One in Ten Americans Thinks AI is Conscious
- Carneades.org
- 1 hour ago
- 8 min read
Around 10% of Americans think that our current artificial intelligence models (LLMs like ChatGPT) are conscious according to a study by Carneades.org conducted in March of 2026. The sentiment was tested with two different questions. On the first question, when asked to select from a list of things that have consciousness, 9.4% of Americans selected "Artificial Intelligence Systems." On the second question, 11.2% of Americans agreed (6.5%) or strongly agreed (4.7%) with the statement "AI LLMs (like ChatGPT) have consciousness."
Overall, artificial intelligence systems ranked below most other groups of things in terms of the share of Americans who thought they had consciousness. More than twice as many people (26%) thought that plants had consciousness, and around six times as many thought that animals had consciousness (61%). Artificial intelligence beat out only normal computers and smartphones (3.3%) and rocks (3.7%).
The Philosophy of Mind
An important debate in the philosophy of mind surrounds what exactly constitutes consciousness and who or what has it. This arises from the fact that as philosophers describe it, consciousness is inherently not something that can be measured by anyone outside of your own mind. Consciousness, in simple terms the first person experience of sensing and responding to the world is something that is inherently mental, not physical and so cannot be detected by physical sciences directly.
As an aside, when neuroscientists speak of consciousness, they are not actually talking about mental states, they are talking about measurable "Neural Correlates of Consciousness". But these can't be proven to be linked in any way to actual consciousness, since consciousness itself cannot be measured. A demonstration of this issue is succinctly presented by David Chalmer's thought experiment on philosophical zombies (people who have all of the physical attributes of other humans, but not the mental states or consciousness).
Since we cannot detect consciousness with scientific means, it falls to philosophers to work to determine whether something has consciousness. While it may be difficult to prove that your neighbors are not actually philosophical zombies, most people think their fellow adults have consciousness (the study finds that 96% of Americans think human adults have consciousness). However the question gets more contentious when you try to determine if something like a computer or an artificial intelligence system has consciousness.
A famous proposal for determining whether computers have consciousness was proposed by the computer scientist Alan Turing (called the Turing test) which argues that once a computer chat bot can fool a human into thinking that they are real, they have consciousness (as they have the same claim to consciousness that any other human does). Many modern LLMs can fool at least some people into thinking that they are human. There are some philosophers who have pushed back on the idea that passing the Turing Test is sufficient for demonstrating consciousness, most notably John Searle with his famous Chinese Room thought experiment. For more on these thought experiments, see the videos below.
According to our survey, most Americans agree more with Searle than Turing. For more on the survey itself and the methodology, check out this blog post.
One in Five Younger Americans thinks AI has Consciousness
The biggest demographic predictor of belief in AI consciousness is age. Younger Americans are much more likely to think that AI is conscious no matter how the question is phrased. Using a linear probability regression model, Americans under 45 were 13.7%*** more likely to select AI as something that is conscious, and 5%*** more likely to select normal computers and smartphones as things that are conscious. Millennial and Gen Z Americans were 17.2%*** more likely to agree with the claim that LLMs like ChatGPT have consciousness. Overall around 16% of Millennials and 18% of Gen Z selected AI as something with consciousness and around 20% of each group agreed or strongly agreed with the claim that AI LLMs like Chat GPT have consciousness.
Race and Philosophy Education Associated with Views on AI Consciousness
Looking through the demographics, after age, the strongest associations with the view that AI has consciousness could be found in race and philosophy education. Using a linear probability regression model, non-white people are 11.7%*** more likely to agree or strongly agree with the claim that LLMs have consciousness and 11.0%*** more likely to select AI systems as something that has consciousness. Within this group, Black people are especially likely to believe that AI systems have consciousness: they are 13.4%*** more likely to agree or strongly agree with the statement that LLMs like ChatGPT have consciousness and 18.1%*** more likely to select artificial intelligence systems as something with consciousness.
Philosophy education is also related to views on AI consciousness. The more philosophy education one has, the more likely one is to believe in AI consciousness. Having a philosophy degree increases the chances of selecting AI as something with consciousness by 4.6%*, while having no philosophy education decreases the chances of selecting AI as something with consciousness by 6.5%***.
Skepticism, Determinism, and Moral Intent connected to beliefs about AI Consciousness
There are several philosophical positions which are associated with belief in AI consciousness. Using a dominance analysis, we can filter out which other components of the survey are the most predictive of belief in AI consciousness. Beyond the obvious other questions about AI consciousness, the most predictive questions on a dominance analysis are questions about skepticism, determinism, and whether intent or impact is more important morally.
