X (Twitter) Fake Follower Percentage Calculator

The calculator needs two counts from an audit you carry out yourself: how many followers you checked and how many you flagged. It does not look at any account.

1. Write down your criteria before you start, so every account is judged by the same rule.
2. Pick followers at random. One way is to take every 60th follower in the list when you have 12,000 followers and want 200.
3. Open each picked profile, apply your criteria, and tally flagged or not flagged.
4. Type the sample size, the flagged count and your total followers, then choose a confidence level.

A fake follower percentage is the share of an account's followers that are not genuine people or organizations. Nobody outside X can measure it directly. X does not publish the figure for any account, and no outside tool can see the signals X uses.

What you can do is audit a sample by hand and estimate. This page handles the arithmetic of that estimate and states how uncertain it is.

The X Authenticity policy, dated April 2025 and read on October 3, 2026, describes what X treats as inauthentic. It says accounts must be legitimate, genuine and transparent as to their source, identity and popularity. Among the things it names:

• Fake personas: manufactured identities, which may use stock, stolen or AI-generated profile photos, copied bios or misleading profile information to deceive.
• Unauthorized automation: automated or scripted accounts that do not comply with the X Developer Policy.
• Metric inflation: coordinating with or compensating others to inflate follows, likes, reposts or views.

A flag in your audit is your own judgement against your own criteria. It is not a finding by X, and it does not prove that an account breaks a rule. A quiet account with no photo can belong to a real person who only reads.

1. The tool divides the flagged count by the sample size n to get p.
2. It picks z for the confidence level: 1.645 for 90%, 1.960 for 95% and 2.576 for 99%.
3. It computes the lower and upper Wilson limits with the formula above.
4. It subtracts the lower limit from the upper limit for the interval width.
5. It multiplies p and both limits by your total followers.

The Wilson method is described in the NIST e-Handbook of Statistical Methods. NIST notes that its lower limit cannot be negative and that it works for almost all sample sizes and shares. The plain formula p +/- z x sqrt(p(1 - p) / n) fails that test: for 1 flagged account in 50 it gives a lower limit of -1.9%, while Wilson gives 0.4% to 10.5%.

Example values only. An account has 12,000 followers. The owner audits every 60th follower, 200 accounts in all, and flags 24 of them. The confidence level is 95%.

• Flagged share = 24 / 200 x 100 = 12.0%.
• Wilson interval at 95% = 8.2% to 17.2%.
• Interval width = 17.2 - 8.2 = 9.0 points.
• Not flagged = 200 - 24 = 176, or 88.0% of the sample.
• Estimated flagged followers = 12.0% x 12,000 = 1,440.
• Range in followers = 984 to 2,068.
• Sample covers = 200 / 12,000 x 100 = 1.7% of followers.

The honest statement is a range: between about 984 and 2,068 of the 12,000 followers would be flagged by this owner's criteria. At 90% confidence the interval narrows to 8.7% to 16.3%. At 99% it widens to 7.3% to 19.2%.

Sample Size

A larger sample narrows the interval. The table shows a 10% flagged share at five sample sizes, at 95% confidence.

Going from 50 to 200 accounts cuts the width by about half. Plan a sample with the sample size calculator.

How You Pick the Sample

The interval is valid only for a random sample. The first 200 accounts in a follower list are one slice of the list, not a random draw. Taking every 60th account, or using random positions, spreads the sample across the whole list.

Your Criteria

Strict criteria flag few accounts and loose criteria flag many. The percentage measures your rule as much as your followers. Keep the rule fixed between audits.

Confidence Level

A higher confidence level gives a wider interval from the same data.

Followers You Cannot Open

Protected, suspended or deleted accounts show little. Decide in advance whether they count as flagged, not flagged or skipped.

A sample audit reports a share you observed with a stated margin, and an automated checker reports a figure from rules it does not show. Neither can see inside X.

