Artificial intelligence is no longer a theoretical issue in litigation. Law firms are developing AI policies, courts have issued guidance on attorney use of generative AI, and litigants are increasingly encountering AI-generated work during discovery. More recently, attention has begun to shift to another question: what happens when expert witnesses incorporate AI into their work?
The answer is still developing. While AI has the potential to assist experts in a variety of ways, courts are beginning to consider whether—and under what circumstances—its use becomes part of an expert’s methodology. As those questions make their way through the courts, litigants should expect increased attention to how AI is used, how its outputs are validated, and whether that process is subject to discovery.
AI Is Blurring the Line Between Work Product and Methodology
The reliability of expert testimony has never depended solely on an expert’s conclusions. Courts also examine the methodology used to reach those conclusions, evaluating whether the opinions are based on sufficient facts and reliable principles and methods under Rule 702 of the Federal Rules of Evidence.
Generative AI introduces a new wrinkle to that familiar framework. Unlike traditional analytical software—which generally performs defined calculations or applies established models—generative AI can assist with research, summarization, drafting, and even the development of analytical approaches. That creates a more nuanced question: when AI contributes to an expert’s work, is it simply another productivity tool, or has it become part of the methodology supporting the expert’s opinions?
Why does that distinction matter? Because the focus shifts from the technology itself to the expert’s process. If AI contributes to the development of an opinion, litigants may increasingly examine how it was used, what role it played in the analysis, and whether the expert can fully explain and defend the resulting opinions.
Importantly, this does not suggest that AI is incompatible with expert testimony. Nor does it require a new evidentiary framework. Rather, it raises practical questions about transparency. If AI contributes to an expert’s analysis, can the expert explain precisely how it was used? Can they distinguish their own reasoning from AI-generated content? Can they independently validate every opinion they ultimately present?
When the Process Becomes Discoverable
One of the first disputes to illustrate these issues emerged in Conservation Law Foundation, Inc. v. Shell Oil Company et al. There, the plaintiff’s expert used a generative AI tool to narrow a large volume of the defendants’ document production into a workable subset for their analysis. The federal court then ordered production of the expert’s AI prompts after concluding they were relevant to evaluating the expert’s methodology and therefore discoverable. Although that order has since been stayed pending further review, the dispute is notable less for its outcome than for the questions it raises.
Historically, discovery disputes involving experts have focused on familiar subjects: the data reviewed, the assumptions made, the analyses performed, and the basis for the expert’s conclusions. The Shell dispute suggests that AI-assisted work may increasingly become part of that conversation when it contributes to the development of an opinion.
The discussion is no longer limited to individual court decisions. Organizations including The Academy of Experts and practitioners writing on AI-assisted expert testimony have begun publishing guidance addressing how experts should incorporate AI into their work. Although the recommendations vary, they consistently emphasize human oversight, documentation, confidentiality, and the expert’s continuing responsibility for every opinion ultimately presented. Together, these efforts suggest the profession is beginning to treat AI not simply as another productivity tool, but as a component of the expert’s analytical process that may warrant additional scrutiny.
Professional Judgment Must Remain Transparent
Emerging guidance consistently returns to one principle: regardless of how AI is used, responsibility for the final opinion never shifts from the expert. AI may assist with organizing information, summarizing materials, conducting preliminary research, or drafting portions of a report, but it cannot replace the expert’s independent judgment or professional responsibility.
That expectation is not new. Courts have never accepted an opinion simply because it was generated by software, and there is little indication that generative AI will be treated differently. What may be changing is the level of documentation expected to demonstrate that judgment if AI becomes part of the expert’s workflow.
More important than preserving every AI interaction is preserving the expert’s reasoning. If AI contributes to organizing information, identifying themes, or drafting preliminary language, the expert should still be able to explain how those outputs were evaluated, what was accepted or rejected, and why the final opinion reflects their own independent analysis. A clear record of that process may become increasingly valuable if an expert’s work is later examined during discovery or cross-examination.
Ultimately, the issue is not whether AI was used. It is whether the expert can demonstrate that technology informed the analysis without replacing the expert’s own professional judgment. The credibility of the opinion will continue to depend not on the software, but on the expert’s ability to explain and defend every conclusion presented to the court. As Eleanor Hamilton, DOAR’s VP of Experts, said: “The experts who will thrive are not the ones who avoid AI; they are the ones who can hand a court their prompts and still stand behind every conclusion. An expert who can’t reconstruct how a tool shaped their analysis faces more than a credibility problem; they also face a potential discoverability problem before they even reach the stand. Explainability is becoming its own form of credibility.”
Cross-Examination Is Likely to Follow the Methodology
Historically, cross-examination has focused on familiar questions: Were the underlying facts complete? Were the assumptions reasonable? Was the methodology applied consistently? AI is unlikely to replace those lines of inquiry, but it may expand them. Attorneys may begin asking not simply whether AI was used, but how it influenced the expert’s analysis, what safeguards were employed to verify its outputs, and whether the expert exercised independent judgment at each stage of the process.
Confidentiality may also become part of that discussion. Depending on the platform used and the safeguards in place, experts and counsel will need to consider whether confidential client information, proprietary data, or other protected materials are appropriate for use with AI tools. While many organizations have already developed internal AI policies to address these concerns, questions surrounding data security and confidentiality will likely continue to evolve alongside the technology.
Ultimately, these inquiries are less about artificial intelligence than about credibility. An expert who can clearly explain how AI was used—and just as importantly, where independent professional judgment was exercised—is likely to be in a stronger position than one who cannot. In that sense, AI creates another avenue for testing an opinion, but not a fundamentally different standard for evaluating it.
New Technology, Existing Principles
Much of the public conversation surrounding AI assumes that entirely new rules will be needed to govern its use in litigation. Recent developments suggest a more measured approach may be emerging.
The disputes surfacing today are not asking courts to decide whether AI is permissible. Instead, they raise familiar questions in a new context. Was the expert’s opinion independently formed? Can the analytical process be explained and defended? Was AI used responsibly and appropriately within that process? Those are extensions of the same principles that have long governed expert testimony.
As additional cases work their way through the courts, more specific guidance will undoubtedly develop. Questions surrounding disclosure, discoverability, confidentiality, and documentation will continue to be refined as AI becomes more common in litigation. But the broader trajectory suggests that AI may ultimately be evaluated through existing evidentiary principles rather than an entirely separate framework. The fundamentals have not changed.
For experts and the attorneys who work with them, that distinction is significant. The technology will continue to evolve, but the qualities that make expert testimony persuasive remain unchanged: a reliable methodology, independent professional judgment, and the ability to clearly explain and defend every opinion offered.
