What Federal Judges Think About AI — And What It Means for Your Case
Two major judicial surveys released this spring paint a striking picture of where federal judges stand on artificial intelligence: judges are increasingly using AI tools, but the rules governing that use — and governing AI-generated evidence — vary wildly across courtrooms.
Judges Are Using AI – But Expectations Remain Inconsistent
A survey by Northwestern Pritzker School of Law and The Sedona Conference queried over 500 federal judges and found that more than 60% of responding judges have used AI at least once, and nearly one in three permit AI use within their chambers. The most common use? Legal research — with Westlaw AI leading the pack, followed by ChatGPT.
But adoption is uneven. Only about 22% of judges use AI on a daily or weekly basis, and nearly 40% never use it at work. Almost half received no formal AI training. The takeaway: you cannot assume your judge is AI-fluent — or AI-skeptical.
Judicial policies remain fragmented. Some judges reported permitting limited AI use within chambers, while others described formal restrictions or outright prohibitions, and many indicated they had no formal AI policy at all. The result is a patchwork of judge-specific expectations that can vary significantly from courtroom to courtroom.
The survey also suggests that federal judges are drawing an important distinction between AI as an organizational tool and AI as a substitute for legal judgment. Respondents appeared more comfortable using AI to summarize records, review documents, or assist with research than relying on it to generate substantive legal analysis or make decisions. Several judges described AI as a useful “starting point,” but not something that eliminates the need for independent review.
For litigants, that distinction matters. The practical issue is no longer simply whether AI may be used, but whether lawyers can verify, explain, and defend AI-assisted work product if challenged.
Courts Are Getting Serious About Deepfake Evidence
On the evidence side, courts are grappling with a different AI challenge: fabricated audiovisual evidence. A Federal Judicial Center survey of 931 federal judges found that very few judges — only about 15, or roughly 2% — have encountered a genuine deepfake challenge in litigation. But the number will grow, and courts are already setting the ground rules.
The survey found that most judges would require a party raising a deepfake challenge to make an initial evidentiary showing before they would act. A bare assertion that evidence is AI-generated will not suffice. Courts want proof.
That approach is now influencing proposed changes to Federal Rule of Evidence 901. The Advisory Committee on Evidence Rules is now considering a formal amendment to Federal Rule of Evidence 901 that would codify this standard. The proposed Rule 901(c) would require a challenging party to present evidence sufficient to support a finding of fabrication before a court takes any action on the challenge. The rule puts the burden squarely on the challenger — and raises the bar above suspicion.
Importantly, the surveys suggest that courts are not seeking to create entirely new evidentiary systems for AI-generated evidence. Instead, judges appear inclined to adapt existing authentication and admissibility principles to emerging technologies.
What This Means for Litigants
For litigants appearing in federal court, these surveys carry a clear message:
Know your judge's AI policy. Before submitting AI-assisted work product, check whether your judge has issued standing orders or local rules on AI use and disclosure.
Authenticate digital evidence carefully. If you plan to challenge AI-generated or deepfake evidence, prepare to back it up. Courts will not act on your say-so. Line up expert support and documentary evidence before you raise the issue.
Verify everything AI produces. Judges who use AI still expect lawyers to catch AI errors. The survey authors put it plainly: "trust but verify" is the price of admission for any AI-assisted work product.
Use AI strategically. Judges appear most receptive to AI serving as a litigation-support tool for organization, summaries, and document review rather than as a replacement for legal analysis.
The federal judiciary is not moving toward either blanket acceptance or blanket rejection of AI. Instead, courts appear to be developing a cautious, task-specific approach that permits efficiency-enhancing uses while continuing to insist on human judgment, accountability, and evidentiary reliability.