how does reading pixels get you banned over anything else you mentioned
Scripts using vision are better at performing less predictable actions, but this comes with downsides...
They need access to the game’s pixels
A vision script has to constantly “see” the screen. That usually means:
Capturing frames from the game window
Hooking into graphics APIs (DirectX/OpenGL/Vulkan)
Or using high-frequency screen capture calls
From Blizzard’s perspective, this can show up as:
Unusual rendering hooks
Repeated frame readbacks from the GPU
External processes interacting tightly with the game window
Many vision bots don’t just passively capture—they also:
Draw overlays (bounding boxes, highlights, debug info)
Inject code into the game process to get cleaner frames
Use DLL hooks to intercept rendering
Anti-cheat systems are often designed to flag exactly these behaviors, regardless of intent.
Even if the script avoids injection, it still leaves traces:
A companion process constantly reading screen data
High-frequency calls to OS APIs like screen capture or window scraping
Suspicious combinations of tools (automation + capture + input simulation)
Blizzard doesn’t need to “see your bot”—they can detect that something unusual is interacting with the game in real time.
Timing correlation between “vision” and input
A vision bot works like:
capture frame → analyze → decide → act
That loop creates a subtle pattern:
Input events occur shortly after frame captures
Reaction times cluster tightly around processing delays
At scale, that can look like:
Consistent “perception → action” latency
Much tighter than human reaction variability
So even though it’s reacting to visuals, it can still look mechanically coupled in a detectable way.
Blizzard correlates system performance telemetry.
This is why, like I said, a simple humanized script moving/clicking repeatedly between 2 coordinates, with just a slightly randomized target and timing, is actually harder to detect than a much more sophisticated vision bot.