X's median engagement rate hit in 2026, a 9.0% year-over-year drop and the steepest decline of any major platform in the benchmark (Sociavault, 2026). Meanwhile the average user is spending more time on X than they were two years ago. This Twitter/X engagement data study analyzes posts published between January 2024 and May 2026 to answer one question: is your engagement bad, or has the platform gotten harder?
There's a step before that question, and almost nobody takes it. Before you attribute a decline to your content or to the platform, you have to know which formula produced the number you're staring at. The same set of posts can yield engagement rates that differ by 4.7x depending on whether clicks are counted and whether you take the median or the mean. A lot of the "our X engagement dropped 9%" slides going into QBRs right now are comparing incompatible math.
TL;DR:
- Formula choice moves engagement rate more than content quality does. On Ordinal's dataset, the same posts return,, or depending on the definition, a spread.
- X's median engagement rate fell from 1.34% in 2024 to 1.22% in 2025 to in 2026 (Sociavault). Ordinal's median across posts sits closer to because the account mix and denominator differ.
- Threads average a engagement rate in Ordinal's data, roughly plain text and video.
- Average impressions per post fell 5% year over year while average retweets per post rose 35%.
- Text-only posts average 35,975 impressions versus 18,334 for media posts, but media converts harder at 0.79% versus 0.68%.
What Counts as Engagement on X (And Why Your Number Might Be Wrong)
Engagement rate on X is (likes + replies + reposts + quote posts) divided by impressions, multiplied by 100. That's the definition Ordinal uses across all posts in this study. Every part of that formula is a choice, and each choice moves the output enough to change what you conclude about your own performance.
The Same Posts, Three Different Numbers
When Ordinal ran the identical dataset through three common definitions, the results were (median, engagement actions only), (median, including clicks), and (mean, including clicks). Same posts, same window, same accounts, and a 4.7x spread from formula choice alone.
Clicks are the first fork. Profile clicks and link clicks get counted as "engagements" by most native dashboards, and they're far more common than replies or quotes, so folding them in inflates the number by more than 50% in our data. Neither definition is wrong. A benchmark built on one and a report built on the other aren't comparable.
The second fork does more damage.
Mean engagement rate on social data is close to useless because the distribution isn't normal. A handful of posts pull 40x median impressions and drag the average up, which is how you end up with an "benchmark" that almost no individual post reaches. Sort your posts by engagement rate and take the middle value instead, because the median describes what a typical post from your account does.
How to Tell Which Formula Your Tool Is Using
Take one post you know well and calculate it by hand: likes plus replies plus reposts plus quotes, divided by impressions. Then compare that to what your dashboard shows for the same post. If the tool's number is meaningfully higher, it's counting clicks. If your account-level number is much higher than the median of your hand-calculated posts, it's reporting a mean. Native X analytics does both by default, which is why the dashboard number almost never matches a published benchmark.
Fix this before you present anything. If you switched tools last year and saw engagement "drop 40%," you probably changed denominators. Our guide on how to calculate engagement rate walks through the arithmetic.
X's Median Engagement Rate Is and Falling
The platform median for 2026 is 1.34% in 2024, 1.22% in 2025, and in 2026 (Sociavault, 2026). Two consecutive years of decline is a pattern rather than a bad quarter.
Ordinal's median across posts is closer to, and we're not going to paper over the gap. Sociavault's benchmark draws on a different account mix, and our sample skews toward B2B and startup accounts posting at higher volume, which pulls the median down. It also reflects the strict denominator described above. Both numbers are defensible. Mixing them is what isn't.
For self-assessment: above 1% is strong for a B2B account. Between 0.5% and 1% is normal. Under 0.5% points at format or timing rather than the platform, because plenty of accounts in the same sample clear 1% in the same weeks.
Time Spent Is Up and Engagement Is Down
These two facts don't contradict each other. X has roughly 388 million monthly active users and a potential advertising audience of 557.5 million (Onclusive, 2026), and average daily time spent rose to 32 minutes in 2026 from 24 minutes in mid-2024 (Blankspaces, 2026).
The audience is there and paying more attention than it was.
What changed is the denominator. Longer sessions mean more posts consumed per session, and every additional post in the feed is another post yours competes with. Engagement rate is a ratio, and impressions are growing faster than interactions which indicates that the platform is compressing.
Ordinal's data shows the mechanic cleanly. Average impressions per post fell 5% year over year while average retweets per post rose 35%. Reach costs more to earn now, but the people who do see your post are more willing to pass it along. If you want the distinction between those two numbers spelled out, we covered reach versus impressions separately.
None of this argues for pulling budget out of X. The read from the 2026 data is that X's ranking model rewards posts that earn amplification in the first hour, so concentrated, high-signal posting beats volume.
Engagement by Format: Threads Beat Everything Else
Threads average a engagement rate in Ordinal's analysis of X posts, roughly plain text and video. Video, the format most teams default to because it feels like the safe modern choice, is the weakest performer in the set at around 0.54%.