Older versions of this page guessed a fake percent from an engagement rate and a table of "healthy" rates. X publishes no such table, so that method was removed. The margin behind any survey-style estimate is explained on the margin of error calculator.

Vetting a Partner Account

Before a paid collaboration, audit a sample of the partner's followers. Report the interval, not only the single percent.

After a Sudden Jump in Followers

Audit the newest followers as their own group. Track the jump itself with the X follower growth calculator.

A Yearly Health Check

Repeat the audit with the same criteria and sample size. Overlapping intervals mean the change may be chance.

Reporting Suspect Accounts

X Help says you can report inauthentic accounts with the in-app reporting flow. An audit gives you a list to review, and X decides what breaks its rules.

1. Trusting an Instant Fake Follower Percent

A tool that asks only for a follower count and engagement numbers cannot see a single follower. Its percent comes from invented benchmarks.

2. Auditing the Top of the List

A slice of the list is not a random sample, and the interval does not apply to it.

3. Quoting the Percent Without the Interval

5 flagged in 50 is 10.0%, with a 95% interval of 4.3% to 21.4%. The single figure hides a spread of 17.0 points.

4. Treating Inactive as Fake

An account that never posts can be a real reader. Automation that follows the X Developer Policy is also allowed, and X Help describes an automated account label, currently in testing, for accounts that post automated content.

5. Buying Followers to Fix a Ratio

X Help prohibits using or promoting third-party services or apps that claim to add followers. It also prohibits follow churn and duplicating another account's followers.

6. Changing Criteria Between Audits

Two audits with different rules cannot be compared. Simple shares for any count are on the percentage calculator.

The arithmetic is exact for the counts you enter. The estimate is only as good as the sample and the criteria behind it.

The interval assumes a sample that is small next to the follower count. When you audit a large part of your followers, the true uncertainty is smaller than shown, so the interval errs on the wide side.

No. A calculator sees only the numbers you type. It cannot open an account, so it cannot tell a real follower from a fake one. This page estimates from a sample you audit yourself.

Audit a random sample of followers by hand, count the accounts you flag, divide by the sample size and multiply by 100. Add a confidence interval, because a sample never gives the exact share.

More accounts give a narrower interval. At a 10% flagged share and 95% confidence, 100 accounts give 5.5% to 17.4% and 400 accounts give 7.4% to 13.3%.

A Wilson interval is a range for a true proportion, built from a sample. NIST describes it as a method whose lower limit cannot be negative and that suits almost all sample sizes.

No official figure exists. X does not publish a typical fake follower share, and outside estimates use their own definitions. Compare your audits with each other, using the same criteria.

Set your own written criteria. The X Authenticity policy mentions stock, stolen or AI-generated profile photos, copied bios and misleading profile information as marks of a fake persona. One sign alone proves nothing.

Yes. The X Authenticity policy prohibits compensating others to inflate follows, and X Help prohibits using third-party services or apps that claim to add followers.

The Authenticity policy lists anti-spam challenges, restricted reach, temporary loss of features and suspension. Severe violations can lead to permanent suspension at first detection.

The share is 0%, and the upper limit is still above zero. With none flagged in 400 accounts, the 95% Wilson interval runs from 0.0% to 1.0%.

No. The maths runs in your browser and nothing is sent to us. Save keeps a result only in this browser.

• NIST/SEMATECH e-Handbook of Statistical Methods, section 7.2.4.1: Confidence intervals (National Institute of Standards and Technology, read 2026-10-03; the Wilson method for a confidence interval on a proportion, with a worked example).
• Authenticity policy (X Help Center, read 2026-10-03; policy dated April 2025: inauthentic accounts, engagement spam, enforcement and reporting).
• About following on X (X Help Center, read 2026-10-03; follow churn, duplicating followers and third-party services that claim to add followers).
• About profile labels and checkmarks on X (X Help Center, read 2026-10-03; automated account labels and what checkmarks mean).