Reproducing the image in this way is essentially a manual low-pass filter (although with little control over the parameters), so it's certainly one valid data point with which to test the hypothesis.

This can be thwarted fairly easily https://arxiv.org/pdf/1607.02533.pdf