Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models
Open accessThe DOI matches its Crossref record; 157 of the 158 references it cites were found in a registry or an index, and 1 are not checked yet; 29 of the 52 sources its claims cite were checked for support.
Abstract
Abstract Do large language models (LLMs) know the law? LLMs are increasingly being used to augment legal practice, education, and research, yet their revolutionary potential is threatened by the presence of “hallucinations”—textual output that is not consistent with legal facts. We present the first systematic evidence of these hallucinations in public-facing LLMs, documenting trends across jurisdictions, courts, time periods, and cases. Using OpenAI’s ChatGPT 4 and other public models, we show that LLMs hallucinate at least 58% of the time, struggle to predict their own hallucinations, and often uncritically accept users’ incorrect legal assumptions. We conclude by cautioning against the rapid and unsupervised integration of popular LLMs into legal tasks, and we develop a typology of legal hallucinations to guide future research in this area.
Verified 2026-11-19 · 1 reference not checked yet
Overview
CoChat Summary
TL;DR
This paper provides the first systematic evidence that public LLMs frequently hallucinate on verifiable questions about U.S. federal case law, with hallucination rates ranging from 58% for ChatGPT 4 to 88% for Llama 2. It also documents correlates of hallucination and two additional failure modes: susceptibility to false legal premises and poor calibration about their own accuracy.
Key findings
- Hallucinations are widespread across all tested LLMs on direct, verifiable federal case questions: ChatGPT 4 hallucinates 58% of the time, while Llama 2 hallucinates 88%; GPT 4 performs best overall, followed by GPT 3.5, PaLM 2, and Llama 2.
- Hallucination rates increase with task complexity and vary by court hierarchy, jurisdiction, case prominence, and case year. For reference-free high-complexity tasks, the reported rates are lower bounds: at least 49% for factual background and procedural posture, at least 59% for core legal question, and at least 63% for central holding; doctrinal agreement is near 0.5, little better than random guessing.
- LLMs often accept false legal premises: on the false dissent premise, GPT 4 hallucinates 0.691 at SCOTUS and 0.842 at USCOA; on the false overruling premise, GPT 4 hallucinates 0.531, GPT 3.5 0.821, PaLM 2 1.000, and Llama 2 0.027. Llama 2’s low contra-factual rate reflects its tendency to reject premises, which harms its performance on other tasks such as Existence.
- Calibration is poor and systematically overconfident: pooled expected calibration error is 0.057 for PaLM 2, 0.099 for GPT 3.5, 0.190 for GPT 4, and 0.421 for Llama 2. Although GPT 4 has the lowest raw hallucination rate, it is less calibrated than PaLM 2 and GPT 3.5.
Methods
- The authors develop 14 legal knowledge QA tasks grouped into low, moderate, and high complexity. The first nine are reference-based and scored against metadata from the Caselaw Access Project, Supreme Court Database, Appeals Courts Database Project, Library of Congress, and Shepard’s Citations; the final five are reference-free and detected via self-contradiction between repeated stochastic responses, using GPT 4 for contradiction labeling.
- They draw 5,000 cases each from SCOTUS, USCOA, and USDC using stratified random sampling, then query four off-the-shelf LLMs—OpenAI ChatGPT 4, OpenAI ChatGPT 3.5, Google PaLM 2, and Meta Llama 2—under zero-shot and three-shot prompting. They execute more than 800,000 queries, over 200,000 per LLM, generally at temperature 0 for reference-based tasks.
- Analyses estimate population-level hallucination rates, heterogeneity across task, court, jurisdiction, prominence, and year, plus contra-factual bias and expected calibration error.
Limitations
- The paper states that reference-free hallucination rates are only lower bounds because self-contradiction guarantees hallucination but logically consonant answers may still be false; it also cannot capture repeated incorrect answers as hallucinations, and notes that the SCOTUS Subsequent history estimate likely understates hallucination by a larger margin.
- The paper states that it relies on GPT 4 for contradiction detection, though it validates this with human intercoder reliability (Cohen’s κ: Coder 1 vs. GPT 4 = 0.87; Coder 2 vs. GPT 4 = 0.77; Coder 1 vs. Coder 2 = 0.79). It also notes that closed-source models do not expose conditional probabilities, so calibration is estimated by sampling at temperature 1, and that ex post temperature scaling improves but does not fully solve calibration.
- Evident from the text: the study tests specific public model versions available at the time, so results may not generalize to newer LLMs or to retrieval-augmented, fine-tuned, or proprietary legal systems; the tasks are limited to published federal case law and to metadata available for constructing ground truth.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
16 Valid · 5 Not valid · 9 Not checked · 105 Own results and interpretations
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“since hallucinations of some kind are generally inevitable at the token level (Kalai and Vempala 2023; Xu, Jain, and Kankanhalli 2024)”
What the sources say
- Disputedcites Kalai and Vempala 2023 matched to “Calibrated Language Models Must Hallucinate” (2023)Experimental — accuracy not yet measured
The Introduction (p. 1) explicitly says the paper's lower bounds for calibrated predictors of facts should not be taken to mean hallucination is inevitable. The discussion on p. 2 distinguishes its semantic-level calibration from token-level calibration, so it does not establish inevitability at the token level. (LLM, conf=0.87)
Checked by a reasoning model — may be wrong.
