An Old Standard for New Tools: What Judges Expect from AI in Legal Work

Reveal spoke with judges across California’s superior courts and federal courts to understand how they are navigating the rise of AI in legal workflows and evidence. The findings are clear: judges aren’t necessarily resistant to technology, but they demand defensibility and competency. The lawyers who embrace AI thoughtfully, validate their tools, and communicate transparently will be best positioned to succeed.

“Technology is here to stay, and the quicker lawyers learn to harness these tools effectively, the better equipped they’ll be.”  — U.S. Magistrate Judge Allison Goddard, U.S. District Court for the Southern District of California

An age-old stereotype holds that courts lag when it comes to technological innovation. Famously, Chief Justice John Roberts asked what the difference was “between email and a pager,” and Justice Antonin Scalia wondered whether text messages from a workplace affair case could simply be printed and distributed. To their credit, that was 2010, and the judicial system itself favors deliberation, reactivity, and caution by design.

These principles inherently contrast with the rapid pace of modern technology and innovation. While tech companies race toward disruption, the judicial system serves as a critical check and balance. Court decisions determine AI governance standards and work to thoughtfully ensure that the implementation of technology aligns with legal and democratic principles.  

The legal tech industry sits closer to this tension than most. At Reveal, we get a front-row seat to how judges view artificial intelligence, both in legal workflows and in the evidence that comes before them. We’ve been having ongoing conversations with practicing judges and lawyers to understand how they’re experiencing technology development on the courtroom floor.  

Judicial Mood: Technology Optimism

Headlines are buzzing with horror stories of lawyers getting caught red-handed with AI hallucinations in their case materials. With courts imposing costly fees, sanctions, and professional risk mounting, many legal departments are hesitant to implement GenAI.  

In 2025, Thomson Reuters published a report on government legal departments noting that “36% of court and legal professionals surveyed described their agency’s attitude toward GenAI as pessimistic or apocalyptic.”  

Yet, when we dug deeper into the court's response to GenAI use, we found that they aren’t anti-tech, but pro-defensible tech. Judges are seeing the same benefits many legal professionals are: time savings, cost reduction, fewer manual errors, and scalability. The State Bar of California released an Advisory Regarding Artificial Intelligence (AI) Hallucinations where it stated that, GenAI “has a wide-ranging application for the practice of law and administrative functions of the legal practice.”  

They recognize the value GenAI offers to the legal profession and the foothold it’s established in the space. But they pair that acceptance with expectations for how attorneys use AI, stressing that, “while these tools can be helpful in streamlining some aspects of legal work, attorneys must use them in a manner consistent with their duty of competence, diligence, and responsibility as managerial and supervisory lawyers” (The California State Bar).

They ask lawyers to use common sense and critical caution when leveraging GenAI. They remind attorneys that AI functions as a tool and, like a chainsaw or a hammer, any tool can cause damage if not used correctly. So, they encourage lawyers to learn how to use it properly, promoting a “run toward it, not away from it” mindset.

“I think we should be excited. I think we should be cautious and skeptical about technology, but I think it’s here, and you need to run toward it, not run away.” — U.S. Magistrate Judge Allison Goddard, U.S. District Court for the Southern District of California

Judge Goddard captures the duality the legal profession is navigating: AI can be genuinely transformative if lawyers implement it thoughtfully. Courts will hold practitioners to that standard.

GenAI and TAR: Encouraged and Normalized, But Must Be Validated

Judges are embracing GenAI and technology-assisted review (TAR) as mainstream tools for document review. Judge Goddard stated, “Those who know how to harness [AI] are going to be the most successful. I think both [TAR and GenAI] are better than human review and hold a lot of promise.”  

Across our conversations, judges acknowledged that AI would continue to play a central role in legal practice, and they actively encouraged lawyers to invest the time required to harness it effectively.

“Lawyers can do it as well if you take time to learn the technology. Technology has always been helpful to the legal profession.”  — Judge Socrates Peter Manoukian, Santa Clara County Superior Court

That said, enthusiasm for AI adoption comes paired with a consistent caution against deploying these tools without guardrails. “Technology can step in and help us. But we should still maintain some skepticism to ensure that the tools are accurate and meeting standards,” Goddard advised.  

Judges and Bar Associations expect legal teams to uphold rigorous standards when deploying AI, including transparency about how tools work and pragmatism about their limitations. They acknowledge that the technology is still maturing and encourage legal teams to invest in tools that continue to improve.

“Make sure the technology is working and push that it gets better and more accurate. We’re in an information revolution, we’re swirling in the tornado.”  — U.S. Magistrate Judge Allison Goddard, U.S. District Court for the Southern District of California

A hammer in a tornado is an apt metaphor for this moment: powerful, potentially dangerous, and in need of a steady hand. There’s real uncertainty around the deployment of GenAI tools and judges are actively working to establish the frameworks to manage it.

