A California courtroom in 2026 became ground zero for one of the most consequential shifts in personal injury litigation in recent memory. When a defense forensic expert admitted under cross-examination that he had used ChatGPT to calculate future medical costs — without proper methodology documentation, peer-reviewed sourcing, or disclosure to opposing counsel — the jury didn’t just award damages. They sent a message. The $10 million verdict that followed has reverberated across courtrooms nationwide, forcing attorneys, insurers, and expert witnesses to reckon with a fundamental question: what does AI expert testimony credibility actually mean in 2026, and who is responsible when it fails?
The Watershed Verdict: What Happened in California
The 2026 California personal injury case that sparked this national conversation involved a defense-retained forensic economist tasked with countering the plaintiff’s damage calculations. According to California Personal Injury Law Updates 2026, the expert’s final report contained medical cost projections that opposing counsel found inconsistent with documented billing records and standard actuarial methodologies. During voir dire examination of the expert witness, it emerged that portions of the future medical cost analysis had been generated using ChatGPT — a large language model not designed for forensic economic analysis and incapable of accessing real-time, jurisdiction-specific medical billing data.
The expert had neither disclosed this AI-assisted methodology in his written report nor flagged it during initial depositions. When the admission surfaced at trial, the credibility of the entire defense damages framework collapsed. The jury, already attentive to growing cultural awareness of AI limitations, awarded the plaintiff $10 million — a verdict widely attributed in part to the defense expert’s credibility collapse. This single case has become a defining reference point for AI expert testimony credibility challenges throughout 2026 and beyond. It also arrives at a moment when artificial intelligence is no longer a futuristic concept in California personal injury law — it is already inside the courtroom, the insurance adjuster’s office, and the claims process, meaning that understanding how AI is changing things could be the difference between a fair recovery and a settlement that falls far short of what injured victims are owed.
For plaintiffs evaluating what their own cases may be worth, using a reliable personal injury settlement calculator based on documented methodologies remains a critical first step before engaging expert witnesses or entering settlement negotiations.
Why AI-Assisted Expert Analysis Creates Legal Vulnerability
The Daubert Standard and AI’s Reliability Problem
Under the Federal Rules of Evidence Rule 702, expert testimony must be grounded in sufficient facts or data, the product of reliable principles and methods, and applied reliably to the facts of the case. California’s equivalent evidentiary framework imposes similar gatekeeping obligations. Most AI-generated evidence enters the courtroom through expert testimony subject to Federal Rule of Evidence 702, as interpreted by the Supreme Court in Daubert, which requires trial courts to act as “gatekeepers” to admit only expert witness testimony that is based on sufficient facts or data, is the product of reliable principles and methods, and reflects a reliable application of those principles to the facts of the case. AI language models like ChatGPT fail on multiple prongs of this standard: they cannot cite verifiable, jurisdiction-specific data sources; they do not apply consistent actuarial or medical cost methodologies; and their outputs cannot be independently replicated or audited in any meaningful forensic sense.
When a defense expert uses AI to generate damage estimates without disclosure, they are essentially presenting unverifiable outputs as professional analysis. Used well, AI can help an expert organize information, test scenarios, and explain complex material more clearly — but used poorly, it can produce outputs that look persuasive but are difficult to explain, difficult to reproduce, or harder to defend under the rules that govern expert testimony. The hallucination problem compounds the risk: Stanford researchers documented error rates of 69–88% for general-purpose LLMs on legal queries, with even paid legal AI tools hallucinating at rates of 17% or higher.
The Documentation and Disclosure Gap
While there is currently no single unified guidance specifically governing expert witnesses’ use of AI, they must always comply with their duties under existing procedural rules — meaning expert evidence presented to the court should be the independent product of the expert, providing objective unbiased opinion in relation to matters within their expertise, and stating the facts or assumptions on which their opinion is based. The documentation gap created by AI use is particularly dangerous. That reliability concern ties directly to Daubert and Federal Rule of Evidence 702 — generative-AI tools raise pointed questions about reproducibility, since the same prompt can yield different outputs, models are updated and retired, and the evidentiary foundation shifts beneath the expert’s opinion.
