Studies reveal promise and limits of AI in jury deliberations
Recent studies examine AI’s mixed success in aiding jury deliberations and selection.
Why it matters: Understanding AI’s role in jury processes could reshape courtroom dynamics, legal strategy, and ethical standards in trials.
- UNC Chapel Hill’s 2025 mock trial tested AI jurors using ChatGPT, Claude, and Grok models.
- AI-based jury research platforms provide rapid liability and damages insights to attorneys.
- A 2025 study found GPT-4, Claude, and Gemini models systematically fail to replicate juror reasoning.
- AI-assisted jury selection tools risk perpetuating racial and gender biases, potentially harming fairness.
In fall 2025, the University of North Carolina at Chapel Hill School of Law conducted a mock trial using AI jurors powered by models such as ChatGPT, Claude, and Grok. The experiment aimed to evaluate AI’s capability to participate in jury deliberations based on a real juvenile case, revealing both potential benefits and significant limitations. According to UNC Chapel Hill’s analysis, AI can assist attorneys by rapidly assessing liability, comparative fault, and damages through emerging AI-powered jury research platforms within litigation consulting.
However, a rigorous 2025 study by Hayley Stillwell and Sean A. Harrington tested GPT-4, Claude, and Gemini’s performance in mock trials and found they consistently failed to simulate how human jurors interpret evidence and make judgments. Systematic errors in AI outputs stemmed from hidden prompts, content filters, and demographic flattening which distorted decision-making across different demographic variables. Stillwell and Harrington emphasized that today's large language models fall short of replicating juror reasoning but could potentially assist judges and lawyers if refined with transparent methodologies and real human data.
Furthermore, Alexandria Serra's 2026 research highlights serious concerns about AI in jury selection. AI-assisted tools risk encoding historical racial and gender biases into the jury selection process, driving strike decisions at scale and thereby undermining fairness.
As Kathryn M. Rattigan noted, "An AI simulation may consistently measure individual reactions to a case summary yet still fail to capture how an actual jury reaches a verdict." This underscores the vital role human deliberation continues to play despite AI’s growing presence in legal procedures.
By the numbers:
- Fall 2025 — UNC Chapel Hill’s AI juror mock trial date
- August 11, 2025 — Publication date of Stillwell and Harrington’s AI jury simulation study
- 2026 — Year Alexandria Serra published research on AI and jury selection bias
Yes, but: While AI still struggles to fully replicate juror reasoning, advances using transparent methods and human data may enhance its role as a legal aid rather than a replacement for human judgment.
What's next: Future research and refined AI development are needed to address embedded biases and improve AI’s utility in courtroom applications.