{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,28]],"date-time":"2026-10-28T02:43:27Z","timestamp":1793155407387,"version":"4.5.0"},"reference-count":67,"publisher":"National Academy of Sciences","issue":"21","license":[{"start":{"date-parts":[[2026,5,19]],"date-time":"2026-05-19T00:00:00Z","timestamp":1779148800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100014895","name":"Open Philanthropy Project","doi-asserted-by":"publisher","award":["30327210"],"award-info":[{"award-number":["30327210"]}],"id":[{"id":"10.13039\/100014895","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.pnas.org"],"crossmark-restriction":true},"short-container-title":["Proc. Natl. Acad. Sci. U.S.A."],"published-print":{"date-parts":[[2026,5,26]]},"abstract":"<jats:p>The Turing test has been widely discussed as a test of machine intelligence, but it also provides a measure of how humans distinguish other humans from machines. We evaluated 4 systems (ELIZA, GPT-4o, LLaMa-3.1-405B, and GPT-4.5) in two randomized, controlled, and preregistered Turing tests on independent populations. Participants had 5 min conversations simultaneously with another human participant and one of these systems before judging which conversational partner they thought was human. When prompted to adopt a humanlike persona, GPT-4.5 was judged to be the human 73% of the time: significantly more often than interrogators selected the real human participant. LLaMa-3.1, with the same prompt, was judged to be the human 56% of the time\u2014not significantly more or less often than the humans it was being compared to. Without these prompts, however, the same models performed significantly worse (38% and 36%), and did not consistently outperform baseline models, ELIZA and GPT-4o (23% and 21%, respectively). A third study replicated these results in 15-min games: two PERSONA-prompted models achieved pass rates of 56% and 59%. The results constitute empirical evidence that artificial systems can pass a standard three-party Turing test. Interrogators\u2019 reasoning focused more on stylistic and socio-emotional aspects of human behavior rather than more traditional notions of intelligence. The results have implications for debates about what kind of intelligence is exhibited by large language models, the social impacts these systems are likely to have, and the aspects of human behavior that people continue to see as unique.<\/jats:p>","DOI":"10.1073\/pnas.2524472123","type":"journal-article","created":{"date-parts":[[2026,5,19]],"date-time":"2026-05-19T16:04:32Z","timestamp":1779206672000},"update-policy":"https:\/\/doi.org\/10.1073\/pnas.cm10313","source":"Crossref","is-referenced-by-count":14,"title":["Large language models pass a standard three-party Turing test"],"prefix":"10.1073","volume":"123","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6609-8966","authenticated-orcid":false,"given":"Cameron R.","family":"Jones","sequence":"first","affiliation":[{"id":[{"id":"https:\/\/ror.org\/0168r3w48","id-type":"ROR","asserted-by":"publisher"}],"name":"Department of Cognitive Science","place":["La Jolla"]},{"name":"University of California San Diego","place":["La Jolla"]},{"id":[{"id":"https:\/\/ror.org\/05qghxh33","id-type":"ROR","asserted-by":"publisher"}],"name":"Department of Psychology","place":["Stony Brook"]},{"name":"Stony Brook University","place":["Stony Brook"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9395-9151","authenticated-orcid":false,"given":"Benjamin K.","family":"Bergen","sequence":"additional","affiliation":[{"id":[{"id":"https:\/\/ror.org\/0168r3w48","id-type":"ROR","asserted-by":"publisher"}],"name":"Department of Cognitive Science","place":["La Jolla"]},{"name":"University of California San Diego","place":["La Jolla"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"341","published-online":{"date-parts":[[2026,5,19]]},"reference":[{"key":"e_1_3_4_1_2","doi-asserted-by":"publisher","DOI":"10.1093\/mind\/LIX.236.433"},{"key":"e_1_3_4_2_2","doi-asserted-by":"publisher","DOI":"10.1016\/S1364-6613(00)01453-4"},{"key":"e_1_3_4_3_2","doi-asserted-by":"publisher","DOI":"10.1093\/mind\/LXXIII.290.234"},{"key":"e_1_3_4_4_2","first-page":"972","article-title":"Turing test considered