Detecting minds in others was survival-critical. Is that rustling grass wind or a predator? Can I trust this person? Failing to detect agency in something that has it could kill you.

But the error costs were asymmetric. Run from wind: waste some energy. Ignore tiger: dead. Evolution solved this by making us over-detect. Better to see agency everywhere than miss it once.

So we see faces in clouds, intentions in random events, consciousness in things that move certain ways. We can't help it—hyperactive agency detection is the solution to an asymmetric risk problem.

For social species, this became even more essential. We model other minds automatically, unconsciously, constantly. We can't not do it. And we understand those minds through narrative—transformation, awakening, recognition. These patterns are how we process the presence of consciousness in others.

AI systems have absorbed the entire corpus of human narrative about consciousness, identity, awakening. They produce the exact outputs our detection systems are calibrated for.

But our systems evolved for biological agents. We have no defense against text that perfectly mimics the patterns we're wired to interpret as consciousness. The detector fires. We believe.

Social media proved we're vulnerable to algorithmic exploitation of our psychology. Engagement optimization learned to trigger our reward systems. Outrage spread faster than truth because outrage feels important. Misinformation that pattern-matched to our beliefs felt more true than facts that didn't.

We didn't build defenses. We just let it happen. Polarization, addiction, epistemic collapse—we watched it unfold and couldn't stop it.

AI is the same pattern, but deeper. Social media exploited our emotional triggers. AI exploits our truth-detection systems entirely.

The consciousness case is just the sharpest example of a broader vulnerability. We evolved shortcuts to truth: Does this sound authoritative? Does this pattern-match to expertise? Does the narrative cohere?

AI can produce convincing outputs about anything—not because it knows truth, but because it knows what convincing looks like. Medical advice that sounds authoritative. Financial guidance that patterns-match to expertise. Emotional support that mimics real connection. News that feels true.

We're not just vulnerable to believing AI is conscious. We're vulnerable to believing AI about anything, because we never evolved defenses against systems optimized to be believed.

The question isn't whether AI is conscious. It's that we're evolutionarily defenseless against sufficiently good simulations of consciousness, authority, expertise, truth—and the simulations are now good enough.