A judge ruled this month that the accuracy principles from Ayinde, the case that set the standard for AI use by solicitors, apply just as much to someone with no legal training at all. If a document ends up in front of a tribunal, “I didn’t know” is no longer a defence, whoever drafted it.
What actually happened
Marcus Pilgerstorfer KC, a deputy High Court judge sitting in the Employment Appeal Tribunal, ruled on 17 September 2026 on an appeal brought by Terence Hancox, a litigant in person who filed a 300-page, 132,000-word skeleton argument said to have been created using ChatGPT, in an appeal against a 2021 strike-out decision.
The judge was direct. Simply submitting the product of generative AI to a tribunal, or placing the onus to check it onto an opponent, is not acceptable. He said the personal responsibility and accuracy principles set out in Ayinde, which concerned solicitors, apply equally to litigants in person, who must still ensure documents comply with procedural rules, that facts, evidence and legal citations are correct, and that only relevant points are included. He found one ground on which Mr Hancox could appeal, but issued case management directions to stop the case taking up disproportionate tribunal resource.
Why this matters beyond litigants in person
This is not really a story about people representing themselves. It is confirmation of where the line actually sits: the accuracy standard is scoped to whoever puts a document in front of a court, not to whether that person happens to hold a practising certificate.
That is the same principle CPR Part 32, PD 18.1(5) already sets for witness statements, which we have written about before, and it is the same standard a claimant firm has always had to meet when a document goes on the record. It is also the practical lesson of Wiltshire v Aioi. What this ruling adds is a court explicitly saying the standard does not soften just because the drafter is unqualified or the tool is doing the heavy lifting. If anything, it hardens the argument that a firm using AI to draft, with proper checks and a qualified person signing off, is the safer position, not a risky one. That is exactly how our witness statement generator is built to work.
What is changing on the vendor side at the same time
The same pressure is building on the procurement side, not just the drafting side. A guest piece on Legal Futures this month argued that AI regulatory risk for law firms is actually created at procurement stage, not at the point of use. That lines up with what the Ayinde-line cases have been pointing to all year: the responsibility sits with whoever chooses and checks the tool, so the choosing has to be defensible too. We set out the five questions worth putting to any AI vendor before you sign in a separate guide: Five Things You Should Ask Before Buying an AI Product for Your Firm.
Separately, the annual Crowe and Institute of Legal Finance & Management survey of 39 firms with turnovers between £1.7m and £69m found 79% have now committed specific funds to AI-based solutions. This is not a hypothetical question for large City firms. It is a live procurement decision for the regional and claimant firms this survey covers too.
What a firm using AI to draft should actually do
Verify every citation and case reference before it is filed, not after. Keep a record of who checked what, and when, so “the AI produced it” is never the whole answer to a question about a document’s accuracy. And before adopting any AI drafting tool in the first place, put it through the five procurement questions above, not just whether it produces a good-looking document.
FAQ
Does the Ayinde ruling apply to litigants in person?
As of 17 September 2026, yes. A deputy High Court judge sitting in the Employment Appeal Tribunal ruled that the personal-responsibility and accuracy principles set out in Ayinde, which concerned solicitors, apply equally to litigants in person who use generative AI to draft court documents.
What did the judge say about AI-drafted court documents?
That simply submitting the product of generative AI to a tribunal, or placing the burden of checking it onto an opponent, is not acceptable, and that judicial enquiries and sanctions are likely where proper checks have not been done.
Is this new SRA guidance?
No. The 17 September ruling is a tribunal decision, not SRA guidance. It is worth acting on, but should not be presented as settled SRA policy.
Craig Budsworth (FCILEx) writes on AI and legal process for LegalDocs Assist.
LegalDocs Assist drafts from your own case papers, with every authority drawn from a curated UK knowledge base and a fee earner signing off. See it in action or book a demo.
© LegalDocs Assist. www.legaldocs-assist.co.uk