- Partly supportedcites Xu, Jain, and Kankanhalli 2024 matched to “Hallucination is Inevitable: An Innate Limitation of Large Language Models” (2024) sourceExperimental — accuracy not yet measured
Section 3.2 (p. 7) concludes that LLM hallucination cannot be eliminated, supporting the inevitability claim. But Section 2.1 (p. 2) describes token-by-token generation, while the paper defines hallucination by comparing complete output strings with ground truth; it does not establish inevitability specifically at the token level. (LLM, conf=0.85)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
Every source the paper cites for this claim was checked and supports it.
In the paper
“as Kalai and Vempala (2023) show, LMs that assign a positive probability to every response token must hallucinate at least some of the time”
What the sources say
- Supportedcites Kalai and Vempala (2023) matched to “Calibrated Language Models Must Hallucinate” (2023)
The Introduction (p. 1) explicitly says that an LM predicting every string with positive probability will necessarily hallucinate with positive probability. This supports the claim's paraphrase that such hallucination occurs at least some of the time. (LLM, conf=0.93)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Xu, Jain, and Kankanhalli (2024) agree, arguing that 'hallucination is inevitable for any computable LLM, regardless of model architecture, learning algorithms, prompting techniques, or training data.'”
What the sources say
- Supportedcites Xu, Jain, and Kankanhalli (2024) matched to “Hallucination is Inevitable: An Innate Limitation of Large Language Models” (2024) source
The Introduction (p. 2) states the attributed claim nearly verbatim. It frames the result under the paper’s formal definition of hallucination and says it applies to real-world LLMs. (LLM, conf=0.95)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“increased reliance on LLMs also has the potential to produce a kind of legal monoculture (Creel and Hellman 2022; Kleinberg and Raghavan 2021), with users being fed information from only a limited subset of judicial sources that elide many of the deeper nuances of the law.”
What the sources say
- Partly supportedcites Creel and Hellman 2022 matched to “The Algorithmic Leviathan: Arbitrariness, Fairness, and Opportunity in Algorithmic Decision-Making Systems” (2022) sourceExperimental — accuracy not yet measured
Section 3.a, “Why arbitrariness at scale arises” (page unavailable), says that reliance on similar datasets can standardize different algorithms and produce uniform outcomes. That supports the general concern about algorithmic monoculture, but the paper does not address LLMs feeding users a limited subset of judicial sources or eliding nuances of law. (LLM, conf=0.84)
- Partly supportedcites Kleinberg and Raghavan 2021 matched to “Algorithmic monoculture and social welfare” (2021) sourceExperimental — accuracy not yet measured
The Introduction (p. 1) describes algorithmic curation making choices more homogeneous, and the Conclusion (p. 12) says shared evaluations can cause valuable options to be overlooked. Those passages support the general possibility of monoculture, but the paper examines algorithmic screening, not LLMs or whether users receive a limited subset of judicial sources that omits legal nuance. (LLM, conf=0.87)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: unclear — It is unclear whether it could be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“the problem of legal hallucination has so far only been studied in closed-domain applications, such as when a model is used to summarize the content of a given judicial opinion (Deroy, Ghosh, and Ghosh 2023; Feijo and Moreira 2023) or to synthesize provided legal text (Savelka et al. 2023)”
What the sources say
- Partly supportedcites Deroy, Ghosh, and Ghosh 2023 matched to “How Ready are Pre-trained Abstractive Models and LLMs for Legal Case Judgement Summarization?” (2023) sourceExperimental — accuracy not yet measured
The Abstract describes a study of hallucinations in summaries generated from Indian court case judgments, supporting this paper as an example of a closed-domain legal application. The Introduction’s “Novelty of this work” section says that, to the authors’ knowledge, hallucination had not previously been studied in legal summarization. Neither passage establishes the broader claim that legal hallucination had only been studied in closed-domain applications. (LLM, conf=0.84)
- Not checked — no full text of the sourcecites Feijo and Moreira 2023 matched to “Improving Abstractive Summarization of Legal Rulings through Textual Entailment” (2023) source
- Partly supportedcites Savelka et al. 2023 matched to “Explaining Legal Concepts with Augmented Large Language Models (GPT-4)” (2023) sourceExperimental — accuracy not yet measured
The Proposed Method (pp. 6–7) describes an augmented setup that synthesizes retrieved case-law sentences into an explanation, supporting the example attributed to Savelka et al. But it also studies a baseline that generates explanations without those retrieved sentences, and the Conclusions (p. 13) report hallucinated case-law citations in that baseline. The paper therefore does not support the claim that legal hallucination had only been studied in closed-domain applications. (LLM, conf=0.87)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
60 more claims · 3 with their cited sources checked
Create a free account to see all 135 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 42
- It may state two separate claims as one. 44
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In high-stake domains such as legal consultation or scientific research, these hallucinations can lead to severe misjudgments and erode user trust [190,191].”
