News|Crimson raises $2.5m seed round
Crimson
Back to Blog
Best Practices

AI-Assisted Witness Statement Preparation: Best Practices and Ethical Boundaries

The Civil Justice Council has proposed requiring lawyers to declare that AI was not used to generate the content of trial witness statements. Here is what that means in practice, and where AI can still add value without crossing the line.

Witness statements sit at a unique intersection of legal process and human memory. They are supposed to reflect what a witness personally observed, in their own words, about facts relevant to the dispute. In England and Wales, Practice Direction 57AC reinforces this with specific requirements: the statement must contain only matters of personal knowledge, must be in the witness's own language, and must be accompanied by both a signed confirmation of compliance from the witness and a certificate of compliance from the solicitor.

Against that backdrop, the question of whether and how AI can assist in the preparation of witness statements has become increasingly urgent as the courts and regulators move toward formal guidance.

The Proposed Rules

In February 2026, the Civil Justice Council (the body that advises the Lord Chancellor and the judiciary on the development of the civil justice system) published a consultation on the use of AI in preparing court documents. The consultation closed on 14 April 2026, and its proposals for witness statements were notably stricter than for other types of litigation documents.

For trial witness statements prepared with the involvement of a legal representative, the CJC proposed a rule requiring a declaration that AI has not been used to generate the content of the statement, including by way of altering, embellishing, strengthening, diluting or rephrasing the witness's evidence. Administrative uses such as transcription and spell-checking would fall outside the declaration requirement, but anything that touches the substance of what the witness is saying would be caught.

The Law Society's response went further, recommending that all witness statements (not only those covered by PD 57AC) should carry a declaration of non-AI use. Its position is clear: AI should not be used to alter, embellish, strengthen, dilute or rephrase a witness's evidence in any context.

Godwin v Godwin and Judicial Scepticism

The early judicial signals reinforce this direction. In Godwin v Godwin [2026] EWHC 923 (Ch), HHJ Klein dealt with a defendant and his former wife who had used ChatGPT to assist with their witness statements. They described it as a "digital assistant" used for grammar, spelling and presentation, claiming that ChatGPT had not added, removed or rearranged any words from their first drafts.

The judge was unpersuaded. He noted that both witnesses were clearly sophisticated people capable of using standard word processing tools for spell-checking, and that there was no good reason for them to have used ChatGPT at all. The concern was straightforward: once AI has been applied to the text, the court cannot be confident that the final wording is truly the witness's own. The judge approached the relevant evidence with caution as a result.

The case illustrates a broader principle. Even where the AI intervention is described as minimal, the court will treat the resulting statement with scepticism because it cannot verify the boundary between what the witness wrote and what the AI contributed.

The "Own Words" Requirement

PD 57AC requires the statement to be in the witness's own words so far as practicable, to reflect their personal recollection, and to identify any documents the witness was referred to during preparation. The solicitor must certify compliance. These requirements create a direct conflict with any process that involves AI generating, restructuring or rephrasing the witness's account.

Even if the AI is instructed only to "tidy up" the language, the output draws on patterns learned from millions of other documents, producing phrasing that is definitively not the witness's own. This is the same concern that has always applied to solicitors over-engineering statements, but AI makes it easier to do at scale and harder to detect.

Trial Statements vs Procedural Statements

It is worth noting the distinction between trial witness statements and the procedural statements that lawyers prepare in support of interim applications. A partner's witness statement in support of a freezing injunction or a summary judgment application is a different animal: it is typically drafted by the legal representative, setting out the factual and procedural background for the court, and the professional responsibility for its contents sits squarely with the lawyer who signs it. The CJC's proposals reflect this distinction. For non-trial statements prepared by legal representatives, the CJC has suggested that no additional AI rules are needed beyond the existing requirement that the document bears the name of the lawyer taking responsibility for it. The stricter proposals around AI declarations apply specifically to trial witness statements, where the witness's own voice and recollection are what the court is relying on.

Legitimate Uses of AI in the Preparation Process

None of this means AI has no role in the witness statement process. There is a clear distinction between using AI to generate the content of the statement and using it to support the preparation work around it.

AI can assist with transcribing witness interviews, structuring interview notes into a logical order for the solicitor to work from, identifying documents in the case file that relate to topics the witness has raised, cross-referencing what the witness has said against contemporaneous documents to flag inconsistencies that need exploring in follow-up sessions, and collating the document list that PD 57AC requires to accompany the statement.

These are tasks that support the solicitor in preparing a compliant statement in a way that should not change the witness's words. The output of the statement itself remains a collaboration between the witness and the legal representative, as it always has been, with AI accelerating the surrounding preparation work.

Practical Recommendations for Practitioners

Given the direction of travel, firms may want to consider the following:

  • Draw a clear internal line between the statement itself (where AI should not change the witness's words) and the preparation work around it (where AI can add considerable value).
  • Ensure that junior team members and paralegals understand where that line sits. The temptation to run a draft through an AI tool to "clean it up" is obvious and easy to act on without thinking through the compliance implications.
  • Document your process, including what AI tools were used and for what purpose. This puts you in a stronger position if compliance is ever questioned. Tools designed specifically for litigators tend to have built-in audit trails that record exactly what was asked and what was returned, which can be invaluable in demonstrating that the AI was used appropriately.
  • Brief your witnesses early that they should not independently use AI to prepare or refine their statements before sending them to you. This avoids problems that are difficult to fix after the fact.

The regulatory position is still developing, and the CJC consultation may result in formal rule changes later in 2026. But the direction is clear enough to act on now.

If you would like to see how Crimson accelerates the witness statement preparation process while respecting the boundaries set out above, request a demo.

Mark Feldner

Mark Feldner

Co-Founder & CEO, Crimson

Co-Founder & CEO at Crimson. Former litigator at Clifford Chance, WilmerHale, and Willkie Farr & Gallagher with 8 years of experience in complex disputes.

Get Started

Ready to transform your litigation practice?

Join leading law firms using Crimson to streamline document review, build chronologies, and win more cases.

Request a demo