How listings work
XBotRadar is a registry of dataset membership, not a bot detector. Here is exactly where the data comes from, what a flag means, and how to get a wrong listing removed.
Where the data comes from
Human scouts in our community watch X giveaways for coordinated bot rings — clusters of accounts with matching creation windows, templated bios, and scripted entry behaviour. When a scout maps a ring, they compile it into a dataset ("farm") with a description of the observed pattern and the giveaways where it was seen.
Named admins review and upload each dataset through a gated import flow. Every entry permanently records which dataset it came from, who uploaded it, and when. Nothing enters the database anonymously, and public reports never go straight in — they are leads for scouts, not listings.
What a flag means (and what it doesn't)
A flag means: "this username appeared in one or more community-submitted datasets on the dates shown." That's it. It is not a factual verdict that the account is a bot, and it should never be the only reason to exclude someone from anything.
X usernames can be renamed and re-registered by different people over time. A handle flagged last year can belong to someone new today — one reason every listing shows its dates and why disputes exist.
Confidence levels
The username appears in multiple independently-submitted datasets, or the dataset pattern was corroborated by more than one scout.
A single vetted dataset from a named scout with a documented pattern.
Imported from an external or community source that we could not fully verify. Treat with extra caution.
Disputes & removal — always free
Anyone can dispute a listing — no account needed. While a dispute is open the listing shows as "under review" and should be treated as unlisted. A human reviews every dispute; wrong listings are cleared, and cleared listings stay publicly visible as cleared, because we don't hide our corrections.
We never charge for removal. Anyone asking you to pay to be delisted from XBotRadar is a scammer.
The AI advisory is not a listing
When a lookup returns no record, you can optionally ask an AI model for a second opinion. It runs only in that case — never alongside a flag — and it reads only the profile details you paste in yourself. We do not fetch anything from X, and we do not verify what you paste.
That read is generated fresh for each request. It never becomes an entry, it is never reviewed and published as a listing, and it is not counted in any figure on this site. The bands mean exactly what the panel says they mean — that some number of details in your snapshot line up — and nothing more. Ordinary real accounts routinely produce unusual signals, and practised bot farms routinely produce ordinary ones.
If a read looks wrong, the feedback link on the panel goes to the same human queue as disputes.
Mistakes we can make
Curated data is human data: handles get recycled, scouts can misjudge a cluster, an imported list can contain collateral accounts. That's why every listing carries provenance, why batch imports can be reverted wholesale, and why the audit trail records every change. If we got it wrong, dispute it — that's what the process is for.