AI agents now browse the web, fill forms, create accounts and click buttons — at scale, around the clock, increasingly well. CAPTCHAs can't tell them apart anymore; modern models solve puzzles cheaply. useHUMA answers the question that matters in the agentic era: is there an actual human behind this session? — from behavior alone, in one API call, with zero PII.
| Signal | Human | AI agent / bot |
|---|---|---|
| Zero-latency actions | Humans read, hesitate, hover. Load-to-action takes seconds. | Agents act the instant the DOM is ready — often under 800ms. |
| Cursor trajectory | Organic curves, overshoot, micro-corrections on the way to a button. | No mouse at all, or surgical straight lines teleporting between elements. |
| Typing cadence | Irregular inter-key timing, typos, bursts and pauses. | Perfectly uniform intervals — or the whole field set in one programmatic write. |
| Scroll rhythm | Variable speed, momentum, re-reading sections. | Metronomic increments or instant jumps to anchors. |
| Touch physics | Tap duration 80–200ms, pressure variance, natural swipe arcs. | Zero-millisecond taps with no pressure profile. |
| Session history | Behavior consistent with past sessions over time. | First-seen fingerprints acting with suspicious confidence. |
Stop agent-driven fake accounts before they hit your database, burn your email quota, or farm your free tier.
Keep autonomous spam out of contact forms, waitlists and surveys — without making humans solve puzzles.
Detect credential stuffing and hijacked sessions with continuous post-login monitoring, not just a gate at the door.
Research panels, reviews and UGC platforms certify that responses come from people — not synthetic respondents.