Can You Use ChatGPT to Prepare Court Documents? What the CJC’s 2026 AI Report Means for Litigants in Person

JSH LAW · COURT DOCUMENTS & AI
By Jessica Susan Hill · Published 2 November 2026 · England and Wales

Can you use ChatGPT to help prepare court documents? It can explain unfamiliar procedures, help organise evidence and make a chronology easier to follow. But it can also invent case law, misstate facts or subtly change the meaning of a witness’s account.

On 29 October 2026, the Civil Justice Council (CJC) published its Use of AI for Preparing Court Documents: Final Report. The 50-page report contains 15 recommendations on the responsible use of artificial intelligence in civil proceedings in England and Wales.

For people representing themselves, the report raises a central question: when AI helps prepare a document, who checks that it remains accurate, authentic and appropriate to put before the court?

1. What has the Civil Justice Council recommended?

The CJC does not propose a blanket ban on AI. Instead, it recommends a functional, risk-based approach focused on whether an AI tool generates or influences substantive content and what risks arise for the integrity of proceedings.

• Litigants in person: the CJC and Civil Procedure Rule Committee (CPRC) should urgently consider appropriate guidance for unrepresented people using AI in civil proceedings (Recommendation 2).
• Clear rules: the CPRC should consolidate the courts’ expectations and sanctions for AI misuse in a plain-language Practice Direction (Recommendation 3).
• Legal arguments: no general new AI-use declaration is proposed for legal representatives preparing statements of case, skeleton arguments or similar advocacy documents. They remain responsible for the accuracy of their work (Recommendation 4).
• Witness statements: the CPR should be amended to require factual witnesses to declare that AI has not generated or influenced the substantive content of their statements (Recommendation 6).
• Expert evidence: the report proposes disclosure where AI generated or influenced the substance of an expert’s report, so that the method and conclusions can be scrutinised (Recommendation 9).
• Translations: the CJC recommends continued professional human certification and that AI-assisted translations be supplied with their original text (Recommendation 8).

Important: these are recommendations for possible rule changes. The report itself did not introduce a new compulsory AI declaration, and it should not be reported as if it did.

2. Witness statements: protect the witness’s own evidence

A witness statement is evidence. It is not simply a chance to produce the most persuasive version of a story. The facts must reflect what the witness actually knows or genuinely recalls, with uncertainty preserved.

Suppose a parent writes, “I cannot remember whether the conversation was on Monday or Tuesday. I felt frightened.” AI might rewrite that into a definitive account of a Monday incident, with stronger language and an implied conclusion about the other parent’s intentions. It may sound more polished while being less accurate.

The CJC’s proposed declaration is aimed at that danger: protecting the substance of the witness’s evidence. It is not a proposal to ban every AI-enabled function, such as accurate transcription or formatting that does not alter content.

3. Can AI help with a position statement or legal argument?

AI can help organise submissions, suggest headings, identify questions requiring research and explain procedural terms. It cannot be treated as a reliable legal authority in itself.

One serious risk is a fictitious or misrepresented case citation. In Ayinde v London Borough of Haringey; Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin), the Divisional Court considered cases arising from actual or suspected use of unverified AI-generated material placed before the court.

Before relying on a judgment, find the original decision, check its neutral citation, read the passages that supposedly support your argument and establish whether it applies to the correct proceedings. A real case with the wrong interpretation can be just as misleading as an invented case.

For a practical framework, see JSH Law’s Seven-Point Legal AI Trust Test.

4. What does the report mean for family proceedings?

The distinction between civil and family procedure matters. The CJC’s proposed amendments concern proceedings under the Civil Procedure Rules (CPR). Most family proceedings are governed by the Family Procedure Rules 2010 (FPR). The CJC’s recommendations do not automatically change the FPR.

Nevertheless, the risks are relevant to family litigation. Practice Direction 22A, paragraph 4.1, already requires family witness statements, if practicable, to be in the witness’s own words; paragraph 4.3 requires a distinction between personal knowledge and information or belief, with sources identified.

Under FPR Part 17, witness statements must be verified by a statement of truth. Rule 17.6 provides for possible contempt proceedings where false statements are made without an honest belief in their truth.

In D (A Child) (Recusal) [2025] EWCA Civ 1570, the Court of Appeal addressed problems arising from erroneous citations of authority in family proceedings. The judgment reinforces the need to verify legal sources instead of placing untested AI answers before the court.

In children proceedings, special care is necessary with disputed allegations, professional assessments, the child’s expressed wishes and feelings and any findings already made. A system must not describe disputed assertions as if the judge has proved them.

5. What about confidential court documents?

Family case files can contain children’s names, school records, medical information, safeguarding allegations, police records, financial disclosure and private correspondence. An AI product accepting a document does not establish that uploading it is lawful or appropriate.

For certain private children proceedings, communication of information is governed by provisions including FPR rule 12.73 and Practice Direction 12G. They are not a blanket licence to transmit confidential case records to an external AI provider.

Before entering information, consider the court’s directions, the recipient, the purpose of processing, retention and training settings, the rights of other people named in the file and whether data minimisation can achieve the task. Replacing names with initials does not necessarily remove all identifying information.

6. Seven precautions before using AI-assisted court documents

1. Identify the document type. A chronology, position statement, witness statement, expert report and legal submission do not all carry the same requirements.
2. Preserve original evidence. Keep the emails, messages, documents and source records in their original form.
3. Check every important fact. Verify dates, quotations, names and numbers against reliable source material.
4. Keep witness evidence authentic. Do not let AI embellish memories, introduce facts or erase uncertainty.
5. Verify every legal proposition. Read the current legislation, rule or judgment rather than relying on an AI summary.
6. Protect confidential information. Check whether a specific upload is permissible before supplying case material.
7. Review before filing. Confirm the latest directions, deadlines, page limits, statement of truth and the final document you will sign or submit.

Our existing, longer guide explains the workflow: How to Use ChatGPT as a Litigant in Person: A Practical Guide to Preparing Your Court Case. It covers case organisation, chronologies, document audits, confidentiality, legal research and quality control.

7. What should happen next?

The CJC has highlighted an important access-to-justice issue. Many unrepresented parties may find that AI provides useful help with burdensome paperwork. But if professional users receive detailed safeguards while litigants in person receive no accessible explanation of the risks, the technology could also deepen procedural disadvantage.

Useful guidance should address the tasks people actually undertake, not simply list technology brands. The key questions are whether AI has affected facts or arguments, whether the result can be verified, and who remains responsible for it.

JSH Law’s position: AI can assist people preparing their cases. It must not become the author of their facts, a substitute for independent verification or a reason to compromise confidential evidence.

For the wider regulatory context, see our new analysis of Australia’s proposed frontier AI safety regulator and what the UK can learn about accountability. It compares enforceable duties, incident reporting and the safeguards needed where AI is used in legal proceedings.

Primary sources and further reading

• CJC, Use of AI for Preparing Court Documents: Final Report, 29 October 2026
• Family Procedure Rules, Practice Direction 22A
• Family Procedure Rules, Part 17
• Ayinde and Al-Haroun [2025] EWHC 1383 (Admin)
• D (A Child) (Recusal) [2025] EWCA Civ 1570

Reviewed 2 November 2026. This article provides general information rather than individual legal advice. JSH Law Ltd provides non-reserved legal support and is not authorised or regulated by the Solicitors Regulation Authority.