From AI Agents vs. Agentic AI: A Conceptual taxonomy, applications and challenges (2026) · cites this paper as [190]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Recent research has shown that leading language models – ChatGPT, Deepseek, Grok, PaLM, Claude, Qwen, and Llama – were on average producing hallucinations in 75 % of the cases (Dahl et al., 2024), with many users unaware of this prevalence (Williamson & Prybutok, 2024).”
From If ChatGPT can do it, where is my creativity? generative AI boosts performance but diminishes experience in creative writing (2025) · cites this paper as Dahl et al., 2024
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Dahl et al. [7] showed that even models with high bar exam scores systematically fabricate case references when applied to real legal tasks.”
From Citation Grounding: Detecting and Reducing LLM Citation Hallucinations via Legal Citation Graphs (2026) · cites this paper as [7]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“a confident legal summary can make uncertain advice feel authoritative [93]”
From Caring Without Feeling: Affective Dynamics as the Control Layer of Human-AI Agent Collaboration (2026) · cites this paper as [93]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Previous work has found that general- purpose LLMs hallucinate on legal queries on average between 58% and 82% of the time (Dahl et al. 2024).”
From Hallucination‐Free? Assessing the Reliability of Leading AI Legal Research Tools (2025) · cites this paper as Dahl et al. 2024
Experimental — accuracy not yet measured
10 more claims about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ E.1 Fake Case Existence Task · p. 61
“GPT 4 and Llama 2, on the other hand, appear immune from this behavior, but recall from Table 2 that this is just bias in the opposite direction: they simply prefer to deny the existence of any case, real or fake.”
§ 6 Discussion · p. 38
“| Existence | Zero-shot | 0.204 (0.006) | 0.004 (0.001) | 0.054 (0.003) | 0.303 (0.006)”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 5.1.2 Hallucinations Vary by Court · p. 24
“We find that across all tasks and all LLMs, hallucinations are lowest in the highest levels of the judiciary, and vice-versa (Figure 3).”
§ 6 Discussion · p. 38
“| Citation | Few-shot | 0.689 (0.007) | 0.641 (0.007) | 0.833 (0.005) | 0.953 (0.003) | 0.468 (0.007)”
Citations
Cites
The Measurement of Observer Agreement for Categorical Data
Landis, J. R., & Koch, G. G. (1977) · Biometrics · doi:10.2307/2529310 · Cited by 81,148
A Coefficient of Agreement for Nominal Scales
Cohen, J. (1960) · Educational and Psychological Measurement · doi:10.1177/001316446002000104 · Cited by 42,398
Tversky, A., & Kahneman, D. (1974) · Science · doi:10.1126/science.185.4157.1124 · Cited by 28,810 · cited 1×
“…(the tendency of an LLM to agree with a user’s preferences or beliefs, even when the LLM would reject the belief as wrong without the user’s prompting; Sharma et al., 2023; Wei et al., 2023) or general cognitive error (Tversky and Kahneman, 1974; Jones and Steinhardt, 2022; Suri et al., 2023).”
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Kaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,690
Survey of Hallucination in Natural Language Generation
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023) · ACM Computing Surveys · doi:10.1145/3571730 · Cited by 5,025
On calibration of modern neural networks
Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017) · arXiv · doi:10.48550/arxiv.1706.04599 · Cited by 3,435
Barocas, S., & Selbst, A. D. (2014) · doi:10.2139/ssrn.2477899 · Cited by 3,048
On the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676
Self-citationLlama 2: Open Foundation and Fine-Tuned Chat Models
Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.09288 · Cited by 2,604
Research works printed but not linked: 10
- Ambrogi, Bob. 2023. 'As Allen & Overy Deploys GPT-based Legal App Harvey Firmwide, Founders Say Other Firms Will Soon Follow.' LawSites.
- Chien, Colleen V., Miriam Kim, Raj Akhil, and Rohit Rathish. 2024. 'How Generative AI Can Help Address the Access to Justice Gap Through the Courts.' Loyola of Los Angeles Law Review, forthcoming.
- Draper, Chris, and Nicky Gillibrand. 2023. 'The Potential for Jurisdictional Challenges to AI or LLM Training Datasets.' In Proceedings of the ICAIL 2023 Workshop on Artificial Intelligence for Access to Justice. Braga, Portugal: CEUR Workshop Proceedings.
- Dworkin, Ronald. 1986. Law's Empire. Cambridge, MA: Harvard University Press.
- Luft, Joseph, and Harrington Ingham. 1955. 'The Johari Window as a Graphic Model of Interpersonal Awareness.' In Proceedings of the Western Training Laboratory in Group Development. University of California, Los Angeles, Extension Office.
- Mayson, Sandra G. 2019. 'Bias In, Bias Out.' The Yale Law Journal 128 (8): 2122-2473.
- Romoser, James. 2023. 'No, Ruth Bader Ginsburg Did Not Dissent in Obergefell - and Other Things ChatGPT Gets Wrong about the Supreme Court.' SCOTUSblog.
