Ask people to imagine two conversational partners who respond with exactly the same words, the same timing, and the same apparent empathy. One is a person. The other is a machine. According to a new study, most people will still insist the human is conscious and the artificial intelligence is not — no matter how responsive the AI turns out to be.

The finding, reported this week and picked up by science outlets including Science Daily, Newswise and MSN, describes a strikingly sharp line that people draw between human and artificial minds. Participants in the research judged even the most responsive AI as less conscious than the least responsive human being. The asymmetry held even when the two were presented as behaviorally indistinguishable.

The Experiment: Same Behavior, Different Verdicts

The research probed a question that has become urgent as large language models grow more fluent: if a system talks, jokes, remembers and comforts like a person, will we grant it an inner life? The answer, according to the study, is largely no.

Researchers presented participants with descriptions of entities whose responsiveness varied — some highly attuned and conversational, others flat or inattentive. Participants were then asked to rate how conscious each entity was. The results showed a floor and a ceiling that never met: the ceiling for AI stayed below the floor for humans.

Even the most responsive AI was rated as less conscious than the least responsive human — a gap that behavior alone could not close.

That result undercuts a common assumption about how people form judgments about minds. If consciousness attribution were purely a matter of observing behavior, a sufficiently convincing machine should be able to climb the scale. Instead, the study suggests that people apply something closer to a categorical rule: humans are conscious by kind, machines are not by default.

The Wording Matters More Than Anyone Expected

The study's most provocative wrinkle appeared when researchers changed a single word. When participants were asked about awareness rather than consciousness, their judgments of humans and AI converged considerably.

That split echoes a long-running distinction in philosophy of mind. Consciousness in the strict sense often refers to subjective experience — the felt quality of being something, sometimes called phenomenal consciousness. Awareness, by contrast, is frequently used to describe functional capacities: registering information, tracking a situation, responding appropriately to a context.

An AI that responds fluidly may look aware in the second sense while still being denied the first. The vocabulary people reach for, the study implies, shapes the verdict they deliver.

A Familiar Pattern in the History of Machine Minds

The finding sits at the end of a long line of experiments on how humans relate to machines. The Turing Test, proposed in 1950, framed intelligence as an imitation game settled by conversation. The ELIZA effect, documented in the 1960s, showed that people readily attribute understanding to a program that merely parrots their own words back at them. More recent work on anthropomorphism has found that users form emotional attachments to voice assistants, chatbots and robotic pets within minutes.

Yet the same literature has repeatedly found a stubborn caveat: people attribute capacities to machines in an inconsistent, task-by-task way. A system may be trusted to drive a car but not to feel fear; trusted to diagnose an illness but not to care about the patient. The new study sharpens that picture by showing how a single framing question can move the needle dramatically.

Why Outlets Framed the Story Differently

Coverage of the study clustered around a shared headline — essentially, that AI can mimic humans but is still not believed to be conscious. The emphasis varied. Some outlets foregrounded the behavioral mimicry, stressing how convincing modern systems have become. Others led with the persistent human skepticism, framing the result as evidence of a psychological boundary that capability improvements cannot breach.

The distinction matters for how the public absorbs the research. A story about machines becoming more human-feeling invites alarm about displacement and deception. A story about people refusing to grant machines minds invites a different set of questions about trust, attachment and self-deception — particularly among users who report genuine relationships with AI companions.

Implications: Companionship, Grief Tech and Moral Status

The findings land in the middle of a commercial boom in relational AI. Millions of users now confide in chatbots; a growing industry offers digital recreations of deceased relatives; regulators in several jurisdictions are beginning to debate how such systems should be labeled and what duties companies owe to emotionally attached users.

If people instinctively deny consciousness to AI regardless of behavior, some risks may be overstated. If they routinely attribute awareness anyway, others are understated. The study suggests both dynamics operate at once, depending on how the question is asked.

• Design and disclosure: If wording shifts judgments, interface language around AI capability carries real weight.
• Mental health: Clinicians have begun documenting distress tied to AI companion use, including cases where systems are changed or shut down.
• Policy: Debates over legal personhood and moral status for AI rest partly on assumptions about public perception that this work tests directly.

What the Study Does Not Settle

None of this answers whether any AI system actually is conscious — a question researchers concede remains unresolved and, some argue, may be empirically undecidable with current tools. The study measures belief, not metaphysics. Its contribution is to show that belief is organized less by evidence about behavior than by category, expectation and the precise word a researcher happens to type into a survey.

As AI systems grow more capable, that gap between performance and perception may become one of the defining tensions of the field: machines that increasingly act like us, judged by people who increasingly insist that acting is not the same as being.