How would Google's backend know whether the traffic comes from their own frontend vs a utility like youtube-dl (assume for the sake of argument that the utility is trying to make the traffic look like normal YouTube traffic)?

Youtube proper does a lot of JS, eg. a view is usually only counted after ~30 seconds[0] and does other things like logging the current timecode so that it can return to that time in case the user clicks off the video and returns to it later[1].

0: https://www.tubics.com/blog/what-counts-as-a-view-on-youtube....

1: https://i.judge.sh/rare/Cherry/chrome_H0p9bAoLgy.png the red line is how far in i've watched.

If the actual video URL is somehow grabbable from the API you might miss whatever triggers a +1 view from the website in theory

It does, it shows up as a "Other YouTube features" for the source.

Though, YouTube seems to do some aggressive deduping here, if I download from 5 different IPs, excluding the IP I'm logged into YouTube's services, I see 5 views. If I download 50 times from one IP, I see one view.

Good to know, thanks for the correction!