There’s no way to distinguish between zombie mimicry (learned role-play or scripted content) and organic thought, ie thinking resulting from quantum phenomena akin to that in biological processes.

The only real test is AI agents that independently learn to self-correct and self-regulate. Memory constraints are not the bottleneck here because tiny multi-cellular/animal brains do this all the time.

Absent validation and success in AI self-correction and AI self-regulation, the only thing we can say here is there is a mass delusion between humans that their AI agents are sentient or conscious.