ChatGPT can be extraordinarily helpful to a litigant in person. It can explain unfamiliar language, bring order to a disorganised case file, identify gaps, create a chronology and help you express your case clearly. What it cannot do is relieve you of responsibility for the facts, the law or anything you put before the court.

Representing yourself is difficult for reasons that have little to do with intelligence. You may be dealing with a frightening court process at the same time as separation, financial pressure, allegations, housing problems or uncertainty about your children. You are expected to absorb unfamiliar rules, understand court orders, preserve evidence, meet deadlines and present the relevant parts of your case calmly.

That is precisely where ChatGPT can help. Used properly, it can operate as a patient case-management and drafting assistant: available when you need to work, able to explain the same point more than once and capable of turning a mass of documents into a structured plan.

Used carelessly, however, it can confidently invent a case, misstate a rule, change the meaning of your evidence or produce a polished document that is simply wrong. The answer is not to reject the technology. It is to use it for the right tasks, set it up carefully and verify everything that matters.

What is a litigant in person?

A litigant in person is someone who conducts court proceedings without a solicitor or barrister acting for them. You have the right to represent yourself, although you can still obtain legal advice on particular issues or instruct a lawyer for part of the case. The Government’s guidance confirms both the right to speak for yourself and the continued option of obtaining legal advice.

This guide is written principally for civil and private family proceedings in England and Wales. Different courts, tribunals and types of case have different rules. Criminal proceedings, appeals, judicial review, public law children cases and complex financial claims require particular care. Scotland and Northern Ireland have separate systems.

What ChatGPT can—and cannot—do for your case

The best description is litigation support assistant. ChatGPT can help you do the work more systematically, but you remain the person conducting the case.

Why you should create a dedicated ChatGPT Project

Do not run an entire case through one ordinary, ever-lengthening chat. Create a dedicated Project for that one case. Projects are designed to keep relevant chats, uploaded files and project-specific instructions together. ChatGPT can then work with the context inside the Project as the case develops.

When creating it, choose project-only memory if that option is available. That keeps the Project’s context within the case rather than drawing upon conversations outside it. OpenAI’s current guidance says project-only memory must be selected when a new Project is created; an existing Project cannot simply be converted afterwards.

Before uploading anything: privacy, confidentiality and redaction

This is the step too many people skip. Court documents may contain personal data, confidential information, medical material, financial records and information about children. Family proceedings involving children are also subject to specific restrictions on communicating information from the proceedings. Do not assume that information may be uploaded to a consumer AI service merely because you are using it privately.

Change your ChatGPT data setting

On a personal ChatGPT account, go to Settings > Data Controls and turn off “Improve the model for everyone”. OpenAI states that new conversations will then remain in your history but will not be used to train ChatGPT. This is sensible, but it is not the same as a legal assurance that every document is appropriate to upload.

Use data minimisation

Upload only what is reasonably necessary for the task. Work from redacted copies and retain the untouched originals separately. Depending on the case, consider removing:

• children’s full names, dates of birth, schools and addresses;
• home addresses, telephone numbers and private email addresses;
• bank details, signatures, passport numbers and official identifiers;
• medical identifiers or medical history unrelated to the issue;
• unnecessary information about third parties; and
• legally privileged or confidential material you may not be entitled to disclose.

Use a proper PDF-redaction tool. Drawing a black rectangle over text may leave the underlying words recoverable. Never redact your only copy of a document and never alter an original piece of evidence.

Project files are intended to persist. OpenAI’s current retention guidance says files uploaded to a Project are retained until the Project is deleted and are then scheduled for removal within 30 days, subject to limited legal or security exceptions. Keep your own case record and delete the Project when it is no longer properly required.

How to set up your litigation Project from the beginning

1Create a neutral, private Project

Select New Project in ChatGPT’s sidebar. Give it a neutral name, such as:

Avoid placing a child’s full name or highly sensitive allegation in the title. Select project-only memory at creation and do not share the Project.