Experts should understand the strengths and limitations of every AI platform they use — without that understanding, they may struggle to explain their methodology during cross-examination, and this weakness may undermine credibility before judges and juries alike. The documentation and disclosure gap also creates exposure at the firm level. Courts have flagged AI-fabricated citations in roughly 1,600 cases, and ABA Formal Opinion 512 places the duty to verify on the lawyer, not the software.
The Broader Landscape: Social Inflation and Nuclear Verdicts in 2026
The California AI expert credibility case did not emerge in a vacuum. It is part of a dramatically escalating personal injury verdict environment. Nuclear verdicts continue to be a major issue: in 2024 alone, there were a total of 135 nuclear verdicts, a 52% increase over 2023, with total awards exceeding $31 billion, while liability claim costs have grown roughly 57% over the past decade, largely due to social inflation. The RPS 2026 Q2 Umbrella and Excess Market Update found that median nuclear verdicts have increased more than 25% over the past decade, outpacing inflation, and that in many cases less than 15% of an award is tied to actual economic loss, with the remainder driven by non-economic and punitive damages.
The average bodily injury payout has climbed to approximately $29,100 per injured person, with bodily injury claims rising about 11% over the past two years — and the cost per claim rising at three to four times the rate of overall inflation. Industry research from the brokerage firm CRC Group estimates that economic inflation combined with social inflation has added more than $90 billion to personal auto insurance losses over the past decade. Meanwhile, third-party litigation funding is contributing to longer settlement times and higher-severity claims, particularly in mass tort and complex liability cases, with verdicts increasingly piercing primary limits and reaching into excess and umbrella coverage.
2026 is shaping up to be a pivotal year for third-party litigation funding (TPLF), with landmark legislative proposals, judicial decisions, and regulatory developments across the U.S. In February 2026, Senator Chuck Grassley introduced the Litigation Funding Transparency Act of 2026, which proposes to require disclosure of outside investors in federal class actions and multi-district litigation, restrict funders from controlling legal strategies, and bar their access to confidential discovery materials — legislation currently pending in the Senate Committee on the Judiciary. Within this environment, a defense expert whose AI-assisted methodology collapses under cross-examination is not merely losing a damages argument — they may be handing plaintiff counsel the narrative that converts a reasonable settlement into a nuclear verdict.
One key trend in 2026 is that jury behavior feels far less predictable than before: two similar cases can produce completely different outcomes simply because they were filed in different regions, as local attitudes now play a much bigger role in how jurors view injury claims, large settlements, and corporate responsibility — and in some cities, juries remain highly sympathetic toward injured victims, especially when large companies or commercial insurers are involved. When a defense expert’s AI methodology is exposed as undisclosed and unreliable, that dynamic intensifies dramatically.
What Courts Are Doing About AI Expert Testimony Credibility in 2026
Emerging Disclosure Protocols and Judicial Orders
The most significant development in 2026 regarding AI expert testimony credibility came from a federal court in Connecticut. On May 18, 2026, Magistrate Judge Thomas O. Farrish of the U.S. District Court for the District of Connecticut ordered the plaintiff in Conservation Law Foundation, Inc. v. Shell Oil Company, et al. to produce the generative AI prompts that its expert witness, Dr. Naomi Oreskes, used in preparing her expert report. This appears to be the first federal court decision requiring an expert witness to disclose AI prompts as part of discoverable methodology for the expert’s opinions.
The plaintiff argued that the prompts were used only to “cull a large document universe” and were “never considered by the witness in forming her opinions,” but the court was unpersuaded: if the prompts shaped which documents the expert ever saw, then they shaped the evidentiary foundation of the opinion, and a party cannot meaningfully test the reliability of that opinion without understanding how the universe of considered material was narrowed. The ruling’s core logic — that the way an expert uses AI is part of the expert’s methodology, and methodology is discoverable — is intuitive enough that other courts are likely to find it persuasive.
A parallel legislative development is also reshaping the playing field. The U.S. Judicial Conference’s Committee on Rules of Practice and Procedure has proposed Federal Rule of Evidence 707, which would subject any AI-generated evidence offered at trial without a corresponding human expert to a reliability test. The Advisory Committee will reconvene in mid-2026 to consider public comment, with a timeline for adoption in 2027 at the earliest — and if adopted, this would be the first explicit rule requiring disclosure of AI in federal proceedings, while states would remain free to adopt Rule 707 entirely, in part, or implement their own systems.
