ChamberOne

UK: judicial guidance on AI for lawyers appearing in court

By the ChamberOne team 5 minute read

The judiciary of England and Wales has published guidance on artificial intelligence. It is addressed to judicial office holders and their support staff, not to advocates. Even so, it tells barristers and solicitor advocates a good deal about how the bench thinks about these tools, and the courts have separately said, in strong terms, what happens when unchecked AI output reaches a court.

Key points

  • The judiciary's AI guidance is addressed to judicial office holders, but it shows how the bench views confidentiality, accuracy and responsibility.
  • Avoid entering private or privileged material into public AI tools, and check the terms of any tool you use.
  • Open and read every authority before citing it; a fluent citation is not evidence that a case exists.
  • Regulators and the courts have said that relying on AI output is not a defence for false citations.
  • Read the current judicial guidance, the judgment and your own regulator's guidance rather than relying on summaries.

Who the guidance is for, and why advocates should read it

The guidance published on the judiciary website applies to judicial office holders for whom the Lady Chief Justice and the Senior President of Tribunals are responsible, and to their clerks, judicial assistants, legal advisers and other support staff. The version on the website at the time of writing is dated October 2025 and replaces an earlier version.

It is not guidance to advocates, and this article does not suggest otherwise. But if you are going to appear before a judge who has been told to be aware of AI-related risks, it helps to know what that judge has been told. Check the judiciary website for the current version, as the guidance has been updated more than once.

Confidentiality: what goes into public tools

The guidance reminds judicial office holders not to enter private information into public AI tools, and it points to where to report inadvertent disclosures as data incidents. The principle is one that advocates can easily apply to their own practice.

For a barrister or solicitor advocate, that means thinking about everything that goes into a tool while preparing a case: instructions, witness statements, draft skeleton arguments, disclosure documents. Public tools may keep or use what is entered. Check the terms of any tool you use, and the guidance from your regulator and your Inn or professional body about confidentiality and technology.

Accuracy and the problem of invented material

The guidance draws attention to the risk that AI tools produce information that is inaccurate, incomplete or misleading, and that they can invent cases and quotations. This is no surprise to anyone who has tried a general-purpose chatbot on a legal question, but it is notable that the point is made in judicial guidance.

The practical consequence for advocates is that a judge may be sensitive to authorities that do not look right. A citation that cannot be found, a quotation that does not appear in the judgment or a proposition that the case does not support are all more likely to be noticed than before.

Responsibility stays with the person

The foreword to the guidance emphasises the personal responsibility of judicial office holders for all material produced in their name. The same thought applies in reverse to advocates. A skeleton argument is the work of the advocate whose name is on it, whoever or whatever drafted the first version.

The Solicitors Regulation Authority has published guidance on the misuse of AI that makes a related point: AI has no legal personality, and those it regulates remain accountable for their work, however it was prepared. The Bar Council also publishes material on AI for barristers. Read your own regulator's current guidance, since this article summarises only part of it.

Warnings from the courts about fake citations

A Divisional Court judgment from 2025, Ayinde v London Borough of Haringey, dealt with material put before the court that contained false citations, in circumstances where the use of generative AI was suspected. The court made clear that those who use these tools to conduct legal research must verify the results before relying on them, and that the consequences of failing to do so can include referral to the regulator.

The SRA's guidance refers to that judgment and says that relying on AI output is not a defence, and that a supervisor who fails to review adequately may be in breach if false citations reach the court. Read the judgment itself for the court's own words and for the range of responses it describes. This article does not attempt to summarise every part of it.

A practical routine for the advocate

Whatever your views on AI tools, a simple routine protects you. It costs time, but much less than the alternative.

  1. 1

    Open every authority in a trusted source, read the passage and confirm that it supports the point as stated.

  2. 2

    Check every quotation word for word against the judgment or statute.

  3. 3

    Confirm that each authority is still good law and from the right jurisdiction.

  4. 4

    Keep a note of which tool, if any, assisted with the draft, and what you checked.

  5. 5

    Make sure junior colleagues and those instructing you know the standard that you apply.

  6. 6

    If an error is discovered after filing, tell the court promptly and plainly.

Do
  • Open every authority and read the passage
  • Check every quotation word for word
  • Confirm each authority is still good law
  • Keep a note of which tool helped and what you checked
Don't
  • Cite an authority you have not opened and read
  • Delay telling the court if an error is found after filing
Checking AI-assisted court documents

What the guidance does not tell you

The judicial guidance is not a rule book for advocates, and nothing in it should be read as a statement of what the court will require of you in a particular case. Practice directions, the rules of the court you appear in and your regulator's guidance are the places to look for obligations. Individual judges may also have their own expectations about disclosure of AI use in the preparation of documents, so check the current practice direction and any relevant directions in the case.

Developments are continuing. The judiciary website also carries an interim report and consultation on the use of AI for preparing court documents, which is worth reading for where practice may be heading. Return to the primary sources periodically instead of relying on an article, including this one.

Instructing solicitors and supervising juniors

Advocates also receive drafts from others. A pupil's note of law, a solicitor's draft grounds or an expert's report may have been produced with help from a tool. Ask, politely, how authorities were checked. It is a normal question now, and the answer tells you how much of your own checking is needed.

Where a draft arrives with citations you do not recognise, treat them as unverified until you have read them yourself. Supervisors should build checking time into the timetable, as a rushed review is where errors survive.

Sources

  1. Judiciary of England and Wales: Use of AI for Preparing Court Documents, interim report and consultation
  2. Judiciary of England and Wales: Artificial Intelligence (AI) Judicial Guidance
  3. Solicitors Regulation Authority: Misuse of AI
  4. Judiciary of England and Wales: Ayinde v London Borough of Haringey (judgment)

This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.

Read next

UK judicial guidance on AI for lawyers in court | ChamberOne