2Add strong Project instructions

Project instructions matter because they tell ChatGPT how to behave throughout the case. Open the Project menu, select Project settings and adapt the following:

3Keep a separate master file outside ChatGPT

Your ChatGPT Project is a working tool, not the official home of your case. Maintain a separate, backed-up folder containing the documents exactly as received or filed. A simple structure is:

1. 00 – Master index and case dashboard
2. 01 – Applications and statements of case
3. 02 – Sealed court orders and directions
4. 03 – Correspondence
5. 04 – Evidence and exhibits
6. 05 – Witness statements
7. 06 – Law and verified authorities
8. 07 – Hearing documents and bundles
9. 08 – Drafts and superseded versions

Name files consistently, for example:

Do not overwrite filed documents. Mark versions clearly: DRAFT v1, DRAFT v2 and FINAL FILED 18-07-2026. Keep screenshots and message exports in their original form with dates, context and metadata wherever possible.

4Upload the foundation documents first

Do not begin by tipping hundreds of files into the Project. Start with the documents that define the proceedings:

1. the current application, claim, defence or response;
2. all current sealed court orders, beginning with the most recent;
3. the parties’ principal statements or pleaded positions;
4. the key evidence needed to understand the immediate issues; and
5. any applicable court guidance or directions supplied specifically for the hearing.

Make sure scanned PDFs are searchable and legible. If a document contains handwritten notes, missing pages or poor scans, tell ChatGPT not to guess.

5Ask for a document audit before analysis

Check the resulting inventory yourself. If ChatGPT has misunderstood a file, correct the record immediately.

Create the five working documents that control the case

A well-run case is not a pile of evidence. It is a structured explanation of the issues, the relevant facts, the evidence supporting those facts, the applicable law and the order the court is being asked to make. These five living documents provide that structure.

Prompt for a sourced chronology

Prompt for an issues-and-evidence matrix

Prompt for court orders and deadlines

Use separate chats for separate jobs

Within the Project, create clearly named chats rather than allowing one conversation to become unmanageable:

• 01 – Case dashboard and directions
• 02 – Document inventory
• 03 – Chronology
• 04 – Issues and evidence
• 05 – Legal research
• 06 – Witness statement drafting
• 07 – Position statement and order sought
• 08 – Correspondence
• 09 – Hearing preparation

This makes it easier to find the reasoning behind a draft and to start again when a conversation has gone in the wrong direction. Save only checked, durable outputs back to the Project sources.

Legal research: the area of greatest danger

ChatGPT can help you formulate a research question and locate potential sources. It must not be trusted to supply legal authorities from memory.

In Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin), the Divisional Court addressed false authorities and inaccurate material placed before the court. The judgment stated plainly that large language models such as ChatGPT are not capable of conducting reliable legal research without proper checking. In one of the matters, citations included a non-existent case while the neutral citation belonged to an entirely unrelated decision.

The risk is not theoretical. AI can produce a case name, citation and quotation that look entirely convincing. The more polished the answer, the easier it is to lower your guard.

A safer legal-research protocol

1. Ask ChatGPT to identify the precise legal question.
2. Ask it to search official primary sources, not merely generate an answer.
3. Open the source yourself.
4. Confirm that the case or rule exists and is current.
5. Read the relevant paragraph in context.
6. Check that the authority actually supports the proposition.
7. Check whether it has been appealed, amended, distinguished or superseded.
8. Record the verified source and the date you checked it.

For England and Wales, begin with official sources such as legislation.gov.uk, the official Civil and Family Procedure Rules, and Find Case Law at The National Archives.

Drafting a witness statement without changing the evidence

A witness statement is evidence. It is not an opportunity for ChatGPT to create the most compelling version of events. Your evidence must remain your own recollection, expressed accurately and with any genuine uncertainty preserved.

AI may help with structure, grammar, repetition and chronology. It must not:

• add a fact you did not provide;
• turn an approximate memory into a precise date or quotation;
• convert suspicion or inference into direct knowledge;
• embellish an event to make it sound more serious;
• remove qualifications that are important to accuracy;
• describe another person’s motive as fact; or
• insert legal argument into factual evidence without good reason.

A safer method is to prepare your own factual account first. Then ask ChatGPT to identify gaps and ask neutral, non-leading questions. Only after you answer those questions should it help organise the material.