- Simshaw, Drew. 2022. 'Access to A.I. Justice: Avoiding an Inequitable Two-Tiered System of Legal Services.' Yale Journal of Law & Technology 24:150-226.
- Tan, Jinzhe, Hannes Westermann, and Karim Benyekhlef. 2023. 'ChatGPT as an Artificial Lawyer?' In Proceedings of the ICAIL 2023 Workshop on Artificial Intelligence for Access to Justice. Braga, Portugal: CEUR Workshop Proceedings.
- Weiser, Benjamin. 2023. 'Here's What Happens When Your Lawyer Uses ChatGPT.' The New York Times.
Cited by
Items 101–200 of 306
Jeannot, G. (2025) · Revue française d administration publique · doi:10.3917/rfap.186.0217 · Cited by 1
Data, Artificial Intelligence, and Platforms as a Triple Challenge for Economic Regulation
Denga, M., & Hornuf, L. (2025) · Advanced studies in diginomics and digitalization · doi:10.1007/978-3-031-81089-3_1 · Cited by 4
Kim, M. S., Lee, J. S., & Bae, H. (2025) · Healthcare Informatics Research · doi:10.4258/hir.2025.31.2.200 · Cited by 4
Large language models to write scientific manuscripts: to be considered but not trusted
Basile, C., Anker, S. D., & Savarese, G. (2025) · Global Cardiology · doi:10.4081/cardio.2025.74 · Cited by 1
AdmPModeler: Modeling Administrative Processes Using Large Language Models. A Case Study
Macrì, M., Luzi, F. D., & Mecella, M. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-01589-1_19 · Cited by 1
Neuwirth, R. J. (2025) · De Gruyter eBooks · doi:10.1515/9783111346632-007 · Cited by 1
Redefining legal access: a RAG-based AI system for Indian law
Mubeen, F., Mehdi, A., Haque, M. A., Nomani, M. Z. M., & Uddin, N. S. (2025) · Human-Intelligent Systems Integration · doi:10.1007/s42454-025-00086-8 · Cited by 1
“Based on a comprehensive review of several popular large language models (LLMs) used in legal work (Dahl et al. 2024), the study found that using these tools without proper oversight could be dangerous.”
Šekrst, K. (2025) · Springer eBooks · doi:10.1007/978-3-032-05562-0_12 · Cited by 3
Emotion in Context: Human and AI Perspectives on the Affective Landscape of Severance
Ortiz, A. M., & García-Gámez, M. (2025) · Corpus Pragmatics · doi:10.1007/s41701-025-00220-6 · Cited by 8
“For example, in legal contexts, LLMs have been shown to hallucinate at least 58% of the time when answering case law questions (Dahl et al., 2024).”
OSIRIS: A Domain-Guided Framework for Tax Foreclosure Automation Using Large Language Models
Braz, F. A., da Silva, N. C., Nishi, L., Siqueira, E., Bonat, D., Corrêa, C. M. B., & Melo, I. (2025) · International Conference on Artificial Intelligence and Law · doi:10.1145/3769126.3769225 · Cited by 1
“Stanford researchers reported hallucination rates of 69–88% for legal queries, underscoring the reliability concerns [5].”
Evaluating Large Language Models as Judicial Decision-Makers
Gazal‐Ayal, O., Elyoseph, Z., & Solomon, A. (2026) · Justice Quarterly · doi:10.1080/07418825.2026.2618254 · Cited by 6
“API-based evaluation is widely recognized as a best practice for ensuring scalability and reproducibility in LLM research across diverse domains (Bang et al., 2023; Chakraborty et al., 2025; Rawashdeh et al., 2024), including legal analysis (Dahl et al., 2024).”
It Doesn’t Mean a Thing (If It Ain’t a Living Thing)
Rolla, G., & Figueiredo, N. M. (2026) · Journal of Artificial Intelligence and Consciousness · doi:10.1142/s2705078526400047 · Cited by 1
Chuan, A., & Nanda, M. (2026) · Anesthesiology · doi:10.1097/aln.0000000000005933 · Cited by 1
Han, J., Burgess, P., & Shareghi, E. (2026) · Artificial Intelligence and Law · doi:10.1007/s10506-026-09506-9 · Cited by 12
Operationalizing Responsible AI Policies with LLMs: an End-to-End Monitoring Prototype
Zieliński, P., Kelly, M., Barreto, R. D. C., & Tahaei, M. (2026) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3772363.3799675 · Cited by 4
“The need for expert human judgment is further motivated by the unreliability of LLMs in legal and policy settings; prior work shows that these models can hallucinate or mis-ground legal content [9, 16, 31], reinforcing the need for human oversight.”
