Will AI Eventually Replace Some Types of Expert Testimony?

AI is starting to change expert work. It is not, however, on a clear path to replacing expert testimony as a whole.

A more useful question for litigators is narrower: which parts of expert work are becoming more automatable, and which parts still depend on human judgment, credibility, and live testimony? Based on what litigation teams are already seeing, the answer is practical. AI can reduce time spent on certain front-end tasks. It is far less capable of replacing the expert functions that matter most when a case is under real pressure.

As Courtney Grubbs-Donovan, Vice President of Research at Expert Institute, explained, “I think certain functions” can be replaced or reduced, particularly “record analysis especially in medical malpractice cases.” Drawing on her experience overseeing complex expert research and case support, Courtney noted that AI can “help you pinpoint what portions of the record you should focus on,” potentially saving time and reducing the amount of expert review needed during the earliest stages of a case.

Where AI is already affecting expert work

The first wave is not replacement. It is triage.

Grubbs-Donovan noted that clients are beginning to use AI-assisted tools for record review and case analysis, especially where the initial task is sorting large volumes of information. In medical cases, that can mean identifying key portions of a chart, flagging potential timelines, or narrowing the set of materials that need close human review.

She also pointed to adjacent uses in litigation prep, including tools that help attorneys research experts, prepare for depositions, and organize case materials. In other words, AI is already changing workflow before it changes testimony.

That matters because expert costs often begin long before an expert ever appears in deposition or at trial. If AI compresses the early review process, some categories of expert work may become narrower, faster, or more targeted.

What AI is unlikely to replace

The core problem for full replacement is straightforward: expert testimony is not just information retrieval.

“You can’t have an AI agent standing up on the stand,” Grubbs-Donovan said. More importantly, expert work in litigation is not limited to generating an answer. It requires the expert to defend methodology, respond to new facts, handle cross-examination, and connect an opinion to the actual record.

That is where human expertise still carries most of the weight.

Grubbs-Donovan put it this way: “At the end of the day, we’re still in a world where a human’s experience and opinion on something is what makes the difference.” AI may help someone reach a conclusion. It is not yet a substitute for the witness who must own that conclusion in a courtroom.

That distinction is especially important in nuanced cases. Even when the underlying subject matter appears technical or guideline-driven, litigation usually turns on context. As she noted, “in these lawsuits there’s always more context to the scenario.”

The risk is overreliance, not use

The most immediate concern is not that experts will use AI. It is that they will use it poorly.

Grubbs-Donovan described a case in which an expert was asked to support a position with data and instead produced what amounted to a list of ChatGPT-generated publications. The sources could not actually be found or were too old to support the point. “You can’t rely on AI,” she said, “you need to have the expert who actually knows their stuff.”

That example captures the real litigation risk. An expert who relies on AI summaries, AI research, or AI-generated citations without independent validation may create avoidable impeachment material.

In practice, that risk shows up in several ways:

• relying on an AI summary instead of reviewing the full record
• citing authorities or publications that cannot be verified
• missing facts or themes that a qualified expert would catch
• adopting conclusions the expert cannot fully explain or defend

Which expert functions may change most

Some work is more susceptible to AI assistance than other work.

Medical record review is one clear example because the records are document-heavy and often structured in ways that lend themselves to summarization and issue-spotting. Grubbs-Donovan observed that this could mean “less of a reliance on actual doctors to do the record review” at the front end, even if physician testimony remains indispensable later.

By contrast, fields that depend heavily on contextual judgment, field investigation, or lived professional experience are less likely to be displaced. Cases involving accident reconstruction, site conditions, forensic inspection, or nuanced standard-of-care disputes still require someone who can connect facts, practice, and real-world judgment in a way that holds up under questioning.

The likely future: faster preparation, not fewer experts

The most defensible prediction is also the least dramatic. AI will likely expedite parts of case preparation for both attorneys and experts. It may reduce some preliminary review work. It may narrow issues earlier. It may make certain kinds of analysis more efficient.

But efficiency is not the same as substitution.

For now, the better view is that AI will reshape expert workflows more than it replaces expert testimony. The cases that matter most still require an expert who has done the work, reviewed the real record, and can defend every opinion when the questioning gets difficult. In litigation, that remains a human job.