Divorce in the age of AI: a new challenge for family law

blogs
November 16, 2026
Posted by:
David Lillywhite
&

Divorce in the age of AI: a new challenge for family law

‍First published in FT Adviser on 12 November 2026

David Lillywhite - Partner at Burgess Mee

Readers of a particular vintage may recognise little Carol Anne intonating ‘they’re heeeere’ just prior to all hell breaking loose in 1982’s Poltergeist. Make no mistake, family lawyers — and, no doubt, many financial advisers — are experiencing a similar, sinking feeling with the rise and rise of AI.

We should all start with the assumption that AI is already being used in some capacity by our clients: the question is when it will appear in the process, and to what end. 

With time and money at an ever-increasing premium, it is understandable why people getting divorced resort to AI, whether they are litigants in person or instructing a lawyer: it is easy to use, efficient and it gives a dopamine rush that often mistakes output for productivity. It might start with a short sense check or a rewritten email to an ex-spouse before graduating to a draft letter to the other side or, worse, a full-blown modelling of the financial settlement. 

While AI can be an effective administrative tool, it can also create significant legal risk and increase costs. Practitioners therefore need to understand how their clients are using AI before any lasting damage is done. 

Risks of using AI in divorce proceedings

English family law is a discretionary system. Lawyers talk to their clients about a ‘bracket of outcomes’ and there is no fixed or binary formula to apply for a quick solution. Every case will turn on its facts, but beware the AI echo chamber that leans into the user’s own inputs.

A client who has already run their case through a generative AI is likely to approach a first meeting with unrealistic expectations, looking only for the adviser to achieve the settlement they have been told they are entitled to. 

Perhaps chief among the real-world risks are the issues of confidentiality and privilege. Clients increasingly put confidential information that should not be placed into the public domain into their AI of choice for analysis and drafting, with no mind to the consequences.

Whether it is one party’s full financial disclosure (and supporting enclosures), solicitor advice, or welfare reports concerning the children, sharing this information with consumer AI tools (and a chatbot is usually the point of entry for most clients) risks sending it out into the ether and constituting a waiver of legal professional privilege — that valuable information barrier which, once gone, is very difficult to recover (see Munir vs Secretary of State for the Home Department [2026] UKUT 81 (IAC)).

This is to say nothing of the fact that simply sharing this information could constitute a contempt of court or possibly even a criminal offence.

AI makes it all too easy to generate long, detailed correspondence, whether with a solicitor or a spouse. The client sees content, but the qualitative barometer will be askew. A solicitor will need to review all the material, as they cannot afford to miss anything that might actually be of use, but this generates an associated expense.

The court is very alive to the issue of cost and not shy about enquiring why a spouse has spent that much more, possibly adjusting the outcome as a result. 

A client who has sent a long, rambling response to a spouse might see this as a strident ‘win’ of reason — a chance to say everything they need to say on a discrete point. A concise, considered response drawing on lived experience and exploring the nuance of the situation will usually be much more effective than bamboozling a spouse with pages of challenge and anticipated counter-argument. 

How the courts are responding 

On August 17 2026, the Solicitors Regulation Authority issued a warning notice to the profession about the misuse of AI and the need for firms to have proper safeguards in place. The SRA’s chief concerns relate to AI ‘hallucinations’ in legal work and submissions to the court and client confidentiality.

The warning notice highlights the risk to clients in relying on incorrect information and poor advice, the increased pressure on the court, which may need additional hearings to properly explore any concerns with evidence, thereby slowing access to justice, and the wider reputational harm to the profession – see R (on the application of Ayinde) vs Haringey LBC [2025] EWHC 1383 (Admin), in which the court was concerned with AI-generated case citations.

In the case of Godwin vs Godwin [2026] EWHC 923 (Ch) — albeit outside the family law sphere — a defendant and a witness had each used a ‘digital assistant’ (revealed to be ChatGPT) to help prepare their statements.

Although both maintained that ChatGPT had not added, removed or rearranged any words from their first drafts, the judge was unable to verify this as those first drafts were not before the court. As a result, the judge had to treat the statements with caution, giving them less weight as he could not be sure they were in their own words.

Practical takeaways for clients

Advisers in the divorce and financial services industries should assume their clients are already using AI and will be stress-testing their advice alongside the latest consumer-facing models. 

For clients, AI should be a tool, not a crutch. It is not a substitute for professional therapeutic support or legal advice. Used in the right way, AI can help to streamline the process and marshal thinking, but so much depends on the prompts that a client uses.

A good prompt can be invaluable, whereas a broad effort dashed off in haste at 3am, combined with a client mistaking sheer volume of output for quality of response, will only incur higher costs and more work. Clients must be made well aware of that in the early stages of any instruction. 

AI can help with general administration and organisation, but it should not be asked what clients might be entitled to or to model an outcome based on the available financial disclosure and factual matrix.

On no account should a client feed an AI anything confidential: once a client has shared that information into a public tool, privilege may have been waived. 

The same can be said for the preparation of any witness statements, which should always be in the client’s own words. Lawyers will be looking harder at statements to see whether there has been any assistance from AI and bringing that to the attention of the court.

The subject will be ripe for cross-examination by experienced advocates looking for weaknesses in the opposing case. The risks are especially high in children proceedings when the court will want to hear from the parents, not whichever model they fed the Cafcass report to the day before the statement was due. 

Conclusions 

AI is not going away. For clients, it promises all of the answers with none of the obvious cost. However, divorce and family disputes rarely lend themselves to quick answers or formulaic solutions; English family law is built on discretion, nuance and the careful balancing of individual circumstances.

For advisers, the challenge is not to compete with AI, but to demonstrate the value of what it cannot provide: strategic counsel, depth and breadth of experience, and human judgment.

David Lillywhite is a partner at family law firm Burgess Mee

recent guides

Your choice regarding cookies
Clicking the Accept All button means you are accepting analytics and third-party cookies. We use cookies to optimise site functionality and give you the best possible experience. To control which cookies are set, click Settings.

request a call back

Complete your details below and we will contact you to book a free, no-obligation consultation at a time convenient to you.
Please note we do not offer Legal Aid

Your submission has been received
Oops! Something went wrong
click