Bot Hound checks whether a public 𝕏 (Twitter) account is a bot. Here's what people ask most. For a walkthrough of the whole thing, see how it works.
The most reliable tell is the post history. Bots struggle to produce a stretch of posts that read like a person - original opinions, real back-and-forth with other accounts, personal details, jokes. Profile metadata like follower ratios and account age helps, but on its own it's weak: plenty of real people have odd-looking numbers, and good fakes are built to look normal.
That's the approach Bot Hound takes. It reads hundreds of an account's recent posts along with its bio and avatar, then returns a bot-likelihood score with the reasons behind it.
Yes. You sign in with 𝕏 so we know who you are and can track your free check and your balance. The sign-in is identity-only - Bot Hound cannot post, send DMs, follow accounts, read your direct messages, or see your followers. The permissions page lists the exact scopes.
No. Bot Hound only reads public data, the same way any person opening the profile would. Nothing is posted, liked, or followed, and the account gets no notification. If you share the report link publicly, that's obviously visible to anyone who sees it.
Usually well under a minute. Bot Hound fetches the profile and recent posts before classifying, so the exact time depends on how much the account has posted.
Verdicts are cached per handle for seven days, so a repeat check of the same account inside that window returns the existing verdict quickly instead of re-analyzing from scratch.
Yes. The analysis is a language model reading the posts rather than a set of English keyword rules, so it works across languages.
It's an AI estimate, not a determination, and every report says so on its face. It's good at catching obvious spam, scam, and lure bots, and reasonable on subtler fakes. It's weakest on accounts with very few posts, because there's little for it to read.
Treat the score as an informed opinion and read the reasons underneath it - that's where the actual argument is.
Usually because there wasn't much to go on: a new account, almost no original posts, nothing but retweets, or a default profile picture. The model is told to assume an account is human unless there's real evidence otherwise, but a thin account gives it very little evidence in either direction.
If you think a verdict about your account is wrong, tag or DM @BotHound_ on 𝕏 and we'll take another look.
No, and Bot Hound specifically does not treat it as a signal. 𝕏 appends numbers to handles when your first choice is taken, so enormous numbers of real people have them. Randomly generated handles only matter alongside other signals.
It's the model's confidence that the account is a bot, from 0 to 100%, sorted into four bands: Looks human under 15%, Some bot signals from 15-45%, Likely a bot from 45-75%, and Very likely a bot above 75%. It isn't a measure of how bot-like the account is - just how sure the model is. There's more detail on how it works.
Your first check is free. After that it's $1 per check with a funded balance, or $1.50 for a single one-off check. No subscription.
Bot Hound's first check is free, with no card required - enough to settle a single account you're suspicious about.
Most other free bot checkers are either heuristic scorers that only look at follower ratios and account age, or archives that no longer score new accounts. X API pricing made free live analysis expensive to run, which is why so few of them still work.
No. A protected account has no public posts to read, so there's nothing to judge. Bot Hound refuses the check up front and you aren't charged.
Bot Sentinel was suspended from X in 2023, and Botometer - the academic bot detection tool - stopped scoring new accounts that same year after X's API pricing changed. That's why people go looking for replacements.
Bot Sentinel's work centers on coordinated disinformation networks rather than answering whether one specific account is a bot. Bot Hound does the narrow version of that job: you give it a single handle, it reads hundreds of that account's posts plus the bio and avatar, and returns a bot-likelihood score with reasons. It does no network or coordination analysis.
Most bot detection tools score accounts with fixed heuristics - follower ratio, account age, posting frequency. Those rules misfire on real people with unusual numbers, and sophisticated fakes are built specifically to pass them.
Bot Hound sends the account's actual post content to a language model, so the judgment rests on whether the posts read like a person rather than on metadata thresholds. The tradeoff is honest: it costs money per check, and it gives you an opinion with reasons rather than a hard number.
No. Bot Hound checks one account at a time - the handle you type in. It doesn't scan your follower list or produce an audience audit.
No. Bot Hound gives you a verdict; blocking is still something you do yourself on 𝕏. The older auto-blocking product and its browser extension have been retired.
Yes. Every report has its own URL with a link preview card, so you can post it on 𝕏 or send it to a client or teammate. Here's a sample report.
Only public 𝕏 data - the public profile, public posts, and avatar of the account being checked. No passwords, no DMs, no private data. Your own sign-in is identity-only. See the privacy policy for the full picture.
Yes. Tag or DM @BotHound_ on 𝕏, or email [email protected]. Contacting us from the account itself lets us confirm you control it.
We review requests in good faith and can take the report offline, re-run it, or remove it permanently. A disputed report shows only an “under review” notice - no score, no band, no reasons.
xAI's Grok. It reads the account's posts, bio, and profile data, and also looks at the avatar image itself to spot AI-generated faces, stock photos, and headshots stolen from real people.
No. Bot Hound is an independent tool, not affiliated with, endorsed by, or operated by 𝕏 Corp.
First check is free.