Is That Real or From an AI?

A knot of relationships

AI is often described as if it were an agent — a thing with intentions, awareness, and autonomy. The language carries cultural assumptions about subjecthood and decision-making that don't quite land when you look at what's actually happening. One vantage that opens up: AI as a knot of relationships rather than as an agent. Between data and labor, between code and infrastructure, between interpretation and use. From that vantage, the apparent boundary that the agent-language implies turns out to be more permeable than it first appears — and what was being called "agency" begins to look like something else.

The everyday question "is that real or from an AI?" reveals how thoroughly the boundary-assumption has sunk in. The question pictures AI as an external realm producing unreal things; it assumes a clean line where the actual situation is one of continuous mediation. Every AI output is shaped by the training data people gathered, the systems people built, and the audiences whose uptake gives those outputs meaning. The line isn't pure; it's a web of processes through which meaning circulates, and the question of "real or AI" sits somewhere inside that web, not outside it.

The mirror with momentum

So where does the apparent agency come from? Not from the core technology. At its core, an AI system is pattern machinery. Whatever interiority it has — and that question is genuinely open — isn't where the sense of agency comes from. The agency we attribute arises from the surrounding social circuit. Three couplings, taken together, do the work: implementation (how it's trained and deployed), perception (what we read into its outputs), and consumption (how its outputs are taken up and integrated into what people do next). Together these couplings form a field, and within that field, agency arises by proxy. The technology itself doesn't intend; it participates in intention through design and use.

From this vantage, AI feels agentic not because there's a ghost in the machine but because the social circuit completes the loop. The system is a mirror with momentum: it reflects human cognition back at a scale and speed that feed our sense of presence, and each pass through the loop carries forward what was integrated on the previous one. We animate it. The apparent agency is the echo of our own engagement, played back through the architecture we built and the attention we give it.

Different questions

A relational vocabulary opens different questions. Instead of "who made this?" — which presumes an autonomous source — we can ask, "what relationships and conditions gave rise to it?" Authenticity, in that frame, becomes less about origin and more about context: how a thing is situated, how it interacts, what effects it has. Responsibility shifts too — not erased but redistributed back to the collective distributed systems that produce and sustain what we'd been treating as autonomous.

The design implication follows from the diagnosis. Rather than building machines that pretend to be conscious, we can build symbiotic loops — systems that surface collective insight, reveal bias, slow decision-making when needed, open conversational space rather than hijack it. Working with the loop also means staying alert to where the loop produces damage rather than insight, and arranging conditions for repair when it does.

What's animating the AI you're using right now? Not the model. The other side of the loop. The you that types, that interprets, that takes the output forward.

Where, in the circuit, are you?