Elamur Research⌕
☰
home / papers / 2025 / People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text

People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text

Paper · 2025 · Jenna Russell, Marzena Karpinska and Mohit Iyyer

Summary

The paper asks whether people can reliably tell modern LLM output from human prose, given that automatic detectors are often inaccurate, easy to evade and hard to interpret. The authors built 300 English non-fiction articles under 1K words, split into five experiments of 60. Each human-written piece, drawn from eight US outlets, was paired with an AI counterpart generated from the same title and subtitle by GPT-4O, CLAUDE-3.5-SONNET or O1-PRO. Some AI texts were also paraphrased or humanized to evade detection. Annotators hired on Upwork gave each article a binary label, a confidence score, highlighted evidence spans and a written justification. They received no training or feedback.

Annotators who rarely use LLMs performed near chance. Those who regularly use ChatGPT for writing were highly accurate, and a majority vote of five such experts got only one article wrong. Only the commercial Pangram detector matched them, while the experts beat the open-source baselines. The authors release the dataset and code, arguing that expert human judges are a practical, explainable option when the stakes are high.

Key points

  • Expert vs. nonexpert annotators: Experts averaged 92.7% TPR at 4.0% FPR. Nonexperts averaged 56.7% TPR at 51.7% FPR, yet still rated their confidence highly (4.03 out of 5).
  • Hardest setting: For O1-PRO articles with humanization, the expert majority vote reached 100% TPR. Binoculars reached 6.7% and Fast-DetectGPT 23.3%.
  • Cues experts relied on: The most common giveaway was typical "AI vocabulary" such as vibrant, crucial and significantly. Formulaic sentence and document structure (for example, vague upbeat endings) and lack of originality came next. Paraphrasing and humanization did not remove all of these signals.
  • LLM-based imitation of the experts: The authors prompted LLMs with a guidebook distilled from expert explanations. GPT-4O with chain-of-thought plus the guidebook reached 78% average TPR, roughly on par with Binoculars and Fast-DetectGPT. The best O1 setup reached 54%, and both setups struggled with humanized text.
  • Cost and resources: Annotators were paid $2 per text, about $4.9K USD in total. Data and code are released at github.com/jenna-russell/human_detectors.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 works

Cite

Cite as Jenna Russell, Marzena Karpinska and Mohit Iyyer. 2025. People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.acl-long.267

@inproceedings{russell2025people,
  title = {{People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text}},
  author = {Jenna Russell, Marzena Karpinska and Mohit Iyyer},
  booktitle = {Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2025},
  doi = {10.18653/v1/2025.acl-long.267},
  url = {https://doi.org/10.18653/v1/2025.acl-long.267}
}

