Damn, I cannot recall where I’ve seen this, but I saw a video about dedicated chips for audio recognition. They said the special phrase is basically the only thing those chips can recognise. After you said the magical words it triggers a response from the device. If this is true for the iPhone? I don’t really know. What boggles my mind about this is that you set up Siri at first and it only responds if it’s the right voice.

For dashcams and other devices I think this might be true since it’s way cheaper than building a software around this task.

So it is kind of listening the whole time. But only with a low power stupid enough chip that recognises the trigger and nothing more.