Home / Tests and Evidence

How to Test If AI Is Conscious: The Indicator Checklist

A row of seed pods on a wall, some glowing and some dark

How do you test if an AI is conscious? There is no single test that can settle it, because nobody can observe consciousness directly, even in other people. The method researchers now use is a checklist: they take the features that leading theories of consciousness say matter, look for those features in an AI system, and treat what the AI says about itself with great caution.

How to test if AI is conscious with indicators

The approach was set out in a 2023 report, Consciousness in Artificial Intelligence, by 19 philosophers, neuroscientists and AI researchers. Instead of asking whether a machine seems conscious, they asked what the best-supported scientific theories say consciousness requires, and turned each requirement into an "indicator property" stated in computational terms.

The indicators came from five families of theory:

  • Recurrent processing theory: signals must loop back, not only pass forwards.
  • Global workspace theory: information must be broadcast to many processes at once, as explained in global workspace theory and AI.
  • Higher-order theories: a system must in some sense represent its own mental states.
  • Predictive processing: the system models its inputs and corrects its predictions.
  • Attention schema theory: the system keeps a model of its own attention.

The more indicators a system has, the more seriously the theories that produced them would take its claim. Applying the list to the AI systems they examined, the authors concluded that no current AI system is conscious.

A follow-up article, accepted in October 2025, develops the idea. It argues that indicators drawn from neuroscientific theories can be tested, and can inform estimates of how likely a particular AI system is to be conscious, rather than giving a simple yes or no.

A lens revealing glowing veins inside a fallen leaf

Why asking the AI does not work

The obvious test, asking, is the weakest. Language models are trained on vast amounts of human writing, much of it about thoughts and feelings, so they can describe experience fluently whether or not they have any. 80,000 Hours makes the point in its review of the field: AI self-reports cannot be trusted, and behavioural, theory-based and architectural tests all have serious drawbacks.

One proposal tried to get round the problem. An article in Scientific American's former blog network described an AI Consciousness Test, a series of increasingly demanding questions about inner experience, and said the AI would have to be kept from learning human ideas about consciousness, so that it could not simply repeat them. That condition is impossible for today's language models, which learn from exactly that material. What chatbots do say about themselves is covered in is ChatGPT conscious.

A closer look: testing whether a model can look inward

A different line of research tests a narrower ability: whether a model can notice its own internal states. In experiments published in October 2025, Anthropic recorded the internal activation pattern for a known concept, injected it into Claude models during unrelated tasks, and asked whether they noticed anything unusual. Even with the best method, Claude Opus 4.1 showed this kind of awareness only about 20% of the time, and it often missed the concept or invented one.

Anthropic was explicit about the limits: its results do not tell us whether Claude, or any other AI system, might be conscious. Noticing an internal state is one of the abilities some theories care about, not consciousness itself.

Checklist: reading a claim that an AI passed a test

  • Which test, from which theory? An indicator only counts if you accept the theory behind it.
  • Was the AI simply asked? Self-reports from models trained on human writing are weak evidence on their own.
  • How many indicators, and how strongly? Assessments now speak in likelihoods, not verdicts.
  • Is the claim about introspection or experience? Detecting internal states is not the same as having experiences, as the Chinese room argument warns in a different way.

More in Tests and Evidence · Report a correction