I read somewhere - I think in a previous thread, about the New Scientist version of this story, where somebody posted text from a site selling ACR services - that the real interest lies in empirical measurement of what adverts you're seeing, and the context that you see them in is of only secondary interest. I guess the ads you see could be matched up to what you buy. It's an "are our adverts working on this person?" question (and thus "can we show that our advertising service is worth buying"), more than a question of "what advert do we send this person next?"

It is in the paper:

"We find that: (1) ACR network traffic exists when watching linear TV and when using smart TV as an external display using HDMI"

Coming back to this, I see it's in the paper. There's a mention of copyright, and then:

> Netflix prefers to have ACR deactivated during its streaming “in order to preserve the integrity of its subscribers viewing experiences and maintain sole control over measurement of its viewership” (Verance, 2020). The same reasoning applies to FAST channels, which LG and Samsung consider to be ”aggregator apps” (Wolk, 2023), where providers may have agreements restricting ACR usage.

In other words, FAST comes from other providers, who may complain. Random content over HDMI doesn't. However, this doesn't explain why they ignore video streamed from phones (tablets, laptops, etc) wirelessly.