>Not sure what this technique is called or how it's done

Downscaling/downsampling attack.

Commonly used against AI systems either to pass filters or poison data.

https://arxiv.org/abs/2010.02456

If you know the sampling rate of the sampling filter you want to pass you can do some tricky encoding to show a different image at that particular resolution bypassing AI/content recognition systems. This said if it starts happening a lot Google will take those images/videos and feed them into a learning system that looks for those patterns and preemptively marks the video.