What the CJC Consultation Means for AI in Expert Reports

An expert used AI to draft his report and accidentally submitted misinformation to the court. The Civil Justice Council cites that case, Kohls v Elison, in setting out how it wants experts to handle AI, and the direction is clear. Agencies and experts that get ready before the rule lands turn a compliance obligation into a selling point with instructing firms.

This piece works through what the CJC has proposed, what it means for you, and how to get ready now. It is written from the CJC’s interim report and the working group June 2026 update on consultation responses. A final report is due later in 2026, and this guide will be updated when it lands. For the wider picture, read our full guide to AI for medico-legal record review and our information for medico-legal agencies.

What the CJC is doing

At its January 2025 strategy meeting, the Civil Justice Council set up a working group on the use of AI in preparing court documents. The group, chaired by Lord Justice Birss, published an interim report and consultation that ran to 14 April 2026. A final report follows later this year. The group published an update on the consultation responses in June 2026. As at August 2026 that update reports strong consensus that no new AI rules are needed for pleadings and advocacy documents, naming witness statements as the principal unresolved question, and setting out a direction of travel that includes introducing proportionate transparency in relation to expert evidence. The expert proposal has firmed up rather than fallen away.

The report covers court documents broadly: statements of case, skeleton arguments, witness statements, and expert reports. For statements of case and skeleton arguments, the CJC’s view is that existing professional duties are enough, and no new rule is needed as long as a named legal representative takes responsibility. Trial witness statements go the other way: the CJC proposes a declaration that AI has not been used to generate their content. Expert reports sit between the two, and that is the part experts and agencies need to read.

The proposal for expert reports

The CJC proposes amending the Practice Direction 35 statement of truth, the declaration an expert signs to verify a report. Under the proposal, an expert would identify and explain any AI used, other than administrative uses such as transcription, and name the tools used.

The CJC’s own wording sets the tone. Provided its use is properly identified and explained, the report says, the nature of an expert’s evidence may properly include AI generated material. This is not a ban. It is a disclosure rule. The CJC treats AI as a legitimate part of expert work, on the condition that its use is transparent and the expert stays accountable for the evidence.

Where record review sits

The report notes that it is difficult to see why using AI as a research tool, or to extract material from voluminous documents, should necessarily be disclosed. That line is often quoted on its own, and the rest of the sentence matters: there will be circumstances, as in Kohls v Elison, in which a failure to disclose could put erroneous evidence before the court.

The drafted proposal is narrower than the aside suggests. It asks the expert to explain any AI use other than transcription or other administrative uses, and the report defines administrative uses tightly: spelling, grammar, transcription, accessibility and formatting, and nothing that generates substantive content. Building a chronology from a bundle does not obviously sit inside that definition, so the safe assumption for an expert is that record review is disclosable. That is workable rather than a problem. Disclosure asks an expert to name the tool and explain what it did, which takes a sentence if the chronology is source-linked and is impossible if the bundle went into a consumer chatbot.

Why the CJC cares about disclosure

The CJC reaches this through the Bond Solon Expert Witness Survey 2025, which found that 20 percent of the 525 experts surveyed had used AI in their role as expert witness. Two concerns drive the proposal, and both bear on an expert’s and agency’s exposure.

• Erroneous evidence. The report cites Kohls v Elison, a Minnesota case in which an expert used AI to draft his report and accidentally submitted misinformation. Undisclosed AI use lets that kind of error reach the court unchecked.
• A level playing field. The CJC notes that failing to disclose AI use undermines the level playing field between experts. If one expert’s report leans on unverified AI and the other’s does not, the court cannot weigh them fairly.

Both concerns point the same way. The court needs to know what AI did and be able to check it. An expert who works from a verifiable, source-linked record answers both concerns before they are raised. One relying on an untraceable consumer tool answers neither.

What this means for you

The rule is not in force, but the direction is set, and the readiness gap is visible now.

• Disclosure is coming, not optional. Once Practice Direction 35 changes, every expert who uses AI beyond admin will declare it in the statement of truth. Anyone already working that way is ready. Anyone improvising is not.
• Verifiability is the test. The CJC’s concern is evidence the court is able to check. An AI-assisted fact with no source trail fails that test. A finding linked to its source line passes it.
• The agency carries the readiness. Instructing firms will ask how your panel handles AI disclosure. A consistent answer across every expert is an agency asset. A different answer from each expert can become a liability.

Getting ready now is cheaper than fixing this later. The agency or expert that builds a compliant workflow before the rule lands meets the change from a position of strength, and tells instructing firms so.

The workflow that already meets the standard

The CJC wants AI use that is identified, explained, and verifiable, with the expert accountable for the evidence. That is the description Health Narrator was built to.

It takes the full bundle and produces a structured, source-referenced chronology, every finding linked to the source line, so verification is one click and the audit trail is built in. That is the record the proposed disclosure rule asks an expert to be able to give: what the tool did, and how the output checks against source. The platform does the extraction while the expert forms the opinion, which is what keeps a disclosure statement short and defensible. That reflects the CJC’s distinction between generating substantive content and handling records. It is NHS Data Security and Protection Toolkit (DSPT) Meets Standards certified and DCB0129 clinical safety compliant, and uses no patient data for model training.

The disclosure rule will ask experts to show their AI use is transparent and checkable. An expert or panel already working that way meets the rule the day it lands.

This guide reflects the CJC’s interim report. When the final report is published later in 2026, this page will be updated to match the confirmed position.

Get ready for the disclosure rule.

Book a demo: calendly.com/ed-healthnarrator/30min