*/
What to do when confronted with AI-generated pleadings and correspondence? Saara Idelbi and Nyasha Weinberg are here to lend a hand, with some assistance from the Bar Standards Board’s AI Guidance
Ah. The bad pleading. 8pt font? Check. Unexpected capitalisation? Check. Unpredictable highlighting? Check.
Once upon a time, these were generated by unrepresented parties (or *gasp* other barristers) with time on their hands and/or a love for grandiose language.
Today, however, we are increasingly confronted with letters, pleadings or submissions that cite no specific facts, throw every cause of action into the mix, quote authoritative-sounding principles with no basis in law, and conclude with total confidence. Deep inside your mind you know it has likely been AI-generated.
Suddenly, the bad pleading can be generated at scale by anyone with a phone or a laptop and AI has changed the bad pleading from an occasional curiosity to a systemic feature of modern litigation.
In May 2026, the Bar Standards Board issued its Guidance on the Use of Artificial Intelligence and Other Technologies. Barristers are now expected to understand AI’s impact not only on their own practice, but also on the benefits and risks of technologies that may be adopted by opposing parties and clients.
What do you do when you are confronted with AI use? We’re here to lend a hand, with some assistance from the BSB guidance.
The reality is that AI expression changes, in the same way that human expression changes over time or with cultural shifts. The best way to spot AI use is to use AI yourself because AI experienced practitioners develop a feel for the texture and the limitations of AI-generated material.
Some recognisable characteristics are set out below. Now, no single feature is conclusive on its own. We must recall that AI is trained on huge volumes of written word, which means its stylistic conventions also appear in ordinary legal writing and would be natural. Presence may be expected (who does not like the rule of three occasionally), but saturation is atypical. And that may alert you to something more concerning:
What should you do about it when you see it? Stay calm. It is easy to become preoccupied with others’ use of AI, but, much as with poorly drafted pleadings in the past, intervention is only necessary when it has a tangible impact on the proceedings.
You don’t need to prove AI use, and trying to will rarely help. Identify the legal defects instead: fabricated authorities, insufficient particulars, misapplied tests and unpleaded facts.
If the document was prepared by a lawyer, do your opponent the professional courtesy of notifying them of the problems you have identified (not just your suspicion of AI use) and give them the opportunity to correct it. If they do not correct it, then you are entitled to inform the court.
If the document was prepared by an unrepresented party, have your solicitor write to the party to flag the issue, highlight the risks they run with using AI (such as the loss of confidentiality), and – this may outrage some – point them towards legitimate sources, such as the website for the procedural rules or Bailii. Above all, direct them to sources where they can obtain legal assistance: pro bono centres in their area or the Law Society website to find a solicitor. Why? Because you will want the court to see that you did everything you could to assist if and when you knock out the claim and/or seek your costs.
Poor authorities, blended tests and/or misidentified causes of action may expose weaknesses in the underlying case. Consider your formal options:
If you are concerned that the unrepresented party may intend to use AI in a court hearing by use of a microphone, notify the court (copying in the other side) and invite the court to remind all that recordings within an AI application are records for the purpose of contempt of court proceedings.
The tools to deal with AI-generated pleadings exist; you may have to use them more frequently.
We all should be familiar with R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin). If you are not, go and read it! Do not place your whole career on the line because you are pressed for time and think the shortcut is worth it.
As part of a barrister’s core duties, the BSB expects you to maintain a sufficient level of competence in technology and AI. Unexpected for some. A worryingly simplistic issue for others.
There remains no uniform definition of AI. So how can you understand it and the technology?!
Many organisations offer courses on understanding AI and if certification gives you comfort, then it undoubtedly will be worth it. But there are plentiful free resources available to help you learn more about AI.*
Time to put that GDPR training to good use. Read those juicy privacy policies! What protections are offered? Where is the data being processed? How confident are you that your work data is secure?
Get into the habit of data minimisation. Do you need to use full names or obvious details when communicating with others about your case?
If you can strip out personal data from the perspective of the recipient (we will not be getting into the DSG Retail Ltd v The Information Commissioner [2026] EWCA Civ 140 decision for the purposes of this article) then even though you are still processing personal data, you can be working in a way that reduces risks to the data subject and your clients’ confidentiality.
Doctors had their Google moment. Lawyers are having their AI moment.
Explain to your clients the risks they take when uploading their own documents. You may have to remind them that AI is trained on the general and that you are trained on the specific. AI is built for probability and your client is unique. Ultimately you may have to ask them to think about whether they would take a chance on suing the AI model if it made a mistake. And remind them of the one question every person should ask AI after it produces an answer: ‘Have you asked your AI what it got wrong?’
AI may be capable of drafting bad pleadings. It may be capable of drafting good pleadings. For practitioners, all we need to do is to deal with those pleadings, without tears.
Saara will be speaking at the Bar Conference on 14 November 2026 in ‘The IT Crowd: 10 ways to get ahead with AI’.
Saara Idelbi is a barrister at 39 Essex Chambers and co-founder of Advocatr, the AI-advocacy training platform. Her specialisms include public law, human rights/equality, data privacy, and challenges concerning the use of technology. She co-authored A Practical Guide to Non-Pecuniary Damages in Human Rights Act Claims (Law Brief Publishing: 2024).
Nyasha Weinberg practises in public law, regulatory, and commercial/construction law, appearing up to the Supreme Court. She has a substantial inquiries practice, including the Lampard, Infected Blood and Grenfell Inquiries, alongside construction adjudication and arbitration work.
Ah. The bad pleading. 8pt font? Check. Unexpected capitalisation? Check. Unpredictable highlighting? Check.
Once upon a time, these were generated by unrepresented parties (or *gasp* other barristers) with time on their hands and/or a love for grandiose language.
