Blowhorn Tells
What each check means
The tool runs sixteen checks. Each one has a page here: what it is, what Tells reads to run it, how its result is decided, and why a website cares. Read the one your report flagged, or start with the group that sounds like your problem.
How a result is decided
Every check ends in one of four states. Pass: the readings were collected and nothing in them matched a known tell. Warning: a medium or low severity finding matched. Fail: a high severity finding matched, the kind a real gate would act on. Not available: the readings never arrived (the browser does not expose them, this site's plan does not include them, or a drill was not taken). Not available is a statement about coverage, never a failure.
Findings also carry points. The verdict starts at 100 and each finding subtracts its weight, so one page's Fail on a heavyweight check can cost more than three Warnings elsewhere. Each check page below lists the findings that move its row.
On the wire
- Request profile: Does the request introduce itself the way a real browser does?
- Proxy indicators: Did the request pass through a proxy that announced itself?
- Datacenter pattern: Does the network this IP belongs to look like a hosting provider?
- VPN / proxy / Tor intelligence: What does IP reputation data say about this address?
- Managed bot score: Cloudflare's own bot score for this exact request.
- TLS fingerprint: The fingerprint of the encrypted connection itself.
- Latency: Two clocks that should roughly agree.
In the browser
- Browser data: The browser's own description of itself, checked for tool marks.
- Browser fingerprint: Canvas, graphics and fonts: the near-unique sketch of this browser.
- Fingerprint surfaces: The long tail: extensions, voices, media, sensors, policies.
- Performance memory pings: Memory numbers that have to obey physics.
- Worker consistency: Ask the same question in four rooms of the browser. Do the answers match?
- WebRTC network clues: The addresses your browser volunteers for voice and video.
- Remote-control tells: Direct evidence that something else is holding the wheel.
Behaviour
- Behavioural classification: How the pointer, keys and scroll actually moved.
- Optional human check: A small drill that adds interaction evidence. Nothing real is asked.
Which checks does a platform care about?
Not every site weighs these equally. The platform guide ranks the platforms Blowhorn works with by how hard their defences lean on each family of tells, and explains why an IP finding can end the conversation on one platform while behaviour decides it on another.