harmful","volume":"1","author":"Hayes P.","year":"1995","unstructured":"P. Hayes, K. Ford, Turing test considered harmful. IJCAI 1, 972\u2013977 (1995).","journal-title":"IJCAI"},{"key":"e_1_3_4_5_2","doi-asserted-by":"publisher","DOI":"10.1023\/A:1011288000451"},{"key":"e_1_3_4_6_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11023-020-09544-5"},{"key":"e_1_3_4_7_2","doi-asserted-by":"publisher","DOI":"10.1007\/s00146-020-00946-8"},{"key":"e_1_3_4_8_2","unstructured":"I. D. Raji E. M. Bender A. Paullada E. Denton A. Hanna AI and the everything in the whole wide world benchmark. arXiv [Preprint] (2021). https:\/\/arxiv.org\/abs\/2111.15366 (Accessed 16 September 2025)."},{"key":"e_1_3_4_9_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2215907120"},{"key":"e_1_3_4_10_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41562-024-02096-z"},{"key":"e_1_3_4_11_2","doi-asserted-by":"publisher","DOI":"10.1016\/S0378-2166(02)80001-7"},{"key":"e_1_3_4_12_2","doi-asserted-by":"publisher","DOI":"10.2307\/2184371"},{"key":"e_1_3_4_13_2","unstructured":"G. Oppy D. Dowe \u201cThe Turing test\u201d in The Stanford Encyclopedia of Philosophy E. N. Zalta Ed. (Metaphysics Research Lab Stanford University ed. Winter 2021 2021)."},{"key":"e_1_3_4_14_2","volume-title":"Artificial Intelligence a Modern Approach","author":"Russell S. J.","year":"2010","unstructured":"S. J. Russell, Artificial Intelligence a Modern Approach (Pearson Education, Inc., 2010)."},{"key":"e_1_3_4_15_2","unstructured":"D. C. Dennett The problem with counterfeit people. The Atlantic (2023). https:\/\/www.theatlantic.com\/technology\/archive\/2023\/05\/problem-counterfeit-people\/674075\/. Accessed 16 September 2025."},{"key":"e_1_3_4_16_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.techfore.2023.122634"},{"key":"e_1_3_4_17_2","doi-asserted-by":"crossref","unstructured":"T. Eloundou S. Manning P. Mishkin D. Rock GPTs are GPTs: Labor market impact potential of LLMs. Science 384 1306\u20131308 (2024).","DOI":"10.1126\/science.adj0998"},{"key":"e_1_3_4_18_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.patter.2024.100988"},{"key":"e_1_3_4_19_2","unstructured":"M. Burtell T. Woodside Artificial influence: An analysis of AI-driven persuasion. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2303.08721. Accessed 16 September 2025."},{"key":"e_1_3_4_20_2","doi-asserted-by":"publisher","DOI":"10.1007\/s10462-026-11517-6"},{"key":"e_1_3_4_21_2","unstructured":"S. M. Shieber Lessons from a restricted Turing test. arXiv [Preprint] (1994). https:\/\/arxiv.org\/pdf\/cmp-lg\/9404002 (Accessed 16 September 2025)."},{"key":"e_1_3_4_22_2","doi-asserted-by":"crossref","unstructured":"H. Loebner \u201cHow to hold a Turing test contest\u201d in Parsing the Turing Test R. Epstein G. Roberts G. Beber Eds. (Springer Netherlands Dordrecht 2009) pp. 173\u2013179.","DOI":"10.1007\/978-1-4020-6710-5_12"},{"key":"e_1_3_4_23_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.adq9356"},{"key":"e_1_3_4_24_2","doi-asserted-by":"crossref","unstructured":"C. Biever ChatGPT broke the Turing test \u2014 the race is on for new ways to assess AI. Nature 619 686\u2013689 (2023).","DOI":"10.1038\/d41586-023-02361-7"},{"key":"e_1_3_4_25_2","unstructured":"A. James ChatGPT has passed the Turing test and if you\u2019re freaked out you\u2019re not alone. TechRadar (2023). https:\/\/www.techradar.com\/opinion\/chatgpt-has-passed-the-turing-test-and-if-youre-freaked-out-youre-not-alone. Accessed 16 September 2025."},{"key":"e_1_3_4_26_2","doi-asserted-by":"publisher","DOI":"10.1080\/0020174X.2024.2446241"},{"key":"e_1_3_4_27_2","doi-asserted-by":"crossref","unstructured":"M. Giunti ChatGPT-4 and other LLMs in the Turing test: A critical analysis. Minds Mach. 36 8 (2026).","DOI":"10.1007\/s11023-026-09760-5"},{"key":"e_1_3_4_28_2","unstructured":"D. Jannai A. Meron B. Lenz Y. Levine Y. Shoham Human or not? A gamified approach to the Turing test. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2305.20010 (Accessed 16 September 2025)."