Human Evaluation of Large Language Models: A Review and Protocol Selection Framework
Brunyé, T. T. (2026) · AI · doi:10.3390/ai7050174 · Cited by 4
A Purpose-Aware Semantic Reasoning Model for Patent Infringement Detection in the DIKWP Network
Guo, Z., & Duan, Y. (2026) · Electronics · doi:10.3390/electronics15112449 · Cited by 1
ONE-CLICK AWAY LEGAL PRACTICE: A Perspective on Modern Legal Practice with AI in the Philippines
Acas, R. T., & Alejo-Abitago, J. (2026) · International Journal of Digital Law and Governance · doi:10.1515/ijdlg-2026-0006 · Cited by 1
“On directly formulated and factually grounded legal inquiries, general-purpose LLMs have also been found to have a number of hallucination issues, ranging from 58 % to 88 % (Dahl et al. 2024; Chang 2026).”
Sobral, L., Muniz, B., Gatto, C., Granja, J. L., Teixeira, T., Mirarchi, C., Pavan, A., Pereira, J. F., Espinho, A., Pires, C. M., et al. (2026) · Building Research & Information · doi:10.1080/09613218.2026.2690114 · Cited by 2
Chen, C. K., Liang, S.-H., & Wang, T. (2026) · Scientometrics · doi:10.1007/s11192-026-05727-6 · Cited by 1
“Dahl et al. (2024) profiled legal hallucinations, finding at least 58% fabrication rates in legal citations.”
White, K., Detert, R., Hornik, S., Guerra-Pujol, F. E., & Rosenthal, L. (2026) · Journal of Accounting Education · doi:10.1016/j.jaccedu.2026.101047 · Cited by 2
Peykani, P., Charles, V., Emrouznejad, A., Ghanidel, S., & Sabour, D. (2026) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-026-10720-6 · Cited by 1
HyperCausal: Visualizing Causal Inference in 3D Hypertext
Bönisch, K., Stoeckel, M., & Mehler, A. (2024) · ACM Conference on Hypertext and Social Media (HT) · doi:10.1145/3648188.3677049 · Cited by 1
URAG: Unified Retrieval-Augmented Generation
Song, Y., Yan, L., Qin, L., Wang, G., Huang, X., Hu, L., & Liu, W. (2024) · International Conference on Communication and Information Processing (ICCIP) · doi:10.1145/3708657.3708763 · Cited by 2
“For example, a recent study showed that state-of-the-art large language models (LLMs) had a hallucination rate of 69% to 88% when dealing with specific legal queries[9].”
Zhang, Y. (2025) · Applied and Computational Engineering · doi:10.54254/2755-2721/2025.tj24496 · Cited by 5
Unveiling Knowledge Boundary of Large Language Models for Trustworthy Information Access
Deng, Y., Li, M., Pang, L., Zhang, W., & Lam, W. P. (2025) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3726302.3731684 · Cited by 5
Parameterized Argumentation-based Reasoning Tasks for Benchmarking Generative Language Models
Steging, C., Renooij, S., & Verheij, B. (2025) · International Conference on Artificial Intelligence and Law · doi:10.1145/3769126.3769230 · Cited by 5
The Rule of Law or the Rule of Robots? Nationally Representative Survey Evidence from KenyaPreprint
Flanagan, B., Almeida, G., Chen, D., et al. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4865782 · Cited by 1
Artificial Intelligence and Analyst ProductivityPreprint
Shanthikumar, D. M., & Yoo, I. S. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5040339 · Cited by 4
Hagan, M. D. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5146722 · Cited by 3
Burgess, P., & Shareghi, E. (2025) · Oxford University Press eBooks · doi:10.1093/9780198945215.003.0093 · Cited by 1
Rodgers, I. T., & Sako, M. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5182132 · Cited by 3
Free LLMs Hallucinate and Rarely Signal Their Limitations in Solving Legal ProblemsPreprint
anon, A., & Figura, J. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5188654 · Cited by 1
Generated Legal TextsPreprint
Wilf-Townsend, D., & Tobia, K. P. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5243382 · Cited by 1
DeepResearch Eco : A Recursive Agentic Workflow for Complex Scientific Question Answering in EcologyPreprint
D’Souza, J., Sander, E. K., & Aioanei, A. C. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.07.14.664755 · Cited by 1
Training for Technology: Adoption and Productive Use of Generative AI in Legal AnalysisPreprint
Chen, B. M., & Bao, H. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6217459 · Cited by 1
Tinker, Tailor, Trust: How Developers Create Privacy Policies With and Without AI
Mayahi, S., Alomar, N., & Malkin, N. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791323
“This raises additional concerns about the documents’ accuracy, since LLMs are well known for their hallucinations [29, 64, 82, 86].”
Bauer, A. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7279079 · Cited by 4
Bridging risk and retention: risk coping with privacy and hallucination in AI
Shin, D., Rasul, A., & Shahzad, K. (2026) · AI & SOCIETY · doi:10.1007/s00146-026-03127-1
“The frequency of such errors is substantial, ranging from 20 to 60% depending on the domain of knowledge (Dahl et al. 2024).”
