Join early access
How it works Features Guides Free tools The story Privacy Pricing FAQ For lawyers Law firms Community & legal aid Compliance Join early access

Using AI for your case? Check what it tells you.

Many people now ask ChatGPT, Claude, Gemini or Copilot about their case, and courts accept that. But these tools can invent cases and quotes, mix up countries, and keep what you paste in. Courts in all six places this guide covers say the same thing: you are responsible for what you put before the court, so check every point yourself.

8 minute readFor England & Wales, Ireland, Australia and New Zealand, with California and Ontario as US and Canadian examplesLast checked 4 October 2026

If a child or anyone is in danger now, call 999 (UK), 112 (Ireland), 000 (Australia), 911 (US and Canada) or 111 (New Zealand).

Step by step

Using AI without getting caught out.

Tap each step. Then choose where you live to see what the courts there say.

General information. Every court is different.

Where you live

General information, checked against official sources on 4 October 2026. It isn’t advice about your case. The rules in their own words ↓

Check the settings before you paste.

Your papers and the tools

What happens to what you upload.

From each company’s own privacy pages, as of 4 October 2026, for personal (not business) accounts. Settings change: check the current page.

  1. ChatGPT.Chats may be used for training unless you turn off “Improve the model for everyone”. Temporary Chats aren’t used for training. Deleted chats are removed within 30 days, unless OpenAI must keep them for legal reasons, as a court order required in 2025.
  2. Claude.Since August 2025 you choose whether chats help train it (the “Model Improvement” setting). If allowed, chats can be kept up to five years; if not, 30 days. Incognito chats are never used for training.
  3. Gemini.With “Keep Activity” on (the default), activity is kept for 18 months and a sample of chats is read by human reviewers. Google asks you not to enter confidential information. Temporary chats are kept for 72 hours.
  4. Gemini Notebook (formerly NotebookLM).Answers from the documents you upload, with citations to them. For personal accounts Google says your data isn’t used for training unless you give feedback. It still can’t tell you whether the law in your papers is current.
  5. Copilot.Conversations may be used for training unless you opt out, are kept for 18 months by default, and some are reviewed by people.
  6. Perplexity.Searches the web and shows sources, which makes checking easier, but its privacy notice says data may improve its AI models. Open each source yourself.
  7. Whatever the tool.Don’t paste anything a court has ordered kept private. Remove or replace names, addresses and your child’s details first.

Companies change these settings often. This summary isn’t advice about any tool; read the current privacy page before you upload papers.

Free tool

Before you rely on it: a checklist.

Tick off each check before an AI answer goes anywhere near the court. Your ticks stay in this browser; nothing is sent anywhere.

A general list, not advice for your country or your case. Courts can have their own rules about AI: check yours.

Writing your own statement? How to put your side on paper →

See it in the app

Answers from your papers, with the page to check.

  1. Add your papers.Their statement, your messages, the letters. Before the language model reads a word, names become tags.
  2. It shows where your papers disagree.Two records that can’t both be right, each quoted with its page, and the question only you can answer: which is right?
  3. It quotes the law where you live. Coming in early accessThe sections that apply to your case, in the law’s own words, with a link to the official text. Change the country above and it changes too.
  4. Every point keeps its page.What it prepares links each point to the paper it rests on, so you can open the page and check it yourself.

Sample case. Names and papers are fictional.

What the rules say

What the courts say, in their own words.

Five questions about using AI on a court case, answered only from court rules, practice directions, official guidance and judgments, with the passage that answers it highlighted.

Where you live

Quoted from the official text. It isn’t advice about your case, and courts are updating these rules fast: check the official site. Individual courts and judges can have their own orders. California and Ontario are shown as US and Canadian examples.

Questions people ask

The questions people ask about AI.

Is it allowed to use ChatGPT or another AI for my case?

Generally yes: courts in all six places this guide covers accept that people use AI, and some have written guidance for it. What they all require is that you check what it gives you and take responsibility for what you file. See “What the courts say” for where you live.

Can AI really make up cases?

Yes. Chatbots can invent cases, citations and quotes that look real. A public database of court decisions about AI errors listed 2,145 decisions on 4 October 2026, more than half involving people representing themselves. Courts have rejected documents, ordered costs and imposed penalties.

Is it safe to upload my court papers?

It depends on the tool, its settings and your case. Family cases about children are confidential in most places, and what you upload may be kept, reviewed by people or used for training. Check the settings, remove names and identifying details, and never paste anything a court has ordered kept private.

Do I have to tell the court I used AI?

Usually not unless the court asks, but there are exceptions: in Ireland’s High Court, affidavits and witness statements must now include a declaration that AI didn’t generate their substance, and some courts have their own orders. Be ready to explain what you used and how you checked it.

Can AI write my statement or affidavit?

Your statement is your evidence, so it should be in your own words. Some courts now forbid AI-generated content in affidavits and witness statements. AI can help you put your facts in date order or spot what’s missing; the words should be yours.

Which AI is best for legal questions?

None of them is a lawyer, and none knows your court or your papers unless you give them. Tools that show their sources, or answer only from documents you upload, make checking easier, but you still have to open the source and read it.

How is PROVEAiBLE different?

Each case is kept on its own. When PROVEAiBLE works on a case, the model is given only that case’s papers and your answers in it — nothing from your other cases, and nothing from anyone else’s. There is no memory that follows you between cases. Names become tags before the language model reads your papers, and what it tells you comes with the passage it rests on, so you can open the page and check it. Like any AI, it can be wrong: that’s why the page is always there.

You’re not alone

You don’t have to do this alone.

AI can help you get organised, but it can’t take responsibility for your case. Free and low-cost legal help exists: check whether you qualify, and get advice on the decisions that matter.

Free and low-cost legal help
    Someone to talk to, any time
      If anyone is in danger

        Showing help for England & Wales. Change it with “Where you live” above.

        “I built this after going through a court case myself.” — Andrew, founder

        With a lawyer or without one, walk in with your papers in order.

        PROVEAiBLE reads your papers, sets out the facts with their pages, shows where the papers disagree, quotes the law for where you live, and prepares your papers for the hearing. It never tells you which legal step to take, or how it will turn out. Law for more countries: coming in early access

        Join early accessFree tools →More guides →
        Sources, checked 4 October 2026

        General information, not legal advice. Laws, rules and court processes change; check the official pages and your own papers before you act.