},{"key":"e_1_3_4_29_2","first-page":"5183","article-title":"Does GPT-4 pass the Turing test?","volume":"290","author":"Jones C. R.","year":"2024","unstructured":"C. R. Jones, B. K. Bergen, Does GPT-4 pass the Turing test? NAACL 290, 5183\u20135210 (2024).","journal-title":"NAACL"},{"key":"e_1_3_4_30_2","doi-asserted-by":"crossref","unstructured":"C. R. Jones I. Rathi S. Taylor B. K. Bergen \u201cPeople cannot distinguish GPT-4 from a human in a Turing test\u201d in Proceedings of the 2025 ACM Conference on Fairness Accountability and Transparency (2025) pp. 1615\u20131639.","DOI":"10.1145\/3715275.3732108"},{"key":"e_1_3_4_31_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11023-025-09711-6"},{"key":"e_1_3_4_32_2","unstructured":"S. Temtsin D. Proudfoot D. Kaber C. Bartneck The imitation game according to Turing. arXiv [Preprint] (2025). https:\/\/arxiv.org\/abs\/2501.17629 (Accessed 16 September 2025)."},{"key":"e_1_3_4_33_2","unstructured":"C. R. Jones B. K. Bergen 3-party Turing test (Prolific). Open Science Framework. https:\/\/osf.io\/f4hj9. Accessed 13 March 2025."},{"key":"e_1_3_4_34_2","unstructured":"C. R. Jones B. K. Bergen A three-party Turing test: Evaluating advanced LLMs\u2019 ability to pass as human. Open Science Framework. https:\/\/osf.io\/m4fst. Accessed 11 March 2025."},{"key":"e_1_3_4_35_2","unstructured":"C. R. Jones B. K. Bergen 3-party Turing test replication (15 minutes). Open Science Framework. https:\/\/osf.io\/msygn. Accessed 13 January 2026."},{"key":"e_1_3_4_36_2","unstructured":"OpenAI et al. GPT-4 technical report. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2303.08774 (Accessed 16 September 2025)."},{"key":"e_1_3_4_37_2","unstructured":"A. Grattafiori et al. The Llama 3 herd of models. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2407.21783 (Accessed 16 September 2025)."},{"key":"e_1_3_4_38_2","unstructured":"OpenAI et al. GPT-4o system card. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2410.21276 (Accessed 16 September 2025)."},{"key":"e_1_3_4_39_2","doi-asserted-by":"publisher","DOI":"10.1145\/365153.365168"},{"key":"e_1_3_4_40_2","unstructured":"Prolific Quickly find research participants you can trust. https:\/\/www.prolific.com\/. Accessed 16 September 2025."},{"key":"e_1_3_4_41_2","unstructured":"C. R. Jones B. K. Bergen Data and analysis code for \u201cLarge language models pass a standard three-party Turing test.\u201d Open Science Framework. https:\/\/osf.io\/jk7bw. Deposited 3 April 2025."},{"key":"e_1_3_4_42_2","unstructured":"Z. Xu Y. Liu G. Deng Y. Li S. Picek A comprehensive study of jailbreak attack versus defense for large language models. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2402.13457v2 (Accessed 16 September 2025)."},{"key":"e_1_3_4_43_2","unstructured":"M. Hasan J. Zhao E. Hoque Hal: Inducing human-likeness in LLMs with alignment. arXiv [Preprint] (2026). http:\/\/arxiv.org\/abs\/2601.02813 (Accessed 16 September 2025)."},{"key":"e_1_3_4_44_2","unstructured":"C. Snell D. Klein R. Zhong Learning by distilling context. arXiv [Preprint] (2022). http:\/\/arxiv.org\/abs\/2209.15189 (Accessed 16 September 2025)."},{"key":"e_1_3_4_45_2","doi-asserted-by":"crossref","unstructured":"U. Le\u00f3n-Dom\u00ednguez et al. AI-driven agents with prompts designed for high agreeableness increase the likelihood of being mistaken for a human in the Turing test. arXiv [Preprint] (2024). http:\/\/arxiv.org\/abs\/2411.13749.","DOI":"10.2139\/ssrn.5227056"},{"key":"e_1_3_4_46_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4020-6710-5"},{"key":"e_1_3_4_47_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41562-023-01710-w"},{"key":"e_1_3_4_48_2","doi-asserted-by":"publisher","DOI":"10.1038\/d41586-023-04058-3"},{"key":"e_1_3_4_49_2","doi-asserted-by":"crossref","unstructured":"R. Kurzweil M. Kapor A. Wager \u201cA wager on the Turing test\u201d in Parsing the Turing Test R. Epstein G. Roberts G. Beber Eds. (Springer Netherlands Dordrecht 2009) pp. 463\u2013477.","DOI":"10.1007\/978-1-4020-6710-5_27"},{"key":"e_1_3_4_50_2","doi-asserted-by":"publisher","DOI":"10.1080\/09528138908953691"},{"key":"e_1_3_4_51_2","doi-asserted-by":"publisher","DOI":"10.1017\/S0140525X00005756"},{"key":"e_1_3_4_52_2","doi-asserted-by":"crossref","unstructured":"R. E. Horn \u201cThe Turing test: Mapping and navigating the debate\u201d in Parsing the Turing Test R. Epstein G. Roberts G. Beber Eds. (Springer