The Effect of Race, Gender and Priming on LLMs' Conviction PredictionsPreprint
Schwartz, T., & Wang, C. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4779332
Han, Z., Battaglia, F., Mansuria, K., Heyman, Y., & Terlecky, S. R. (2024) · Research Square · doi:10.21203/rs.3.rs-5084169/v1
Morhat, P. M. (2024) · Юридические исследования · doi:10.25136/2409-7136.2024.8.71576
Professor GPT: Having a Large Language Model Write a Commentary on Freedom of AssemblyPreprint
Engel, C., & Kruse, J. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4994131
Children, Rights and Algorithms in the Age of Generative AI
Peinado, R. G. (2024) · The European Educational Researcher · doi:10.31757/euer.821 · Cited by 2
“Dahl et al. (2024) warn about the risks of implementing immature technologies in critical environments, recommending prior impact assessments.”
Traditional and Computational CanonsPreprint
Martínez, E. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5155444 · Cited by 5
Acceptable Use Policies for Foundation Models
Klyman, K. (2024) · SuperIntelligence - Robotics - Safety & Alignment · doi:10.70777/si.v1i1.10917
Worker Rights in Collective BargainingPreprint
Arold, B. W., Ash, E., MacLeod, W. B., & Naidu, S. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5190792
The Use of Artificial Intelligence and the Professional Duties of German Lawyers
Rennig, C. (2025) · Studia Prawa Publicznego · doi:10.14746/spp.2025.1.49.3
Worker Rights in Collective BargainingPreprint
Arold, B. W., Ash, E., MacLeod, W. B., & Naidu, S. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5201664
AI and the Future of Disputing: Naming, Blaming, Claiming, and Preventing
Katsh, E., Schwartz‐Maor, T., & Rabinovich‐Einy, O. (2025) · Conflict Resolution Quarterly · doi:10.1002/crq.21477 · Cited by 19
A Bayesian Exploration on the Motivational and Behavioral Impacts of Chatbots in Language Learning
Carlon, M. K. J., Matthews, J., & Kuniyoshi, Y. (2025) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3706599.3720088 · Cited by 1
“Known issues include the generation of “hallucinations” and difficulties in understanding domain-specific contexts, which can undermine trust in AI [10, 19].”
A Method for Detecting Knowledge Conflicts in Chinese Intelligent Agent Interactions
Cheng, H., Gu, Y., Xi, M., Zhong, Q., & Liu, W. (2025) · Interdisciplinary Journal of Information Knowledge and Management · doi:10.28945/5497 · Cited by 1
“They often suffer from hallucination issues – generating content that deviates from factual information – and cannot perceive or reflect real - time updates due to the static nature of their training data (Dahl et al., 2024; Huang et al., 2024; R. Xu, Qi, et al., 2024).”
Global legal skills as a universal language of lawyers
Creciun, N. (2025) · Integrare prin cercetare și inovare. Științe juridice și economice · doi:10.59295/spd2024j.30
From Statement of Facts to Statutory Provisions - Efficient Retrieval of Relevant Legislation
Smywiński-Pohl, A., Król, M., & Mirosław, P. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-97570-7_20 · Cited by 2
Droit et intelligence artificielle : interactions et transformations
Meneceur, Y. (2025) · Annales des Mines - Réalités industrielles · doi:10.3917/rindu1.252.0034
The Ordinary MKeaning Bot: Simulating Human Surveys with LLMsPreprint
Kruse, J. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5378203
Relying on AI in Judicial Decision-Making: Justice or Jeopardy?Preprint
Sourav, M. R. I. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5387247 · Cited by 1
LLM as a law professor: having a large language model write a commentary on freedom of assembly
Engel, C., & Kruse, J. (2026) · Artificial Intelligence and Law · doi:10.1007/s10506-026-09527-4 · Cited by 2
Feng, T. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5498885
Yu, G., Yan, L., Niu, J., Gao, D., & Yuan, X. (2025) · IEEE Advanced Information Technology, Electronic and Automation Control Conference (IAEAC) · doi:10.1109/iaeac65194.2025.11166604 · Cited by 1
“For example, research shows that when answering specific legal queries, the rate of fabricated content by LLMs ranges from 69% to 88% [21].”
Large Language Models as Search Engines: Societal Challenges
Sadeddine, Z., Maxwell, W., Varoquaux, G. P., & Suchanek, F. M. (2025) · ACM SIGIR Forum · doi:10.1145/3769733.3769746 · Cited by 6
“Depending on what the user makes of this answer, the LLM can have an important impact: for instance, it may give a misdiagnosis for an illness, give an inaccurate legal advice [Dahl et al., 2024], or give a wrong or dangerous advice for handling a problem that the user encounters.”
