How to tell if your X followers are bots in 2026

Start with the profile: a default photo, an empty bio, a following count far above the follower count, a very new account and a strange post count are the classic warning signs. Then open the Posts and Replies tabs and read about ten recent posts, because in 2026 plenty of accounts with a photo, a bio and years of history post the same promotion or the same interchangeable reply again and again. No single sign proves an account is automated, so judge what the account does, not who you guess is behind it.

Step 1: run the classic profile checklist

These are the checks most follower tools still run. They take seconds per account and catch the laziest accounts.

• Default profile photo and empty bio. X treats these as risk signals too: its advanced notification filters let you mute accounts with a default profile photo, new accounts, and accounts without a confirmed email or phone number.⁠^[1] Treat that as a hint about where X expects noise, not as proof.
• Following far more than following back. An account following thousands with a few dozen followers is often working the follow-back game. X caps everyone at 400 follows a day and at 5,000 accounts followed until their own follower count grows, and its rules ban "follow churn", following and unfollowing in bulk to inflate a count.⁠^[2] A following count parked near 5,000 next to a tiny follower count is a common shape for that behaviour.
• A handle that looks generated. A first name followed by a long run of digits shows up on throwaway accounts, and also on people who never chose a handle. Weak on its own.
• Account age and location. Tap the join date on a profile to open "About this account", which shows the country or region X associates with it. The Associated Press noted that VPNs can mask that location and that X shows a notice when the data may be off.⁠^[3] A brand-new account is worth a closer look; a location mismatch is a prompt to read more, not a verdict.
• Post count. Zero posts, or tens of thousands of posts from an account a few months old, both deserve a look at the actual posts.
• An "Automated" label. X's automation rules require automated accounts to carry an "Automated" label linked to the human-run account that manages them.⁠^[4] A labelled bot is being honest about what it is; the accounts worth worrying about don't carry the label.

Step 2: know what the checklist misses

Every item above is something an operator can fill in. X's own authenticity policy describes manufactured identities built with stock, stolen or AI-generated profile photos and copied bios.⁠^[5] A blue checkmark doesn't close the gap either: it needs a display name, a profile photo and a confirmed phone number, and X says it means an active Premium subscription, not an ID check.⁠^[6]

So the newer problem looks different. On 24 July 2026 Nikita Bier, then X's head of product, said X had removed 42,000 accounts that were automating replies with chatbots.⁠^[7] They were described by what they did, reply through a chatbot, not by what their profiles looked like. (The purge and the reply side of this are covered in AI reply bots on X.)

Our own reports show why reading posts adds something. Across TrustAudience's 15 current-method public reports, 82% of 1,295 newest-follower profiles had a bio and a custom profile image, and 84% were at least 180 days old. In the same reports, 161 of the 984 followers whose posts we could read (16%) were labelled Repetitive promotion or Repeated reply pattern.⁠^[8] These are two separate counts over the same samples, not a match of individual accounts, but they make the point: a profile filter would wave most of these samples through.

Step 3: read ten recent posts and replies

Open the profile, then the Posts tab and the Replies tab. Ten posts is enough to see a pattern without spending all afternoon. Look for:

1. The same promotion, repeated. The same link, coin, course, giveaway or "DM me" offer in post after post, with the wording barely changed.
2. Replies that would fit under anything. "Great insight!", "This is so true 🔥", or a polished sentence that restates the original post and adds nothing specific. One is normal. Two or more that are interchangeable is a pattern.
3. A Replies tab full and a Posts tab empty. Accounts that exist only to sit under big accounts' posts often have almost nothing of their own.
4. No thread of interest. Real people repeat themselves too, but around their job, their team, their hobby. A timeline that jumps from crypto to fitness to AI tips with the same tone throughout is worth noting.
5. Follow-for-follow language. "Follow back", "F4F", "drop your handle" in the bio or posts.
6. Bursts. Dozens of replies to unrelated large accounts within a few minutes.

What not to use: "this sounds like AI". Researchers have shown that AI-text detectors can be pushed well off course by paraphrasing the text,⁠^[9] and careful human writers get mistaken for machines. Writing that reads as polished, translated or templated is a weak signal. Repetition across posts is a much stronger one, whoever or whatever wrote them.

Step 4: sort by behaviour, not by identity

When you've read a follower's posts, put them in one of a few plain buckets rather than "bot" or "human":

• Substantive: posts with real discussion or information.
• Mixed: ordinary, personal or networking posts, with some promotion.
• Repetitive: the same promotion, or interchangeable replies, again and again.
• Silent: no posts of their own, too little to judge, or a private account. Quiet is not suspicious by itself; most people read more than they post. Silent followers on X goes into how common that is.
• Gone: the profile says the account is suspended or no longer exists (a TrustAudience report calls suspended or deactivated followers Unavailable). Those drop out of your count over time, which is its own topic: see why your X follower count dropped.

Then act on the repetitive ones if they bother you. On the web you can remove a follower from your followers list or their profile without blocking them, though they can follow again later.⁠^[10] Block, mute or report spam when it's more than a nuisance. Don't post screenshots calling someone a bot; you're judging a handful of posts, and you could be wrong.

How many followers to check

You don't need to read your whole list. The newest followers are where a recent burst shows up, and they're the easiest to reach: open your followers list and work down from the top. Thirty accounts gives you a rough picture; fifty is better. If a big share of the first thirty are repetitive, that tells you something happened recently, such as a giveaway, a viral post that drew reply farms, or a paid promotion. If you're trying to judge an account that sold you something, what bought X followers look like covers the specific patterns.

Check it by hand

A 30-minute routine, no tools needed:

1. Open your followers list on the web and note the first 30 accounts.
2. For each, spend 20 seconds on the profile checklist in step 1 and write down any flags.
3. Then spend a minute on the Posts and Replies tabs. Mark the account Substantive, Mixed, Repetitive, Silent or Gone.
4. Count each bucket. Compare the Repetitive count with how many accounts were flagged by the profile checklist alone; the two lists usually differ.
5. Repeat in a month, from the top of the list again, to see whether new followers look different.

TrustAudience does the same reading at a larger scale: it reads up to 10 recent public posts or replies from each of up to 100 of an account's newest followers, labels each one, and shows the counts behind the Audience score (method). It is a recent, non-random sample. It can't establish whether an account belongs to a person, whether a post was written by AI, or the quality of your whole following, and the score does not verify audience quality.