Position statements, applications and written submissions

ChatGPT is often very good at turning long, emotional notes into a shorter and more usable structure. That is valuable, particularly in family proceedings where a parent may understandably want to explain everything that has happened. The discipline is to retain what helps the judge decide the issues and remove what merely expresses anger.

Before drafting, give ChatGPT:

• the type and purpose of the hearing;
• the order listing it and any specific directions;
• the live issues;
• what has changed since the last hearing;
• the evidence relied upon;
• the other party’s position, stated fairly;
• the precise orders sought; and
• any page limit or deadline.

Ask ChatGPT to test your case, not simply agree with you

AI assistants are naturally responsive to the way a question is framed. If you ask for arguments proving that the other party is unreasonable, you are likely to receive them. That can reinforce a mistaken or one-sided view.

Use prompts that force a neutral review:

Sometimes the most helpful answer is that a point is irrelevant, the evidence is weak or the order sought is unrealistic. That is useful preparation, not failure.

Comparing statements and messages properly

ChatGPT can compare long documents much faster than most people, but it should not be invited to label every difference a lie. Memory differs, wording changes and two people can genuinely perceive an event differently.

For message evidence, preserve the full conversation and original export. A selected screenshot may hide context. Ask ChatGPT to identify the message immediately before and after any passage you intend to rely upon.

Preparing for the hearing

ChatGPT can be particularly helpful in the final week, when anxiety makes it difficult to see the case clearly. Ask it to prepare:

• a one-page hearing plan;
• a 30-second introduction explaining who you are and what the hearing concerns;
• the three most important points, each linked to evidence;
• a numbered list of the exact orders you seek;
• likely questions from the judge and truthful, concise answers;
• the weaknesses the other party is likely to raise;
• questions for another witness, phrased as questions rather than speeches;
• a list of documents and page references you need at hand; and
• a practical checklist covering time, location, technology and papers.

Prepare an offline copy of your key documents and speaking notes. Do not assume ChatGPT, internet access or your telephone will be available or permitted during the hearing.

A sensible quality-control check before anything is filed

Every AI-assisted document should pass three separate checks.

1. The factual check

• Is every fact accurate?
• Is an allegation clearly identified as an allegation?
• Can every important assertion be traced to evidence?
• Are dates, names, quotations and page references correct?
• Has the wording changed the meaning of your evidence?

2. The legal and procedural check

• Does every cited case exist on an authoritative database?
• Does the relevant paragraph support the proposition?
• Is the legislation or procedural rule current?
• Does the document comply with the latest order, page limit and deadline?
• Is the correct statement of truth or other required wording included?

3. The strategic and human check

• Does the document answer the question the court must decide?
• Is it proportionate and concise?
• Does it acknowledge material evidence that goes against you?
• Is the order sought precise, realistic and lawful?
• Does the language sound like you and remain something you can honestly adopt?

The mistakes to avoid

• Do not upload the entire unredacted case file without considering confidentiality and necessity.
• Do not ask ChatGPT to “find cases that prove I am right” and paste the result into submissions.
• Do not let it turn allegations into established facts.
• Do not ask it to make your evidence sound stronger or more dramatic.
• Do not rely on an AI-generated deadline, court form or procedural rule without verification.
• Do not use one enormous chat for every aspect of the proceedings.
• Do not replace original evidence with an AI summary.
• Do not file a document you have not read, checked and fully understood.
• Do not use AI in the hearing to record, transcribe or transmit proceedings without express permission.
• Do not mistake fluent language for sound legal analysis.

Used well, ChatGPT can improve access to justice

The justice system requires litigants to present relevant facts and evidence within formal procedures. For a person without representation, the practical barrier is often not the absence of a genuine case but the difficulty of organising and communicating it.

ChatGPT can reduce that barrier. It can make an order easier to understand, expose gaps before a hearing, help a distressed parent write calmly, and turn a chaotic file into a chronology and evidence matrix. It can also help a litigant see that a favourite argument is weak, that a document does not prove what they thought it proved, or that compromise may be the better course.

That is real value. But the value comes from disciplined use. The litigant must remain the source of the evidence, the decision-maker about their case and the person responsible for every word put before the court.