I think that no amount of anecdotal evidence would be enough. For a very long times doctors had anecdotal evidence that bloodletting saved patient, yet they were fooled by their bias. I’m not saying advertising isn’t spying on our microphones, I don’t know, it might be. But it doesn’t seem very plausible to me: the amount of processing necessary, and the amount of network seems way too high. Also, voice recognition is still not great currently, it was even worse years ago.