Belief in the possibility of skeptical scenarios were closely linked with selecting AI as one of the things that has consciousness. Belief that you might be dreaming increase chances of selecting AI as something that is conscious by 11.1%***, and of agreeing with the claim that LLMs have consciousness by 17.1%***. Belief that you might be in the Matrix increases chances of selecting AI as something that is conscious by 14.4%***, and of agreeing with the claim that LLMs have consciousness by 23.1%***.
This link between the possibility of skeptical scenarios and AI consciousness may make sense. If you think that you could be in the Matrix, it seems like your mind could be simulated by a computer. If your mind could be simulated by a computer then a computer could have consciousness.
Belief in determinism is also linked to belief in AI consciousness (determinism increases chances of belief in AI consciousness by 18%***), as is belief that the intent of an action is more morally important than the consequences (increases changes of belief in AI consciousness by 8.9%***). Check out our blog on American's views on free will for more on the connection between determinism and beliefs about AI consciousness. The waterfall charts below show the full dominance analysis and which other variables were the most predictive of belief in AI consciousness.
A Note On Other Studies
Before finishing, a quick note on some other studies which find that much higher shares of Americans believe that AI is conscious (up to 67%). Most of these articles are referencing this study that was published in Neuroscience of Consciousness in 2024. There are a few things to note about the study. First, it had a smaller sample size than the present study (n=300). Second it used only a single question to measure belief in AI consciousness, while our study uses two with nearly identical results despite different phrasings and formulations.
Most importantly though, the question it asked was for participants to rank AI consciousness on a scale of 1 to 100. Of the 300 participants, 100 selected 1, the remaining 200 selected some number above 1 (hence the 67%). Of those that selected a number above 1, the median was only 16 out of 100, something that seems to indicate most of the participants were much closer to thinking AI does not have consciousness than that it does. The authors say that only 23% thought AI was more likely than not to have consciousness, (something much closer to our results). It is likely that the disparity between the results shared here and the results from this other survey are due to the methodology, not due to changes in the public's viewpoints on AI.
Methodology Specifics
Before leaving, here are the specifics of how the questions we have been focusing on was asked. This study was conducted in March of 2026 on a representative sample of Americans. For more on the overall methodology of this survey consult this blog post. As we publish more blogs on this survey the information on the questions for the other philosophical positions will come out. The best attempt was made to make the statements succinctly represent the four positions, while still being accessible to a general audience without too much philosophical lingo. Views on AI consciousness were assessed using two questions. One was multiple choice where participants were asked to select all things that had consciousness (see full question below). The other was a Likert statement were participants were asked their views on the statement "AI LLMs (like ChatGPT) have consciousness."
Which of the following do you believe have consciousness (select all that apply)?
Rocks
Plants
Non-human animals
Human infants
Human adults
Normal computers and smartphones
Artificial Intelligence systems
None of the above/I don’t know
Data like these are a public good that can help us all understand how normal people view important philosophical questions. Our goal is to put out more surveys like this in the future, but we need your help! Have you ever wanted to ask a philosophical question to a representative sample of Americans? Have you ever been curious about how normal people see questions that are important to philosophers? You now have the chance to get your question on our next nationally representative survey! We need around $2,500 to conduct another survey, hopefully in the next few months. Visit our Patreon page to find out more. Here's how you can help:
Raw Data $10: If you want access to the raw data from the survey we just finished, it is available! You can use it to run and publish whatever analyses you want, just 1) don't publish the raw data yourself and 2) make sure to give credit to Carneades.org. We'll even throw in the data from our pre-survey of 100 people.
Likert Statement $30: Want to see how much Americans agree or disagree with a certain statement? For just $30, you can add a statement to our Likert question, where participants will respond to it with Strongly Agree, Agree, Neither Agree or Disagree, Disagree, or Strongly Disagree. The statement must be about philosophy, less than 150 characters, and safe for work. We'll include the raw data from the next survey.
Multiple Choice Question $150: Want to design and ask your own multiple choice question? You can have up to 1000 characters in the question and up to 500 characters in each of the answers. Only up to 10 answers are allowed. The question must be about philosophy and safe for work. We'll include the raw data from the next survey. We will even throw in a Likert statement (as described above) and the raw data from the next survey for free!
We reserve the right to edit any statement for grammar, spelling, or to clarify any philosophical ideas for the respondents. We also reserve the right to reject any statements, but will give a full refund for any that are rejected.
Notes:
All margins of error are calculated at the 95% confidence level.
Dominance analysis is run with all of the views in the study converted to binary or Likert scales. All likelihood estimates are created using a linear probability model with both dependent and independent variable treated as binaries.
*Significant at the p < 0.1 level
**Significant at the p < 0.05 level
*** Significant at the p < 0.01 level
How to cite this report:
Carneades (2026). One in Ten Americans Think AI Is Conscious. Experimental Philosophy. https://www.carneades.org/post/one-in-ten-americans-think-ai-is-conscious
Comments