References

48 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anthropic. 2023. Claude: A language model by anthropic. Accessed: 2025-01-20. find
  2. Anthropic. 2024. Claude 3 model card addendum. https://www-cdn.anthropic.com/ fed9cc193a14b84131812372d8d5857f8f304c52/ Model_Card_Claude_3_Addendum.pdf. Accessed: 2024-12-30. ↗
  3. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… — Language Models are Few-Shot Learners
  4. Tuhin Chakrabarty, Philippe Laban, Divyansh Agarwal, Smaranda Muresan, and Chien-Sheng Wu. 2024. Art or artifice? large language models and the false promise of creativity. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, CHI ’24, New York, NY, USA. Association for Co… find
  5. Yapei Chang, Kalpesh Krishna, Amir Houmansadr, John Frederick Wieting, and Mohit Iyyer. 2024. Post-Mark: A robust blackbox watermark for large language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 8969–8987, Miami, Florida, USA. Association… find
  6. Elizabeth Clark, Tal August, Sofia Serrano, Nikita Haduong, Suchin Gururangan, and Noah A. Smith. 2021. All that‘s ‘human’ is not gold: Evaluating human evaluation of generated text. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International… find
  7. Yao Dou, Maxwell Forbes, Rik Koncel-Kedziorski, Noah A. Smith, and Yejin Choi. 2021. Is gpt-3 text indistinguishable from human text? scarecrow: A framework for scrutinizing machine text. In Annual Meeting of the Association for Computational Linguistics.
  8. Jad Doughman, Osama Mohammed Afzal, Hawau Olamide Toyin, Shady Shehata, Preslav Nakov, and Zeerak Talat. 2025. Exploring the Limitations of Detecting Machine-Generated Text. In Proceedings of the 31st International Conference on Computational Linguistics, pages 4274–4281, Abu Dhabi, UAE. Association…
  9. Abhimanyu Hans, Avi Schwarzschild, Valeriia Cherepanova, Hamid Kazemi, Aniruddha Saha, Micah Goldblum, Jonas Geiping, and Tom Goldstein. 2024. Spotting LLMs with binoculars: Zero-shot detection of machine-generated text. In Proceedings of the 41st International Conference on Machine Learning, volume… find
  10. Liam Dugan, Alyssa Hwang, Filip Trhlík, Andrew Zhu, Josh Magnus Ludan, Hainiu Xu, Daphne Ippolito, and Chris Callison-Burch. 2024. RAID: A shared benchmark for robust evaluation of machine-generated text detectors. In Proceedings of the 62nd Annual Meeting of the Association for Computational Lingui…
  11. Xiaomeng Hu, Pin-Yu Chen, and Tsung-Yi Ho. 2023. RADAR: robust ai-text detection via adversarial learning. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023. find
  12. Liam Dugan, Daphne Ippolito, Arun Kirubarajan, and Chris Callison-Burch. 2020. RoFT: A Tool for Evaluating Human Detection of Machine-Generated Text. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 189–196, Online. Association f… find
  13. Liam Dugan, Daphne Ippolito, Arun Kirubarajan, Sherry Shi, and Chris Callison-Burch. 2022. Real or fake text?: Investigating human ability to detect boundaries between human-written and machine-generated text. In AAAI Conference on Artificial Intelligence. find
  14. Bradley Emi and Max Spero. 2024. Technical Report on the Pangram AI-Generated Text Classifier. arXiv preprint. ArXiv:2402.14873 [cs]. — [arXiv:2402.14873](https://arxiv.org/abs/2402.14873) find
  15. Henrique Da Silva Gameiro, Andrei Kucharavy, and Ljiljana Dolamic. 2024. LLM Detectors Still Fall Short of Real World: Case of LLM-Generated Short News-Like Posts. arXiv preprint. ArXiv:2409.03291 [cs]. — [arXiv:2409.03291](https://arxiv.org/abs/2409.03291) find
