Not really. We would know people by proxy and referral through other real people. Like how real life works. And actually, over a large duration of time, the real nature of people eventually surface - even the nature of those who successfully pretend to be someone else that they are not. I dont expect that it would be different in the case of AI - it should actually be easier to tell that an account is an AI in the long run. Real people are rather sticky in their ways and character for large durations of time. Their quirks show. The AI constantly evolves and changes.

Imagine Google's next Big Thing: Google Advisor. It's an AI that rates all content you consume. It tells you whether it is AI-generated or human-generated, reliably. Web, forums, chats, SMS, e-mail, even billboards and other offline ads. Also images, sound and video, it's multimodal. All your phone calls, video calls, music you listen to, movies you watch. Anything you can point your camera to.

It's free, but you pay with your data, as always. What you consume, when, how and how much. Also in what mood and your emotional reactions to it, via accelerometer and other side channels. You can opt out of the latter two, the switch is buried somewhere deep in the settings.

The real product is ads that are clearly AI-generated but still acceptable by you. Sometimes even likable.