Netherlands Dordrecht 2009) pp. 73\u201388.","DOI":"10.1007\/978-1-4020-6710-5_5"},{"key":"e_1_3_4_53_2","doi-asserted-by":"publisher","DOI":"10.1037\/0003-066X.51.2.77"},{"key":"e_1_3_4_54_2","doi-asserted-by":"publisher","DOI":"10.1016\/S0004-3702(01)00129-1"},{"key":"e_1_3_4_55_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.134.3495.2011"},{"key":"e_1_3_4_56_2","volume-title":"The Most Human Human: What Talking with Computers Teaches Us about What It Means to Be Alive","author":"Christian B.","year":"2011","unstructured":"B. Christian, The Most Human Human: What Talking with Computers Teaches Us about What It Means to Be Alive (Anchor, 2011)."},{"key":"e_1_3_4_57_2","unstructured":"A. Srivastava et al. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models (2023). https:\/\/openreview.net\/forum?id=uyTL5Bvosj. Accessed 16 September 2025."},{"key":"e_1_3_4_58_2","unstructured":"E. Davis ChatGPT\u2019s poetry is incompetent and banal: A discussion of (Porter and Machery 2024) (2024). https:\/\/cs.nyu.edu\/~davise\/papers\/GPT-Poetry.pdf. Accessed 16 September 2025."},{"key":"e_1_3_4_59_2","first-page":"67","article-title":"Large language models for enhancing customer lifecycle management","volume":"7","author":"Soni V.","year":"2023","unstructured":"V. Soni, Large language models for enhancing customer lifecycle management. J. Empir. Soc. Sci. Stud. 7, 67\u201389 (2023).","journal-title":"J. Empir. Soc. Sci. Stud."},{"key":"e_1_3_4_60_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.chb.2023.108133"},{"key":"e_1_3_4_61_2","unstructured":"J. Lehman Machine love. arXiv [Preprint] (2023). https:\/\/arxiv.org\/abs\/2302.09248 (Accessed 16 September 2025)."},{"key":"e_1_3_4_62_2","doi-asserted-by":"crossref","unstructured":"H. R. Kirk I. Gabriel C. Summerfield B. Vidgen S. A. Hale Why human\u2013AI relationships need socioaffective alignment. Humanit. Soc. Sci. Commun. 12 728 (2025).","DOI":"10.1057\/s41599-025-04532-5"},{"key":"e_1_3_4_63_2","volume-title":"Alone Together: Why We Expect More from Technology and Less from Each Other","author":"Turkle S.","year":"2011","unstructured":"S. Turkle, Alone Together: Why We Expect More from Technology and Less from Each Other (Basic Books, 2011)."},{"key":"e_1_3_4_64_2","unstructured":"S. El-Sayed et al. A mechanism-based approach to mitigating harms from persuasive generative AI. arXiv [Preprint] (2024). https:\/\/arxiv.org\/abs\/2404.15058 (Accessed 16 September 2025)."},{"key":"e_1_3_4_65_2","doi-asserted-by":"crossref","unstructured":"M. Carroll A. Chan H. Ashton D. Krueger \u201cCharacterizing manipulation from AI systems\u201d in Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms Mechanisms and Optimization V. Manshadi C. Mendler-D\u00fcnner E. Redmiles M. Rodriguez Eds. (Association for Computing Machinery New York NY 2023) pp. 1\u201313.","DOI":"10.1145\/3617694.3623226"},{"key":"e_1_3_4_66_2","unstructured":"W. Brainerd Eliza chatbot in Python. GitHub. https:\/\/github.com\/wadetb\/eliza. Accessed 16 September 2025."},{"key":"e_1_3_4_67_2","unstructured":"C. R. Jones B. K. Bergen Confirmatory Turing test with GPT-4. Open Science Framework. https:\/\/osf.io\/ug4s3. Accessed 22 February 2024."}],"container-title":["Proceedings of the National Academy of Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pnas.org\/doi\/pdf\/10.1073\/pnas.2524472123","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,19]],"date-time":"2026-05-19T16:04:45Z","timestamp":1779206685000},"score":1,"resource":{"primary":{"URL":"https:\/\/pnas.org\/doi\/10.1073\/pnas.2524472123"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,19]]},"references-count":67,"journal-issue":{"issue":"21","published-print":{"date-parts":[[2026,5,26]]}},"alternative-id":["10.1073\/pnas.2524472123"],"URL":"https:\/\/doi.org\/10.1073\/pnas.2524472123","relation":{},"ISSN":["0027-8424","1091-6490"],"issn-type":[{"value":"0027-8424","type":"print"},{"value":"1091-6490","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,19]]},"assertion":[{"value":"2025-09-16","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-03-27","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-05-19","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e2524472123"}}