Courts are also confronting AI disclosure issues beyond structured expert reports. In Kohls v. Ellison, the plaintiffs’ expert submitted a report that included passages drafted with generative AI, but the role of the tool was not disclosed — the court insisted on transparency, requiring the expert to identify what the AI produced, explain how it was reviewed and edited, and preserve prompts or logs. A March 2026 survey conducted by Northwestern University of 502 randomly selected federal judges found that over 60% use at least one AI tool in their chambers, while about 25% formally permit AI use and 20% ban it — indicating the bench itself is actively grappling with where the line falls.
How Plaintiff Attorneys Are Responding
Plaintiff attorneys in 2026 have developed increasingly aggressive strategies for detecting and challenging AI-assisted defense expert testimony. AI is being used to simulate adversarial questioning — by feeding an expert’s report and prior statements into a model configured to challenge assumptions and probe weaknesses, lawyers can pressure-test testimony before it ever reaches the courtroom, which does not replace traditional mock examinations but enhances them, and on the flip side, lawyers challenging opposing experts can use AI to synthesize prior testimony and technical literature into focused lines of cross-examination that expose contradictions or unsupported assumptions.
An experienced plaintiff attorney understands how AI valuations are generated, what they miss, and how to counter them with properly documented evidence — and defense AI tools can crawl public social media profiles, detect body-movement patterns in videos, and flag posts whose timestamps conflict with injury claims. The risk, however, cuts both ways: expert witnesses present a particularly sensitive variation, as an expert who uses AI to refine opinions, anticipate cross-examination, or pressure-test conclusions risks blurring the line between independent analysis and external influence, and even when opinions remain substantively unchanged, the appearance of AI-assisted refinement can invite aggressive questioning about methodology, authorship, and reliability.
Best practices emerging from 2026 litigation suggest that AI-assisted expert reports should disclose the AI tool by name, version, and vendor; identify the training data category used; state quantitative performance metrics used to validate the model; and document all prompts, queries, or inputs submitted to the AI system.
Systemic Implications for Insurance Defense and Expert Witness Practice
The cascading impact of the AI expert credibility crisis is reshaping how insurance defense teams approach expert witness management. At AIDA’s 2026 Insurance Law Forum: Global Perspectives, presenters demonstrated what an experienced insurance defense attorney can now execute in a single AI-assisted workflow: analyze a plaintiff’s expert report and contrast it with opposing counsel’s position, surface the strongest support from thousands of pages of discovery, build targeted deposition questions to challenge methodology and conclusions, and research complex apportionment doctrine across jurisdictions — tasks that would have collectively consumed days, now completed in hours. But this power cuts both ways: the same AI efficiency that helps build a defense can expose it to devastating credibility attacks if the underlying AI use is undisclosed.
The hallucination sanction record in 2026 has made the stakes unmistakable. In the costliest AI hallucination sanction in U.S. legal history to date, U.S. Magistrate Judge Mark D. Clarke imposed a total of $110,000 in fines and attorneys’ fees against two lawyers in a winery dispute case in Oregon, after an AI tool led human minds astray — and the court dismissed the case entirely. Courts have moved well beyond warnings, with one court stating directly that courts “should begin meeting this challenge with an eye towards deterring similar conduct” that will “necessarily and unfortunately, involve moving beyond admonitions and reprimands into more punitive sanctions” — meaning attorneys and law firms should treat current case law not as a ceiling but as a floor.
For insurers, the financial picture is acute. The divide between property and casualty is the defining feature of 2026 — commercial auto has extended its 59-quarter streak of increases, while umbrella and general liability continued to rise, with lines tied to bodily injury, litigation, claims severity, social inflation, and jury verdicts remaining structurally challenging. Even before reaching a jury, the average cost per claim continues to rise due to medical inflation, repair cost inflation, and litigation-driven settlement values — and even before trial, bodily injury claims awards are now multiples of what comparable claims settled for five years ago. A defense expert credibility collapse — especially one tied to undisclosed AI use — accelerates every one of these pressures.
Frequently Asked Questions About AI Expert Testimony Credibility in Personal Injury Cases
Can a personal injury defendant’s expert witness legally use AI tools to prepare their analysis?
