I wonder if those devices simply looked at the volume and said, "If it's high [perhaps for a few seconds at a time], then it's probably an ad". Seems like a pretty good heuristic.

I had one of those but it was built into our TV, a Magnavox. It relied on a small black frame that local broadcasters used to insert their ads. National broadcaster like NBC would stream to their local affiliates. And the stream would have blank spaces to an insert their own ads. It’s very fast ,0.1s, that your eye would see briefly but to smart enough electronics was very perceptible. Once the commercial block was detected volume goes down. And the subsequent black frame turns it back up.

It's a tiny algorithm listening to output volume and reacting to drastic changes via the VolUp/Dn buttons on your remote (figuratively): https://en.wikipedia.org/wiki/Dynamic_range_compression

"Night mode", if it is present, should also do this to a lesser extent. I am grateful for this feature in movies where light chit-chat is interrupted by window-shattering explosion sounds.

TV ads (almost) never interrupted the flow of the show you were watching, because the show was specifically designed around the ad breaks. Radio ads always came between songs.

YouTube is now just sticking ads in any old where in videos. The middle of a song. The middle of a sentence. It's revolting, and in many cases it ruins the content you're actually trying to watch.