Today, however, we are increasingly confronted with letters, pleadings or submissions that cite no specific facts, throw every cause of action into the mix, quote authoritative-sounding principles with no basis in law, and conclude with total confidence. Deep inside your mind you know it has likely been AI-generated.
Suddenly, the bad pleading can be generated at scale by anyone with a phone or a laptop and AI has changed the bad pleading from an occasional curiosity to a systemic feature of modern litigation.
In May 2026, the Bar Standards Board issued its Guidance on the Use of Artificial Intelligence and Other Technologies. Barristers are now expected to understand AI’s impact not only on their own practice, but also on the benefits and risks of technologies that may be adopted by opposing parties and clients.
What do you do when you are confronted with AI use? We’re here to lend a hand, with some assistance from the BSB guidance.
The reality is that AI expression changes, in the same way that human expression changes over time or with cultural shifts. The best way to spot AI use is to use AI yourself because AI experienced practitioners develop a feel for the texture and the limitations of AI-generated material.
Some recognisable characteristics are set out below. Now, no single feature is conclusive on its own. We must recall that AI is trained on huge volumes of written word, which means its stylistic conventions also appear in ordinary legal writing and would be natural. Presence may be expected (who does not like the rule of three occasionally), but saturation is atypical. And that may alert you to something more concerning:
What should you do about it when you see it? Stay calm. It is easy to become preoccupied with others’ use of AI, but, much as with poorly drafted pleadings in the past, intervention is only necessary when it has a tangible impact on the proceedings.
You don’t need to prove AI use, and trying to will rarely help. Identify the legal defects instead: fabricated authorities, insufficient particulars, misapplied tests and unpleaded facts.
If the document was prepared by a lawyer, do your opponent the professional courtesy of notifying them of the problems you have identified (not just your suspicion of AI use) and give them the opportunity to correct it. If they do not correct it, then you are entitled to inform the court.
If the document was prepared by an unrepresented party, have your solicitor write to the party to flag the issue, highlight the risks they run with using AI (such as the loss of confidentiality), and – this may outrage some – point them towards legitimate sources, such as the website for the procedural rules or Bailii. Above all, direct them to sources where they can obtain legal assistance: pro bono centres in their area or the Law Society website to find a solicitor. Why? Because you will want the court to see that you did everything you could to assist if and when you knock out the claim and/or seek your costs.
Poor authorities, blended tests and/or misidentified causes of action may expose weaknesses in the underlying case. Consider your formal options:
If you are concerned that the unrepresented party may intend to use AI in a court hearing by use of a microphone, notify the court (copying in the other side) and invite the court to remind all that recordings within an AI application are records for the purpose of contempt of court proceedings.
The tools to deal with AI-generated pleadings exist; you may have to use them more frequently.
We all should be familiar with R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin). If you are not, go and read it! Do not place your whole career on the line because you are pressed for time and think the shortcut is worth it.
As part of a barrister’s core duties, the BSB expects you to maintain a sufficient level of competence in technology and AI. Unexpected for some. A worryingly simplistic issue for others.
There remains no uniform definition of AI. So how can you understand it and the technology?!
Many organisations offer courses on understanding AI and if certification gives you comfort, then it undoubtedly will be worth it. But there are plentiful free resources available to help you learn more about AI.*
Time to put that GDPR training to good use. Read those juicy privacy policies! What protections are offered? Where is the data being processed? How confident are you that your work data is secure?
Get into the habit of data minimisation. Do you need to use full names or obvious details when communicating with others about your case?
If you can strip out personal data from the perspective of the recipient (we will not be getting into the DSG Retail Ltd v The Information Commissioner [2026] EWCA Civ 140 decision for the purposes of this article) then even though you are still processing personal data, you can be working in a way that reduces risks to the data subject and your clients’ confidentiality.
Doctors had their Google moment. Lawyers are having their AI moment.
Explain to your clients the risks they take when uploading their own documents. You may have to remind them that AI is trained on the general and that you are trained on the specific. AI is built for probability and your client is unique. Ultimately you may have to ask them to think about whether they would take a chance on suing the AI model if it made a mistake. And remind them of the one question every person should ask AI after it produces an answer: ‘Have you asked your AI what it got wrong?’
AI may be capable of drafting bad pleadings. It may be capable of drafting good pleadings. For practitioners, all we need to do is to deal with those pleadings, without tears.
Saara will be speaking at the Bar Conference on 14 November 2026 in ‘The IT Crowd: 10 ways to get ahead with AI’.
What to do when confronted with AI-generated pleadings and correspondence? Saara Idelbi and Nyasha Weinberg are here to lend a hand, with some assistance from the Bar Standards Board’s AI Guidance
The Chair of the Bar reports back
By Kem Kemal, CEO, Henry Dannell
By BSK Financial Planning
Mário Barroso, Head of R&D and Method Development at AlphaBiolabs
Susie Griffin on the admin that keeps barristers from billable work
Clement Cowley, Partner at The Penny Group, explains how global events can influence the performance of pensions and ISAs, and why taking a long-term approach to investing is often the most effective strategy
Kick off your career at the Bar Council’s Pupillage Fair this October – it’s one of the most valuable opportunities for aspiring barristers, says Amelia Clegg
The golden age of advocacy? David Langwallner pays homage to the filmic if contentious barrister Sir Edward Carson and some of his contemporaries
Henry Fingerhut outlines what the Bar Standards Board AI guidance means for early career barristers
The Ministry of Justice will match-fund up to 100 criminal pupillages, with grants available to criminal chambers across England and Wales. Similar to the COIC scheme, the grants will match-fund pupillages additional to those otherwise offered without government funding
Not yet, says Jacob Turner. Regulation, unlike the technology itself, is a domain the legal profession can still control