Yes — but with significant legal and ethical constraints. Artificial intelligence has entered the expert testimony space not as a replacement for experts but as a tool reshaping how expert testimony is analyzed, prepared, and challenged — and to be clear, no AI system is about to take the stand, but what AI can do is change the mechanics of how expert evidence is reviewed and tested. The legal exposure arises not from using AI, but from failing to disclose it, failing to verify its outputs, or allowing AI-generated content to substitute for independent expert analysis. Courts have been clear that existing procedural and evidentiary rules apply to AI just as they do to other technologies — expert disclosures still must explain the “how” and “why” of an opinion, even when AI is involved.
What is the Daubert standard and how does it apply to AI-assisted expert testimony?
The Daubert standard, derived from Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993) and codified in Federal Rule of Evidence 702, requires trial courts to serve as gatekeepers who admit expert testimony only when it rests on sufficient facts, reliable methodology, and reliable application of that methodology to the case facts. The reliability concern ties the 2026 AI disclosure rulings directly to Daubert and Rule 702 — generative-AI tools raise pointed questions about reproducibility, since the same prompt can yield different outputs as models are updated and retired. Courts applying Daubert must determine whether sufficient information about model methodology is publicly available to evaluate reliability without full code disclosure — and that tension remains unresolved at the circuit level.
How does undisclosed AI use affect a personal injury settlement?
The impact is severe and multidirectional. When undisclosed AI use by a defense expert is revealed during deposition or trial, it triggers an immediate credibility collapse that typically increases settlement pressure on the defense side dramatically. An AI tool that outputs a settlement range for a claim carries implicit authority that can pressure an unrepresented claimant to accept less than their case is worth — but when the defense AI methodology is exposed, that same dynamic reverses, making the defense’s damages case appear engineered rather than evidence-based. Major insurers use AI-driven valuation platforms that generate recommended settlement ranges based on comparable claim data, systems that often undervalue soft-tissue injuries, psychological harm, and future medical needs — but an experienced attorney can counter these valuations with properly documented evidence, expert testimony, and knowledge of how these platforms are calibrated.
What should plaintiff attorneys do to challenge AI-assisted defense expert testimony?
In 2026, the playbook for challenging AI-assisted defense expert testimony has crystallized around several key strategies. First, pursue early and targeted discovery into the expert’s AI tool use. Attorneys nationwide are already rewriting how they manage expert witnesses who use AI in the wake of the Connecticut federal court ruling compelling production of AI prompts. Second, file a Daubert motion challenging reproducibility and methodology documentation. Third, lawyers challenging opposing experts can use AI to synthesize prior testimony and technical literature into focused lines of cross-examination that expose contradictions or unsupported assumptions. Fourth, use the growing body of hallucination case law to illustrate systemic unreliability. A public database maintained by legal researcher Damien Charlotin tracks roughly 1,490 court decisions worldwide — more than 1,000 of them in the United States as of May 2026 — where a party relied on AI-hallucinated material and a court responded.
Will courts require mandatory AI disclosure from expert witnesses in personal injury cases going forward?
The trajectory points clearly toward mandatory disclosure, though the legal framework is still being built. The Advisory Committee’s timeline for adopting proposed Federal Rule of Evidence 707 is 2027 at the earliest — and if adopted, it would be the first explicit rule requiring disclosure of AI in federal proceedings, while states would remain free to adopt it entirely, in part, or implement their own systems. In the meantime, the core logic established by the Connecticut federal court — that the way an expert uses AI is part of the expert’s methodology, and methodology is discoverable — is intuitive enough that other courts are likely to find it persuasive, meaning the question is no longer whether opposing counsel will ask about AI use but how prepared each side will be when they do. For personal injury plaintiffs and their counsel, that shift in the evidentiary landscape represents both a powerful new tool for challenging defense experts and a responsibility to ensure their own experts maintain fully documented, independently verifiable methodologies. Using a transparent, data-driven personal injury settlement calculator remains an essential baseline for building a damages case that can withstand scrutiny in an era where every number will be questioned.
Thomas B. Harrison is a personal injury legal consultant with extensive experience connecting injury victims with qualified attorneys across the United States. He specializes in helping people understand when they need legal representation and how to find the right personal injury attorney for their specific situation. Thomas is not an attorney and the information he provides is for educational purposes only.