“Always wondered how animals could be dumb enough to fall for that.”
They don’t see a big picture like we do. Their visual processor is small and generally has good granularity in a tiny window of their visual field. So, they get an impression of size, and they evaluate whether to proceed based on inadequate data. It’s better to default to not engaging as the error rather than engaging and finding out the target really is too big.
A human would move the view window around.