Von GovTech zu Legal Tech: Was Start-ups im Rechtsbereich wissen sollten
Dötterl, S. (2025) · Edition innovative Verwaltung · doi:10.1007/978-3-658-48647-1_4
Law, Justice, and Artificial IntelligencePreprint
Katz, O., & Zamir, E. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5585290 · Cited by 1
Application of artificial intelligence technologies in criminal proceedings: realities and prospects
Serhii, K. (2025) · Aktual’ni problemi pravoznavstva · doi:10.35774/app2025.03.117 · Cited by 2
Towards Trustworthy Legal AI: A Multi-Agent Approach to Integrating Legislative Knowledge
Matak, M., & Chudziak, J. A. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-09318-9_1
Liu, X., Yun, J., Li, B., Zhang, Y., & Shi, X. (2025) · Communications in computer and information science · doi:10.1007/978-981-95-4088-4_4
Jia, H., Feng, D.-Y., Chen, H.-G., Jiang, S., & Wang, H. (2025) · IEEE International Joint Conference on Neural Network · doi:10.1109/ijcnn64981.2025.11229389
“Similarly, for GAR2 , our further examination reveals that in addition to the reliance on legal knowledge, as seen with GAR1 , hallucinations [20]–[22] also play a detrimental role.”
N-CATS: A Neural Network Classification Automatic Taxonomy System
Carthen, C. D., Lac, L., White, R., Braun, T., Ochoa, J., Tavakkoli, A., & Harris, F. C. (2025) · ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing (SNPD) · doi:10.1109/snpd65828.2025.11254174 · Cited by 1
“These RAG methods are created to expand an LLM’s knowledge beyond its training knowledge and reduce hallucination [13].”
Human-in-the-loop semantic middleware for construction compliance checking and safe product reuse
Adu-Duodu, K., Wilson, S., Li, Y., et al. (2026) · Advanced Engineering Informatics · doi:10.1016/j.aei.2026.105112
A Resilient, Multi-Layered RAG Framework for High-Fidelity Retrieval in Complex Technical Manuals
Huang, T., Zhu, X., Liu, K., & Zhao, P. (2025) · International Conference on Electronic Information Engineering and Computer Science (EIECS) · doi:10.1109/eiecs67708.2025.11283308
“These failures significantly increase the risk of harmful hallucinations [4] , a catastrophic outcome in domains where precision is paramount.”
Debate-Driven Legal Reasoning: Disambiguating Confusing Charges Through Multi-agent Debate
Li, M., Chao, W., Zhou, X., & Luo, Z. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-3343-5_27
Retrieval Augmented Generation-based Large Language Model Chatbot
Yau, L. H., Xian, E. T. Y., Yu, Y. P., & Lim, T. M. (2025) · IEEE International Conference on Computation, Big-Data and Engineering (ICCBE) · doi:10.1109/iccbe65177.2025.11255894
“Despite their capabilities, LLMs show "hallucinations" by generating outputs that are nonsensical, inaccurate, or unfaithful to the input data [16].”
Chanrueang, S., Chormuan, S., Nakpomchin, S., & Sookplung, P. (2025) · International Computer Science and Engineering Conference · doi:10.1109/icsec67360.2025.11298019
“Phenomena such as hallucinations—where models generate plausible but incorrect information—limit adoption in domains that demand rigor and accountability [2].”
Salvo, M. D. (2026) · Digital Law Journal · doi:10.38044/2686-9136-2025-6-2-8-27
What Is the Law? A System for Statutory Research (STARA) with Large Language Models
Surani, F., Gailmard, L., Casasola, A., Magesh, V., Robitschek, E. J., & Ho, D. E. (2025) · International Conference on Artificial Intelligence and Law · doi:10.1145/3769126.3769219 · Cited by 3
Self-citationA Demonstration of a Semi-Structured Legal Reasoning Framework
Westermann, H. (2025) · International Conference on Artificial Intelligence and Law · doi:10.1145/3769126.3769196
“Several studies have further highlighted issues of e.g. hallucination when using LLMs to perform legal tasks [1, 2, 7].”
Cox, C. (2025) · International Conference on Artificial Intelligence and Law · doi:10.1145/3769126.3769194 · Cited by 1
“[19, 38] But these limitations are present in many types of data, and so while they may call into question the feasibility or ethics of particular legal AI applications, such limitations do not generally call into question the basic picture of how opinions are to be understood and used as data.”
Kabakuş, A. T. (2026) · Gazi University Journal of Science Part A Engineering and Innovation · doi:10.54287/gujsa.1819131
Large language models in interpreting legal documents
Corbo, S. (2026) · Elsevier eBooks · doi:10.1016/b978-0-443-33592-1.00004-8 · Cited by 1
Ram, A. J., & van Zijl, W. (2026) · South African Journal of Economic and Management Sciences · doi:10.4102/sajems.v29i1.6348
“This can be linked to current concerns about AI models’ hallucinations, reducing their value for tasks where correct answers, without any creativity on the AI model’s part, are required (Dahl et al. 2024; Roberts et al. 2024).”
