and Wednesday stayed below. A single image was worse than text-only posts over the year and from January to May, and about even with them from June to October.
Limitations
- X only, English only. No data from LinkedIn or other networks; test before you carry it over.
- Not a random sample of X. The creators are tracked because their posts tend to do well, and the library leans towards tech, startups and AI.
- Correlation, not cause. A creator may save their best ideas for video, or write longer when they have more to say. The within-creator comparison removes who is posting, not what they chose to post.
- Rules, not reading. Openings are sorted by rules on the first line, so some posts land in the wrong group. Thread openers are found by a 🧵, "1/" or "thread" in the post, and the same rule also catches some numbered lists that start with "1.". Lines are counted in the first 600 characters, so very long posts may have more lines than counted.
- Views are a snapshot. Read when the post was stored or last refreshed.
Download the data
The aggregates behind every chart are free to use under CC BY 4.0, with a link to this page: viral-posts-study-2026.csv. Each row is one group (an opening, a length, a number of lines, a media type, links, threads, a weekday, or a format) with its post count, creators, viral posts, the viral posts expected from those creators, the difference in percent with its 95% range, and median views against the creators' usual. No post text or creator names are included.
To cite this study
Sünneke, J. (2026, November 4). What made X posts go viral in 2026 [Data set and analysis]. Postel. 297,373 posts from 4,439 creators. https://www.postel.app/viral-posts-study
Frequently asked questions
What makes a post go viral on X?
In 297,373 posts from 4,439 creators in 2026, the posts that went viral more often than their creators' usual led with a number (+17%), carried video (+19%) or two to four images (+16%), were 500 to 999 characters long (+20%) and had no link. Thread openers did best of all (+61%), from few posts.
Do questions work as hooks on X?
Not in this data. Posts whose opening line asked a question went viral 8% less often than their creators' usual (95% range −12% to −5%).
Does post length matter on X?
Yes. Posts under 100 characters went viral 9% less often than usual and posts of 500 to 999 characters 20% more often, each compared with the same creator's usual rate.
Do images help a post go viral on X?
Several do; one doesn't. Two to four images +16%, one image −13%, video +19%, a GIF −33%.
Can I use this data?
Yes, under CC BY 4.0 with a link to this page. The CSV holds the aggregates; there is no post text or creator data.