AI and Your Court Documents: The Assumption vs the Reality

A recent Employment Appeal Tribunal ruling gives us a clear look at how courts are responding to AI-drafted documents from litigants in person. It is worth breaking down, because the gap between what people assume and what the court actually expects is where people get caught out.

The case

A litigant in person filed a skeleton argument running to 300 pages and 132,000 words, created using ChatGPT. He asked the other side to flag any inaccuracies by 4pm the following day. A deputy High Court judge, Marcus Pilgerstorfer KC, made it clear this was not acceptable.

Assumption: If I use AI to draft it, I've done my part.

Reality: The judge said directly that simply submitting the product of generative AI to a court, or placing the onus to check it onto an opponent, is not acceptable. Using the tool is not the finish line. Checking what it produced is the actual work, and it cannot be skipped or handed to someone else.

Assumption: AI is banned in court documents, so if I use it I'm already at risk.

Reality: The judge confirmed the opposite. Using AI to prepare documents like skeleton arguments is not generally prohibited. The risk isn't in using the tool. The risk is in submitting what it produces without checking it.

Assumption: If a lawyer would be blamed for an AI mistake, a litigant in person won't be, since I'm not a professional.

Reality: The judge was explicit that the underlying principles of personal responsibility and accuracy apply to all persons who submit documents to a court or tribunal, not just those with professional duties. Whether you're a solicitor or representing yourself, your name is on what you file, and you are accountable for it.

Assumption: A longer, more detailed document looks more thorough and helps my case.

Reality: The judge said documents should contain only relevant points, with the focus on the central or best arguments, presented clearly and without undue repetition. A 300-page AI-generated skeleton is not thoroughness. It is a sign the document was never properly reviewed, and the judge indicated a document like it is likely to be rejected outright going forward.

Assumption: If the case law and facts in an AI draft look right, they probably are right.

Reality: AI tools can generate confident-sounding but incorrect case citations, misstate legal tests, or get facts wrong. The judge's guidance requires that every factual, evidential and legal point, including references to authorities, be checked and correct before submission. Confidence in the writing is not the same as accuracy in the content.

The point that cannot be stressed enough

Whatever tool you use to prepare a document, whatever your McKenzie Friend or anyone else has helped you draft, the responsibility for what goes to the court sits with you. It cannot be delegated to AI, and it cannot be delegated to the person on the other side of the case. If a document is inaccurate, incomplete, or misleading, that comes back to the litigant who filed it, not the software that helped write it.

What to actually do

• Treat AI output as a first draft, never a final one
• Check every case reference against the original source
• Read the whole document yourself before it goes anywhere near the court
• Keep it focused on your strongest points rather than everything AI can generate
• Get a second pair of eyes on it, whether that's a McKenzie Friend or someone else with relevant experience

This information is general and does not constitute legal advice. If you're unsure how AI use might affect your specific case, speak to a family law professional.