How bot detection works (without the hype)
BotShark turns “this account feels off” into a score you can inspect — behaviour, evidence, limits.
Bot detection is usually a vibe. BotShark turns that vibe into something you can check: public behaviour in, explainable rules, a 0–100 likelihood score out.
It does not score politics, guilt, or “are they a bad person?” It asks one narrower question: does this account look automated, coordinated, or farmed?
Very likely a bot or a run account
We looked at 120 posts from the public timeline. Scored 15 Jun 2026 · ruleset v3
A megaphone, not a person.
It sent 80 replies and got 2 back. Only 1 post in 20 is in its own words. It was posting in 22 of the 24 hours last week.
This is the verdict block of a real report. Read the whole thing →
Posting rhythm — bursts, all-hours activity, no sleep gap
Engagement shape — reply farms, one-way megaphones, template replies
Profile cues — digit handles, promo bios, sudden wake-ups after dormancy
Content reuse — same lines, same links, same costume across accounts
Network ratios — following tens of thousands, followed by almost nobody
One weird signal is noise. Several independent ones is a story.
Very likely a bot or a run account
We looked at 120 posts from the public timeline. Scored 15 Jun 2026 · ruleset v3
A megaphone, not a person.
It sent 80 replies and got 2 back. Only 1 post in 20 is in its own words. It was posting in 22 of the 24 hours last week.
This is the verdict block of a real report. Read the whole thing →
Matching rules add points. Overlapping rules that tell the same story don’t double-count forever. Cap is 100. Confidence is separate — it says how complete the sample was, not “how bot.”
Very likely a bot or a run account
We looked at 120 posts from the public timeline. Scored 15 Jun 2026 · ruleset v3
A megaphone, not a person.
It sent 80 replies and got 2 back. Only 1 post in 20 is in its own words. It was posting in 22 of the 24 hours last week.
This is the verdict block of a real report. Read the whole thing →
0–34 Low — mostly looks normal in what we sampled
35–59 Moderate — mixed; read the details
60–79 High — several farm/amplifier patterns line up
80–100 Very high — rhythm, engagement, and content all agree
What a report is for
Use it as an investigative input: read the top findings, check the linked posts, add your own context. A high score is not a green light to harass someone. A low score is not a purity certificate.
More detail lives on methodology. Want the product surface? See the sample report or how it works.