  16. Matt Gardner, Yoav Artzi, Victoria Basmov, Jonathan Berant, Ben Bogin, Sihao Chen, Pradeep Dasigi, Dheeru Dua, Yanai Elazar, Ananth Gottumukkala, Nitish Gupta, Hannaneh Hajishirzi, Gabriel Ilharco, Daniel Khashabi, Kevin Lin, Jiangming Liu, Nelson F. Liu, Phoebe Mulcaire, Qiang Ning, Sameer Singh, N… — Evaluating Models' Local Decision Boundaries via Contrast Sets
  17. Sebastian Gehrmann, Hendrik Strobelt, and Alexander Rush. 2019. GLTR: Statistical Detection and Visualization of Generated Text. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pages 111–116, Florence, Italy. Association for Computat…
  18. Daphne Ippolito, Daniel Duckworth, Chris Callison-Burch, and Douglas Eck. 2020. Automatic detection of generated text is easiest when humans are fooled. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1808–1822, Online. Association for Computational…
  19. Jiazhou Ji, Ruizhe Li, Shujun Li, Jie Guo, Weidong Qiu, Zheng Huang, Chiyu Chen, Xiaoyu Jiang, and Xinru Lu. 2024. Detecting Machine-Generated Texts: Not Just "AI vs Humans" and Explainability is Complicated. arXiv preprint. ArXiv:2406.18259 [cs]. — [arXiv:2406.18259](https://arxiv.org/abs/2406.18259) find
  20. Marzena Karpinska, Nader Akoury, and Mohit Iyyer. 2021. The perils of using Mechanical Turk to evaluate open-ended text generation. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 1265–1285, Online and Punta Cana, Dominican Republic. Association for C… find
  21. Marzena Karpinska, Nishant Raj, Katherine Thai, Yixiao Song, Ankita Gupta, and Mohit Iyyer. 2022. DEMETR: Diagnosing evaluation metrics for translation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 9540–9561, Abu Dhabi, United Arab Emirates. Associ… find
  22. Marzena Karpinska, Katherine Thai, Kyle Lo, Tanya Goyal, and Mohit Iyyer. 2024. One thousand and one pairs: A “novel” challenge for long-context language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17048–17085, Miami, Florida, USA. Associa… find
  23. John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers, and Tom Goldstein. 2023. A watermark for large language models. In Proceedings of the 40th International Conference on Machine Learning, volume 202 of Proceedings of Machine Learning Research, pages 17061–17084. PMLR.
  24. Kalpesh Krishna, Yixiao Song, Marzena Karpinska, John Wieting, and Mohit Iyyer. 2023. Paraphrasing evades detectors of ai-generated text, but retrieval is an effective defense. In Advances in Neural Information Processing Systems, volume 36, pages 27469– 27500. Curran Associates, Inc.
  25. Yafu Li, Qintong Li, Leyang Cui, Wei Bi, Zhilin Wang, Longyue Wang, Linyi Yang, Shuming Shi, and Yue Zhang. 2024. MAGE: Machine-generated Text Detection in the Wild. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 36–53, Bangk…
  26. Zachary C. Lipton. 2018. The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery. Queue, 16(3):31–57. find
  27. Ning Lu, Shengcai Liu, Ruidan He, and Ke Tang. 2023. Large language models can be guided to evade ai-generated text detection. find
  28. Trans. Mach. Learn. Res., 2024. find
  29. Yongqiang Ma, Jiawei Liu, Fan Yi, Qikai Cheng, Yong Huang, Wei Lu, and Xiaozhong Liu. 2023. Ai vs. human – differentiation analysis of scientific content generation. Preprint, arXiv:2301.10416. — [arXiv:2301.10416](https://arxiv.org/abs/2301.10416) find
  30. Elyas Masrour, Bradley Emi, and Max Spero. 2025. DAMAGE: Detecting Adversarially Modified AI Generated Text. arXiv preprint. ArXiv:2501.03437 [cs]. find
  31. Eric Mitchell, Yoonho Lee, Alexander Khazatsky, Christopher D. Manning, and Chelsea Finn. 2023. Detectgpt: Zero-shot machine-generated text detection using probability curvature. In International Conference on Machine Learning.
  32. OpenAI. 2024. OpenAI o1 System Card. arXiv preprint. ArXiv:2412.16720 [cs].
  33. B Porter and Edouard Machery. 2024. Ai-generated poetry is indistinguishable from human-written poetry and is rated more favorably. Scientific Reports, 14. find