Implementation Challenges of AI in Judicial Decision-Making in Jordan
Alqatawna, A., Czékmann, Z., & Nagy, A. (2026) · The Law and Development Review · doi:10.1515/ldr-2025-0076 · Cited by 2
Diaz, M. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6001374
DSAD: A dual-stream aligned detection framework for LLM-generated exam answers
Wang, J., Li, Y., You, T., Liu, L., & Wang, P. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.131602 · Cited by 2
TOWARDS RESPONSIBLE AI IN LEGAL SERVICES:THE CASE FOR REGULATING CHATBOTS
Mazaraki, N., & Кортукова, Т. (2025) · Bulletin of Taras Shevchenko National University of Kyiv Legal Studies · doi:10.17721/1728-2195/2025/2.130-9 · Cited by 1
AI Hallucinations as an Existential Risk to Human Society: Mitigating the Risks
Goltz, N., & Yuxin, Z. (2026) · BENTHAM SCIENCE PUBLISHERS eBooks · doi:10.2174/9798898812256126040007
Reimagining Reality: Innovations in AI Hallucination Management and Trustworthy Generation
Nag, A., & Chaudhuri, D. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-7534-2.ch012 · Cited by 1
Passing the Buck to AI: How Individuals’ Decision-Making Patterns Affect Reliance on AI
Mei, K. X., Pang, R. Y., Lyford, A., Wang, L. L., & Reinecke, K. (2026) · ACM Transactions on Computer-Human Interaction · doi:10.1145/3786326 · Cited by 8
“Despite their extensive capabilities, current conversational AI tools supported by large language models (LLMs) have been found to generate inaccurate and biased information in various domains including politics, law, and medicine [8, 22, 32, 54, 57, 64].”
Large Language Models as Search Engines: Societal Challenges
Sadeddine, Z., Maxwell, W., Varoquaux, G. P., & Suchanek, F. M. (2025) · HAL · doi:10.1145/3769733.376974
McCallion, B. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6242458
Mohamed, A. R., Elmahjub, E., & Shaban, K. B. (2026) · Intelligent Systems with Applications · doi:10.1016/j.iswa.2026.200727 · Cited by 3
The Hegemonic Function of AI: Cognitive Sovereignty in the Age of Algorithmic MediationPreprint
Calloway, D. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6256058
OPENXRD: a comprehensive benchmark framework for LLM/MLLM XRD question answering
Vosoughi, A., Shahnazari, A., Xi, Y., Zhang, Z., Hess, G., Xu, C., & Abdolrahim, N. (2026) · Digital Discovery · doi:10.1039/d5dd00519a · Cited by 5
“In contrast, recent progress in Natural Language Processing (NLP) driven by Large Language Models (LLMs), such as GPT-based architectures, has transformed various computational domains 37–39 .”
Judge AI: A Case-Study of Large Language Models as Judges
Posner, E. A., & Saran, S. (2026) · Journal of law & empirical analysis. · doi:10.1177/2755323x261433614 · Cited by 28
“Dahl et al. (2024), for instance, report that LLMs hallucinate be-tween 58% and 88% of the time when faced with legal queries and tasks.”
Luo, M. M., Zhang, P. M., Xu, W., Su, H., Zhou, X., Yan, H., & Cummings, J. J. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6464746
Lee, H. (2026) · Journal of Korean Academy of Nursing Administration · doi:10.11111/jkana.2025.0082
Improving AI Accuracy: Identifying and Fixing Hallucinations in Large Language Models
Dadoriya, N., & Chandra, S. K. (2025) · International Conference on Data Science and Intelligent Network Computing (ICDSINC) · doi:10.1109/icdsinc66221.2025.11448207
“NLP extensions with LLM models such as GPT[2], LLaMA[3], and BERT[4] sometimes generate false information as output, which proves to be dangerous when used in crucial domains such as healthcare[5], law[6], and finance[7].”
Plastic Computing: The Cloud Continuum Journey beyond InfinityPreprint
Masip, X., Garcia, J. C., Asensio, A., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6510379
Governing Artificial Intelligence in the Evolving Academic Publishing Ecosystem
Karaarslan, E., & Aydın, Ö. (2026) · Journal of Metaverse · doi:10.57019/jmv.1899774 · Cited by 1
Strąk, T., Piwowarski, M., Tuszyński, M., & Nermend, K. (2026) · IFIP advances in information and communication technology · doi:10.1007/978-3-032-18920-2_42
Research on Citizen Privacy Risk Assessment Method Based on Retrieval-Augmented GenerationPreprint
Qu, J., & Luo, F. (2026) · Research Square · doi:10.21203/rs.3.rs-9015963/v1
Metadata
- Authors
- Dahl, M., Magesh, V., Suzgun, M., & Ho, D. E.
- Affiliations
- Yale University (US)
- Stanford University (US)
- Corresponding author
- Dahl, Matthew
- Venue
- Journal of Legal Analysis
- Publisher
- Oxford University Press (OUP)
- Published
- 2024-01-01
- Volume
- 16
- Issue
- 1
- Pages
- 64-93
- ISSN
- 1946-5319 (ISSN-L)
- 2161-7201
- OpenAlex
- W4400047099
- Cited by
- 559
- References
- 158
- Citations by year
- 2026: 355 · 2025: 148 · 2024: 55
- FWCI
- 677.59
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- Locations
- The Journal of Legal Analysis (publishedVersion, cc-by-nc)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Topics
- Social Sciences › Law › Artificial Intelligence in Law
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Law › Digitalization, Law, and Regulation
- Keywords
- large language models, ChatGPT, legal knowledge, legal reasoning, legal education, legal practice
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.