I think the end state of this is likely a liveness model via video+audio+device heuristics If people want to interact, they need to speak their interactions -> system proves liveness and writes to the network AI slop is then at best the proxy of someone reading a script (which is maybe also detectable with vision tools on eye movements)