Artificial Intelligence in Federal Courts: A Random-Sample Survey of Judges
Click here to access the full report.
EXECUTIVE SUMMARY
The purpose of this study is to understand how, and to what extent, federal judges and other personnel who work in their chambers use artificial intelligence (AI) tools in their judicial work. We selected a stratified random sample of 502 federal bankruptcy, magistrate, district court, and court of appeals judges from a population of 1,738 current federal judges. Of the 502 judges that we surveyed via email, 112 responded (22.3% response rate).
Although a majority of responding judges at least occasionally use AI tools in their judicial work, relatively few report using AI on a daily or weekly basis. Approximately 38% of judges reported that they did not use AI at all in their work. This pattern suggests that AI is present in federal judicial chambers but not yet a routine, embedded part of most judges’ decision-making processes.
Respondents report more frequent use of legal‑specific AI tools integrated into established research platforms (such as Westlaw’s AIAssisted Research and similar tools) than of stand‑alone, general‑purpose AI tools such as ChatGPT, Copilot, or Gemini. This pattern indicates that vendor familiarity and perceived reliability may strongly shape which AI tools judges are willing to deploy in chambers.
Judges’ attitudes toward AI are almost evenly split between optimism and concern. Many respondents simultaneously recognize AI’s potential efficiency gains and express unease about hallucinations, “zombie cases,” and skill atrophy. When AI training is offered by court administration, most judges attend, but a sizeable majority have not been offered such training or are unsure whether training has been available, suggesting unmet demand for high‑quality, judiciary‑specific education on AI.
KEY FINDINGS
The survey reveals the following main insights:
- AI adoption is broad but infrequent: More than 60% of responding judges reported using at least one AI tool in their judicial work. However, only 22.4% reported using these tools on a weekly or daily basis. Conversely, 38.4% of judges have never used any of the AI tools listed on the survey in their work.
- Preference for legal-specific AI tools: Judges are more likely to use “AI for Law” tools (AI tools specifically designed for legal use cases) than general-purpose AI platforms and use them more frequently.
- Legal research dominates use cases: The most common AI use case for judges by far is conducting legal research, reported by 30.0%. This is followed by reviewing documents at 15.5%. Similarly, the top AI use case for others working in chambers is also legal research, at 39.8%, followed by reviewing documents at 16.7%.
- AI training has not been offered to most judges: 45.5% of judges said that AI training had not been provided by court administration, and 15.7% said that they were not sure. Three out of four judges offered AI training attended the training.
- One in three judges permit AI use in chambers: One in three judges permit or permit and encourage the use of AI by those working in their chambers. Only 20.4% of judges formally prohibit AI use, and 17.6% of judges discourage but do not formally prohibit AI use.
- One in four judges has no official AI use policy: About one in four judges (24.1%) reported that they have no official AI policy. If the 17.6% of judges who discourage but do not formally prohibit AI use are also counted as not having an official policy, that number climbs to 41.7% of judges not having an official AI use policy.
- Personal and professional AI use are correlated: A judge’s use of AI in their personal life is correlated with their use of AI in their judicial work. Overall, 38% of judges use AI daily or weekly outside of work. A majority of judges reported that they rarely (26.9%) or never (25.9%) use AI outside of work.
- Judicial outlook on AI is evenly split: Judges are nearly evenly divided between those who are optimistic about AI’s potential for the judiciary and those who are concerned.