Format performance across the dataset:
- Threads: engagement rate, the highest of any format and the only one clearing the platform median by a wide margin
- Image posts: roughly 0.79% engagement rate on an average of 18,334 impressions
- Text-only posts: roughly 0.68% engagement rate on an average of 35,975 impressions
- Video: roughly 0.54% engagement rate, the lowest in the sample
The impressions split is the finding people argue with. Text-only posts reach nearly twice as many people as media posts, but media posts convert a higher share of the people they reach. If the goal is reach, post text. If the goal is interaction per view, add an image.
Threads sidestep the tradeoff because they generate dwell time, and dwell is what the ranking model pays for.
Post length runs the same direction. Posts of 100 to 200 characters produce the highest engagement rate in our data, despite X's expanded character limits.
"We found that when we bold our hooks, we see like a pretty big spike in engagement rate on those posts on LinkedIn. Little tactics like that, it's interesting to see the validation for them." Jeffrey Zhao
Our Twitter analytics breakdown goes deeper on the format-level numbers.
How to Benchmark Your Own X Engagement
- Export your last 90 days of posts from native X analytics. Use the CSV export rather than the dashboard summary, because the summary is pre-aggregated with a formula you can't inspect.
- Calculate engagement rate per post with one consistent formula: likes plus replies plus reposts plus quotes, divided by impressions, times 100. Strip out profile clicks and link clicks.
- Take the median. Sort the column and pick the middle value, because the mean will be dragged up by your two best posts and tell you nothing about a typical week.
- Segment by format before you compare anything. A thread-heavy account and a video-heavy account will produce different medians for reasons that have nothing to do with content quality.
- Compare each format segment against the format-specific numbers above and your blended median against the platform figure. Comparing a video-only account to a blended industry benchmark is how teams end up killing a strategy that was working.
Timing is the other variable worth isolating before you blame format, and we published the current best posting times from the same dataset.
What to Do With This Data
Shift your format mix toward threads first. If threads run plain text and threads are 10% of your calendar, that ratio is the biggest gain available to you this quarter, and it costs nothing but planning time.
Then stop reporting impressions as a headline metric. Impressions per post fell 5% platform-wide while retweets rose 35%, so a flat impressions chart in a QBR mostly describes a platform trend. Report amplification and engagement rate by format instead, and put the formula on the slide so nobody relitigates the math.
Post shorter, in the 100 to 200 character range, and get real engagement onto posts inside the first hour, because early interaction compounds in a recency-weighted feed. For cross-channel context on the same argument, we published broader cross-platform engagement data alongside this study.
Teams running this properly need format labels on every post and analytics that segment by those labels, which is the part spreadsheets break on. Ordinal handles format-level performance tracking and coordinates first-hour engagement across teammate accounts so the comparison stays clean month to month.
The One Audit Worth Running This Quarter
Pull your last 90 days, recalculate every post with one formula you can defend, take the median, and find your thread-to-text ratio. Then move it. If your reported engagement rate changes by more than 30% once you strip clicks out and switch from mean to median, the decline you've been explaining to leadership was arithmetic, and you now have a much better answer for the next QBR than "the platform is declining."
Frequently Asked Questions
What Is a Good Engagement Rate on X (Twitter) in 2026?
Anything above 1% counts as strong. Sociavault's 2026 benchmark puts the platform median at, while Ordinal's analysis of posts finds a median closer to because of differences in account mix. Under 0.5% usually points to format or timing rather than the platform.
Why Is My Twitter/X Engagement Dropping?
Platform-wide engagement fell 9.0% year over year in 2026, the steepest decline of any major platform tracked by Sociavault. Ordinal's dataset shows average impressions per post fell while retweets per post rose 3, so reach is genuinely harder to earn right now.
What Content Format Gets the Most Engagement on X?
Threads. Ordinal's analysis of X posts found threads average a engagement rate, nearly plain text and video. They win because they hold attention longer, and dwell time is what the ranking model rewards.
Do Text-Only Posts Get More Reach Than Posts With Images on X?
Yes, on raw reach. Text-only posts average impressions compared to for media posts in Ordinal's dataset. Media posts convert better per view, 0.79% versus 0.68% engagement rate, so the right call depends on your goal: reach or interaction.
Should I Use Median or Mean to Measure Engagement Rate?
Median, every time. Ordinal found the same dataset produces a median, a median when clicks are included, or an mean, a 4.7x spread from formula choice alone. Mean gets skewed by a handful of viral outliers and makes ordinary performance look broken.
Is X Still Worth Using for B2B Marketing in 2026?
Yes. X has roughly 388 million monthly active users, according to Onclusive's 2026 report, and average daily time spent climbed to 32 minutes from 24 minutes in mid-2024, per Blankspaces. What's changed is that any single post now competes for a smaller slice of a busier feed.
How Long Should My Posts Be on X for Maximum Engagement?
Ordinal's data shows posts between 100 and 200 characters get the highest engagement rate, despite X's expanded character limits allowing much longer posts. More room to write doesn't translate to more interaction per view.
How Do I Calculate Engagement Rate on X?
Add likes, replies, reposts, and quote posts, divide by impressions, then multiply by 100. Leave out profile clicks and link clicks unless you're matching a benchmark that includes them, since folding clicks in can inflate your number by more than 50%.
Related Posts
Looking to learn more? The below posts may be helpful for you to learn more about social media and content marketing.