  34. Giovanni Puccetti, Anna Rogers, Chiara Alzetta, Felice Dell’Orletta, and Andrea Esuli. 2024. AI "News" Content Farms Are Easy to Make and Hard to Detect: A Case Study in Italian. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages… find
  35. Vinu Sankar Sadasivan, Aounon Kumar, Sriram Balasubramanian, Wenxiao Wang, and Soheil Feizi. 2024. Can AI-Generated Text be Reliably Detected? arXiv preprint. ArXiv:2303.11156 [cs].
  36. Chantal Shaib, Yanai Elazar, Junyi Jessy Li, and Byron C. Wallace. 2024. Detection and measurement of syntactic templates in generated text. In Conference on Empirical Methods in Natural Language Processing.
  37. Zhouxing Shi, Yihan Wang, Fan Yin, Xiangning Chen, Kai-Wei Chang, and Cho-Jui Hsieh. 2023. Red teaming language model detectors with language models. Transactions of the Association for Computational Linguistics, 12:174–189.
  38. Irene Solaiman, Miles Brundage, Jack Clark, Amanda Askell, Ariel Herbert-Voss, Jeff Wu, Alec Radford, and Jasmine Wang. 2019. Release Strategies and the Social Impacts of Language Models. ArXiv.
  39. Edward Tian and Alexander Cui. 2023. Gptzero: Towards detection of ai-generated text using zero-shot and supervised methods". find
  40. Vivek Kumar Verma, Eve Fleisig, Nicholas Tomlin, and Dan Klein. 2023. Ghostbuster: Detecting text ghostwritten by large language models. In North American Chapter of the Association for Computational Linguistics.
  41. James Wang, Ran Li, Junfeng Yang, and Chengzhi Mao. 2024a. RAFT: Realistic Attacks to Fool Text Detectors. arXiv preprint. ArXiv:2410.03658 [cs]. find
  42. Tianchun Wang, Yuanzhou Chen, Zichuan Liu, Zhanwen Chen, Haifeng Chen, Xiang Zhang, and Wei Cheng. 2024b. Humanizing the Machine: Proxy Attacks to Mislead LLM Detectors. ArXiv. Publisher: arXiv Version Number: 1. find
  43. Yuxia Wang, Rui Xing, Jonibek Mansurov, Giovanni Puccetti, Zhuohan Xie, Minh Ngoc Ta, Jiahui Geng, Jinyan Su, Mervat Abassy, Saad El Dine Ahmed, Kareem Elozeiri, Nurkhan Laiyk, Maiya Goloburda, Tarek Mahmoud, Raj Vardhan Tomar, Alexander Aziz, Ryuto Koike, Masahiro Kaneko, Artem Shelmanov, Ekaterina… — Is Human-Like Text Liked by Humans? Multilingual Human Detection and Preference Against AI
  44. Alex Warstadt, Alicia Parrish, Haokun Liu, Anhad Mohananey, Wei Peng, Sheng-Fu Wang, and Samuel R. Bowman. 2020. BLiMP: The benchmark of linguistic minimal pairs for English. Transactions of the Association for Computational Linguistics, 8:377– 392.
  45. Qihui Zhang, Chujie Gao, Dongping Chen, Yue Huang, Yixin Huang, Zhenyang Sun, Shilin Zhang, Weiye Li, Zhengyan Fu, Yao Wan, and Lichao Sun. 2024. Llm-as-a-coauthor: Can mixed human-written and machine-generated text be detected? In NAACL-HLT. find
  46. Tianyang Zhong, Zhengliang Liu, Yi Pan, Yutong Zhang, Yifan Zhou, Shizhe Liang, Zihao Wu, Yanjun Lyu, Peng Shu, Xiaowei Yu, Chao Cao, Hanqi Jiang, Hanxu Chen, Yiwei Li, Junhao Chen, Huawen Hu, Yihen Liu, Huaqin Zhao, Shaochen Xu, Haixing Dai, Lin Zhao, Ruidong Zhang, Wei Zhao, Zhenyuan Yang, Jingyua… — Evaluation of OpenAI o1: Opportunities and Challenges of AGI
  47. Ying Zhou, Ben He, and Le Sun. 2024. Humanizing machine-generated content: Evading ai-text detection through adversarial attack. In International Conference on Language Resources and Evaluation. find
  48. Tiffany Zhu, Kexun Zhang, and William Yang Wang. 2024. Embracing AI in Education: Understanding the Surge in Large Language Model Use by Secondary Students. arXiv preprint. ArXiv:2411.18708 [cs] version: 1. — [arXiv:2411.18708](https://arxiv.org/abs/2411.18708) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/russell-2025-people-who-frequently-use-chatgpt/