Twitter (X) Shadowban Statistics 2026: 38.3% Restricted in September
In September 2026, 38.3% of 8,501 Twitter (X) accounts checked had at least 1 of 4 restrictions active, down from 54.3% in August. Weekly and per-type data.
In September 2026, 38.3% of the 8,501 Twitter (X) accounts checked on Shadowban Radar had at least one of the 4 restrictions active on their most recent check of the month. In August 2026, the figure was 54.3%. The rate fell through the first three weeks of September and then held between 32% and 36%.
What share of X accounts were shadowbanned in September 2026?
38.3% of the 8,501 accounts checked in September 2026 were restricted (95% confidence interval: 37.3% to 39.3%), against 54.3% of 1,564 accounts in August 2026. Each account is counted once by its most recent check in the month:
| Month | Accounts checked | Restricted | Not restricted | 95% confidence interval (restricted) |
|---|---|---|---|---|
| August 2026 (August 8 to 31) | 1,564 | 54.3% | 44.8% | 51.8% to 56.7% |
| September 2026 | 8,501 | 38.3% | 61.6% | 37.3% to 39.3% |
The two columns do not add up to 100% because a few accounts gave no clear result and are counted in neither. The September figure is based on 5.4 times more accounts than August's, so its interval is much tighter.
What counts as a shadowban in this data?
Free, no login, results in seconds.
In this report, an account counts as restricted when at least one of 4 checks came back positive on its most recent check of the period. The 4 checks, in this order:
- Search suggestion ban. The account does not appear in the dropdown suggestions when someone types its handle or name in the search box.
- Search ban. The account's posts are missing from X search results.
- Ghost ban. The account's replies are hidden from the conversation they belong to.
- Reply deboosting. The account's replies are ranked below other replies in the conversation.
A shadowban on Twitter (X) is a hidden limit on account visibility: posts, replies or profiles may reach fewer people without a notice. Terms used in this report are defined in the X shadowban glossary.
How did the shadowban rate on X change week by week in September 2026?
The share of restricted X accounts fell from 53.9% of 1,836 accounts in the week of September 1 to 7 to 32.3% of 2,661 accounts in the week of September 15 to 21, a drop of more than 20 points.
| Period | Accounts checked | Restricted | 95% confidence interval |
|---|---|---|---|
| September 1 to 7 | 1,836 | 53.9% | 51.6% to 56.2% |
| September 8 to 14 | 2,467 | 45.5% | 43.6% to 47.5% |
| September 15 to 21 | 2,661 | 32.3% | 30.5% to 34.1% |
| September 22 to 28 | 2,494 | 35.3% | 33.4% to 37.2% |
| September 29 to 30 (2 days) | 1,054 | 34.2% | 31.4% to 37.1% |
Each row counts every account once, by its last check in that period. An account checked in two different weeks appears in both rows.
The first week of September (53.9%) is close to the August figure (54.3%). The last three periods sit in a narrow band, between 32.3% and 35.3%, and their intervals overlap.
Which type of restriction is the most common on X?
Search ban was the most common restriction among restricted X accounts in September 2026, at 53.4%, ahead of search suggestion ban at 46.1%.
| Restriction | Share of restricted accounts | 95% confidence interval |
|---|---|---|
| Search suggestion ban | 46.1% | 43.7% to 48.5% |
| Search ban | 53.4% | 51.0% to 55.7% |
| Ghost ban | 25.6% | 23.5% to 27.7% |
| Reply deboosting | 26.3% | 24.2% to 28.4% |
This breakdown covers the 1,690 restricted accounts among the 4,814 accounts checked from September 11 to 25.
Do X accounts often carry more than one restriction?
Yes: 29.5% of restricted X accounts had two or more restrictions active together in September 2026 (499 of 1,690), and 70.5% had exactly one. The four shares add up to more than 100% for this reason. Each share counts an account only when that check came back positive.
Did the shadowban rate on X go up or down between August and September 2026?
It went down: the restriction rate was 16 points lower in September 2026 (38.3%, 8,501 accounts) than in August 2026 (54.3%, 1,564 accounts), and the two confidence intervals do not overlap.
Here is what the data does and does not say:
- A real gap. The August interval (51.8% to 56.7%) and the September interval (37.3% to 39.3%) are far apart.
- A within-month decline. The drop happened inside September, between the first and third week, not between two months.
- No cause tested. We did not measure why. Who runs a check changes over time, and a different mix of visitors can move the rate without any change in how X treats accounts. Treat any explanation, including ours, as an untested hypothesis.
August covers August 8 to 31 only, so it holds fewer accounts and no per-type breakdown.
How was the September 2026 shadowban rate calculated?
Each rate is the share of accounts whose most recent check in the period found at least one restriction. The rules behind it:
- Source. Every free check run on Shadowban Radar between August 8 and September 30, 2026. The live version of these figures is on X Shadowban Statistics, which is updated every day.
- One account, counted once. An account is counted once per period, by its last check in that period.
- Excluded. Internal checks run by Shadowban Radar on its own accounts, and accounts that X does not serve (suspended or not found), are left out.
- Complete days. Days are UTC days. The period ends on September 30, so no partial day is counted.
- Intervals. Confidence intervals are 95% Wilson score intervals, which stay between 0% and 100% at these sample sizes.
- Privacy. No handle and no individual account is published, only totals and percentages.
- Other reports. The total differs from the country study, which covers August 1 to September 28 and applies its own exclusion rules.
The daily and 7-day figures on the live page can differ slightly from the monthly figures here, because they use a rolling window and not the calendar month.
Can these figures stand for all X accounts?
No. The figures describe people who ran a free check on Shadowban Radar, not all X accounts. Three limits apply to every number in this report:
- A self-selected sample. Everyone behind these numbers chose to run a check, usually because they suspected a drop in reach. This is not a random sample of X accounts, and the real rate across X is almost certainly lower.
- A snapshot per period, not a count of events. The data says how many accounts had a restriction when they were last checked, not how many restrictions started or lifted.
- Not a measure of X's policies. A higher or lower rate says something about the people who checked in that period, and nothing about why X applied a restriction.
For a comparison across visitor countries, see which countries get shadowbanned the most on X. For how long a restriction lasts, see how long a shadowban lasts, and for the 4 restrictions in one guide, is my Twitter (X) account restricted.
Want to know where your own account stands on the 4 checks? Run a free shadowban check.
