Introduction
The recent judgment in Godwin v Godwin [2026] EWHC 923 (Ch) confirms that using AI for the production of trial witness statements may result in such statements being approached cautiously where the Court cannot be satisfied they are in the witnesses’ own words. The case serves as a reminder of the importance of complying with relevant practice directions when preparing trial witness statements and, alongside the Civil Justice Council’s AI consultation, has led regulatory bodies to conclude that AI should not be used for the production of witness statements until further guidance is published.
More broadly, the Civil Justice Council is consulting on refinement of the civil procedural rules and practice directions to improve transparency on the use of AI in the preparation of court documents of all categories – from legal submissions to expert witness reports. The increase in instances of AI-hallucinated material being put before the courts, and the responses to the consultation, all point towards a need for regulatory guidance and court rules that provide certainty, and the challenge of setting standards that can successfully operate in a fast-evolving AI environment. In the interim, legal professionals and their insurers continue to navigate this changing landscape as best they can.
AI Assistance in Godwin v Godwin
The case of Godwin v Godwin involved a determination of how the body of the deceased Mr Godwin should be laid to rest, following disagreement between two of his sons (the Claimant and Defendant in the case).
The case has attracted wider interest because the Court took issue with how the Defendant (Mr Godwin’s son) and a witness in support of the defence (Ms Godwin, the ex-wife of the Defendant) had used AI to produce their witness statements for trial. The witness statements identified that the witnesses had each had ‘limited assistance from a ‘digital assistant for grammar, spelling, and “presentation”’. The judgment identified that what was meant by ‘digital assistant’ was the AI tool ChatGPT.
Both witnesses claimed that they had each prepared the first draft of their statements, which they each uploaded to ChatGPT and that ChatGPT did not add, remove or re-arrange any words when compared to their first drafts. As the Court had not been provided with the draft witness statements, it could not confirm the witnesses’ claims.
His Honour Judge Klein observed, ‘there was no good reason for either the Defendant or Ms Godwin to use ChatGPT in this case. Both struck me as sophisticated people…Both the Defendant and Ms Godwin are sufficiently technically adept to have used ChatGPT. They must have been able to use the spelling and grammar checking facility in Word…’ The Court held that it did not receive a satisfactory explanation from the witnesses about why they chose the phrase ‘digital assistant’ and noted that it is not conventional to refer to an AI tool in that way. His Honour Judge Klein found he could not rule out the possibility that itself was a phrase ChatGPT had contributed to the witness statements. For all these reasons, the judge could not be sure how the AI was used and whether the witness statements were in the witnesses' own words. As such, the judge’s decision was to approach the evidence cautiously.
The decision in Godwin v Godwin is in line with Practice Direction 57AC, which the witnesses in this case were ordered to follow in the preparation of their witness statements for use at the hearing. PD57AC reinforces CPR Practice Direction 32, requiring that trial witness statements should be expressed in the witnesses’ own words.
AI use in witness evidence – the Civil Justice Council’s Consultation and Interim Report
The judgment in Godwin v Godwin coincided with the Civil Justice Council’s (“CJC”) Consultation on the Use of AI in Preparing Court Documents, which concluded in April this year. This considered, amongst other things, legal representatives use of AI for preparing witness statements, including those that fall under PD57AC.
The CJC’s primary role is to advise the Judiciary and the Civil Procedure Rule Committee on civil matters. Although the final report is yet to be published, the CJC’s interim report suggests that the working group undertaking and reporting on the consultation consider supplementing PD57AC with a rule requiring a declaration that AI has not been used for the purposes of ‘generating the content’ of a trial witness statement, which would include altering, embellishing, strengthening, diluting or rephrasing the evidence. The CJC have said they feel this would be consistent with the aims of the Practice Direction and would reinforce the importance of witness statements being in the witness’s own words. The same proposal has been made for trial witness statements falling outside of PD57AC (specifically those subject to CPR Part 32).
The Law Society’s response to the CJC consultation agrees with these proposals, but extended beyond the scope of the consultation to call upon the SRA and HM Courts and Tribunal Service to issue urgent new guidance on the use of AI in court documents in the short term. The Law Society has said that ‘Before any rule change, court guidance could confirm AI should not be used in the production of witness statements and clarify the definition of “producing" witness statements with AI’'.
The Bar Council also provided a response to the CJC’s consultation, saying on witness statements that, ‘many AI tools involve text enhancement or make suggestions about the text, meaning an AI-assisted statement may not be written in the witness’s own words, or could reach conclusions they might not have come up with themselves. We therefore agree with the CJC that a rule should be introduced, requiring witnesses to sign a declaration that AI has not been used to generate their statement.’
It will be interesting to see if these proposals are integrated into the final report which is anticipated to be published later this year. The meaning of “generating content” will, undoubtedly, be subject to scrutiny unless very clear guidance is released to aid in its interpretation. For now, it appears that the safest route is to refrain from the use of AI in the preparation of trial and non-trial witness statements until further guidance is released.
AI use in other court documents
Despite the approach in Godwin v Godwin and the clear objectives of PD57AC as to witness evidence, it is worth noting the positive comments have been made about the use of AI in preparing court submissions. During annual evidence session to House of Lords Constitution Select Committee in June 2026, The Lady Chief Justice, The Right Honourable the Baroness Carr of Walton-on-the-Hill, stated in a discussion about what might be done to help individuals representing themselves (where eligibility for Legal Aid has been narrowed):
‘AI, we are seeing, can really help litigants in person have access to justice because they can get support in drafting their submissions. Now, of course, we as judges are trained to look out for fake hallucinations, fake cases and the like, but many judges are saying to me that they find AI-assisted submissions from litigants in person more helpful, easier to digest, than submissions that are drafted without the support of AI.’
