The AIs will keep getting better. We are not too far off from having AIs whose purpose is to establish an online presence that makes others believe there is a real person behind. The AI could even post (generated) videos of a fake person doing very mundane activities.

In the end, I think we will very much need the web of trust and attestation and a reputation score for agents in it, but it will need to include real-world in-person interactions, a degree of government support (i.e. emitting physical id cards for people) and companies selling cameras which are capable of authenticating their footage and any metadata the hardware can attach (date and time, global localization signals, additional radio background, background aural noise and background electrical noise from the power network).

On the other end of the chain, people who consume content and want to verify their authenticity (i.e., people who read the news) will need to opt into all of this or stick to established media outlets. Perhaps some countries will pass laws that help an ordinary citizen consume truthful news, and the essence and potential abuses of those laws will be very interesting.

I don't think there is a way to have a decently robust network of trust where people know others are people without actually knowing the identity of those other people. So, of course, this web of trust will be used by criminals and governments to find their marks.

The social cost of allowing AIs to pose as humans is so high that legislating against it may be the worth it.