Tech Credibility: Involve Specialists Early

Judges noted concern around the gap between lawyers’ use of technology and their ability to defend those choices in court. Goddard said, “I don’t believe what you tell me when it comes to technology, because I don’t believe that you know what you’re talking about. But I also believe that you’re trying to advocate for your client.” This credibility gap erodes trust in the courtroom. So, judges are pushing lawyers to bring subject matter experts into the process, both to bolster defensibility and to avoid black-box disputes over how a tool arrived at its outputs.

“Get the technical people involved. You’ll build credibility with the court. It’s good for your client.”  — U.S. Magistrate Judge Allison Goddard, U.S. District Court for the Southern District of California

Judges aren’t asking lawyers to become computer scientists or get additional degrees in artificial intelligence. They’re asking lawyers to know their limits and bring in the people who can speak credibly to how the technology operates.

Informal Discovery Conferences: Reduce Cost and Conflict

Judges temper their optimism around AI adoption with a clear push for better communication between opposing counsel. They consistently recommend informal discovery conferences as a cost-effective way to resolve eDiscovery disputes before they escalate.

“90% of discovery disputes that go to informal discovery conferences are resolved there through negotiation—it’s much more efficient.”  — U.S. Magistrate Judge Allison Goddard, U.S. District Court for the Southern District of California

Judges are encouraging lawyers to counteract polarization patterns by investing time upfront to align, especially as eDiscovery review shifts from keyword search to AI-generated prompts.

AI prompts offer more comprehensive search capabilities, but they also require more upfront agreement on scope and methodology. Without that shared foundation, disputes downstream become expensive and difficult to resolve.

“If your methods aren’t reasonable, you could be spending a lot more money trying to fix issues after the fact… that’s why I encourage early, clear discussions.”  — U.S. Magistrate Judge Allison Goddard, U.S. District Court for the Southern District of California

Metadata: The Source of Verification

AI now permeates legal workflows and increasingly shows up in the evidence itself.

“We’re getting lots of ESI questions, deepfakes, photos, videos, etc. Someone will come in and claim that the photo really isn’t them or wasn’t taken by them.” — Judge Danna Nicholas, San Diego County Superior Court

AI-edited content, deepfakes, and tampered evidence are introducing new skepticism around the defensibility of electronically stored information (ESI). As a result, lawyers must work harder to establish the provenance and integrity of their evidence. Judges point to intact metadata as the key mechanism for keeping ESI defensible. As Judge Nicholas put it, “If you’re an attorney, look at the metadata—see when it was changed or altered.”

Conclusion: Defensible, Pragmatic, Transparent Discovery

Judges are optimistic about AI in legal workflows, and they believe the lawyers who succeed will be the ones who learn to harness these tools effectively. But conditions shape that optimism. Judges distrust blind faith in AI outputs and expect legal teams to validate results, involve technical specialists, communicate transparently about methodology, and preserve metadata as a foundation for defensibility.

Across our conversations, the throughline is a demand for accountability and rigor as lawyers adapt to new technology. The encouraging development is that the guardrails judges describe increasingly live inside the tools themselves, which closes the gap between what courts expect and what lawyers can actually deliver.

Validation offers the clearest example. Judges want proof that a tool works before a team relies on it, and that proof can now precede the work rather than follow it. Reveal's aji engine builds this in directly. A calibration review compares an attorney's own coding decisions against the GenAI ratings on a sample set and produces an agreement rate, and a separate validation pass confirms accuracy against a second random set. The lawyer walks into a discovery conference with a number instead of a claim.

Transparency follows the same pattern. Judges want to see how a tool reached its output, and they trust answers they can trace. Reveal designed aji and ASK around that expectation. Every positively rated document carries its reasoning and in-text citations and ASK answers natural-language questions with links back to the source material rather than asking anyone to accept an unsourced summary. The black box opens on its own.

Provenance completes the picture. An AI output holds up only as well as the path back to its source survives. Tools that ground every response in the underlying documents, with citations a reviewer can open and check, keep that path intact and keep the evidence defensible under scrutiny.

For legal teams, the path forward is clear. Embrace what AI makes possible, choose tools that carry the guardrails inside them, and communicate openly about how the work gets done. The pragmatism and transparency judges are asking for is not a new bar. It is the one lawyers have always cleared, now applied to the tools, and the teams that succeed will be the ones whose technology answers for itself the way an attorney always has.

That standard is already within reach. The tools exist to meet it, and the teams that adopt them now will be the ones prepared when the court asks.

Hold your AI to the same standard you hold yourself. See how Reveal builds validation, transparency, and traceability into the tools. Book a walkthrough