This has been widely reported as praise of AI usage by litigants in person, but it is important to remember that the Court has commented that reliance on AI-generated false citations in proceedings is just as unacceptable by a litigant in person or litigation friend as it is by a regulated professional, even though the sanction for registered lawyers will differ.[1]
The CJC’s consultation largely focused on the need for court rules to regulate the use of AI by legal representatives (and not litigants in person). Their interim report proposals for possible rule changes differ as between four categories of court documents, namely; statements of case, skeleton arguments and other advocacy documents, witness statements, and expert reports. In contrast to the preliminary position taken in relation to witness statements (addressed above) the CJC’s proposal for statements of case and skeleton arguments and other advocacy documents was that there is no need for any further rules relating to using AI in the production of these document, as long as they bear the name of the legal representative taking responsibility for the document. Whilst the Bar Council agreed with these proposals, the Law Society leans towards rules that bring about more transparency on the use of AI.
As with witness statements, the CJC’s pre-consultation position for expert reports proposed an amendment be made to the relevant practice direction, here PD35, which provides the specific form needed for the statements of truth within expert work product. The proposal is that the expert explains what use of AI has been made other than for transcription (or other administrative uses) and that the expert identifies the AI tools used. This declaration would be different to that for witness statements as expert evidence can include certain AI generated material. The Law Society endorses this approach, but the Bar Council has raised concerns with a requirement that experts include an explanation on the use of AI, with no guidance as to the level of detail required in the explanation.
Whilst there is broad acceptance that expert witnesses may use AI for legitimate purposes and in a way that does not cross with an expert’s duty to provide their independent views to the court, we may expect some further attention on the level of transparency and explanation by expert witnesses on their use of AI tools.
AI Law Firms
The first court case won by an AI law firm was heard in May, this year. Garfield AI, the first purely AI-based firm to be authorised by the SRA, won at trial. A barrister was used to conduct the advocacy at trial for the claim over unpaid fees for a freelancer, but the background work, including the preparation of witness statements and a case bundle, was carried out by Garfield AI. The key here, however, is that Garfield AI will only take small claims instructions (up to the value of £10,000) and have achieved a specific SRA authorisation for the way they operate. The format of witness statements in the small claims track is also much less strict than those subject to PD 57AC and PD 32.
Continuing risks of using AI
As regards witness statements, it is clear from the case of Godwin v Godwin that using AI for the preparation of trial witness statements will not be looked on favourably by the Court. As such, legal representatives and witnesses should refrain from using AI to ‘generate content’ for witness statements.
Building on this sentiment, the Court of Appeal, on hearing a recent criminal matter, went on to address the use of AI in witness training and its impact on witness evidence at trial, stating “The coaching of witnesses in their evidence, by any means, is not permitted. All witnesses, whether for the prosecution or defence, should be firmly discouraged from resorting to the use of AI to prepare for giving evidence. Doing so could run the risk of unfairness up to the point, in an extreme case, where the court may have to consider excluding evidence.” [2] The judgment acknowledged the CJC’s consultation and related work in exploring the use and regulation of AI-use in civil proceedings, and its commentary on witness preparation have an obvious parallel for civil proceedings.
Beyond witness evidence, there are several recent high-profile cases which have highlighted the need for caution when using AI in legal proceedings and legal submissions more widely. In addition to R (Ayinde) v Haringey LBC [2025] EWHC 1383 (Admin) and Taiwo v Homelets of Bath Ltd [2025] EWHC 3173 (KB) addressed in our Legal Updates of December 2025 and January 2026 (linked here and here), in May of this year, law firm Pinsent Masons was identified by the court as relying on AI generated content which included AI hallucinations and then again relying on an AI generated letter explaining the error. Similarly, in April of this year, US firm Sullivan & Cromwell disclosed to a US federal bankruptcy court that a filing they had prepared contained multiple hallucinations.
These continued instances of hallucinations within legal content and lack of transparency around evidence suggest more than teething problems with the legal profession’s use of AI tools, underlining the importance of the CJC’s consultation process, and legitimising the Law Society’s calls for urgent and interim guidance. The recent activity also serves as a reminder to law firms and all legal professionals of the importance of staying informed of developments in AI-governance within the legal profession, and taking a careful approach to tech adoption in a way that aligns with regulatory obligations.
One particular danger of AI use in the legal sphere which clients are grappling with alongside the legal profession is the risk of losing privilege in otherwise legally privileged documents and advice. In a decision earlier this year by the Upper Tribunal of the Immigration and Asylum Chamber[3], the Court commented that ‘uploading confidential documents into an open-source AI tool, such as ChatGPT, is to place this information on the internet in the public domain, and thus to breach client confidentiality and waive legal privilege’. You can read more about this issue in our article Generative AI and Legal Privilege: Risks and Considerations.
[1] Taiwo v Homelets of Bath Ltd [2025] EWHC 3173 (KB)
[2] [2026] EWCA Crim 918
[3] UK v Secretary of State for the Home Department [2026] UKUT 00081 (IAC)
