As the legal sector pushes ahead with AI adoption, Gabriella Banham takes a look at how the use of AI is impacting claimants. Relying on freely available platforms for legal advice may seem like a cost-effective option but, as Gabriella explains, it will almost always be a false economy.
Artificial intelligence has arrived in the legal sector with all the fanfare one would expect of a transformative technology. It promises efficiency, accessibility and cost reduction. It drafts emails in seconds, summarises dense case law at speed, and at first glance appears to level the playing field for those without legal representation.
Yet behind the polished prose and persuasive tone lies a more uncomfortable reality. AI may sound like a lawyer. It may even look like one, on the page. But it is not one, and the courts are increasingly being asked to deal with the consequences of parties forgetting that distinction.
The seduction of ‘instant law’
For businesses and individuals alike, the appeal of AI-generated legal assistance is obvious. Faced with rising legal costs and increasingly complex disputes, many are tempted to type a question into a chatbot rather than instruct a solicitor.
This is particularly evident among litigants in person who may use AI tools to draft pleadings, research legal authorities, or construct arguments. On one level, this represents a democratisation of legal knowledge. On another, it introduces a fundamental risk: the outsourcing of legal judgment to a system that does not understand the law but merely predicts what it might sound like.
As the Law Society noted, while generative AI offers significant opportunities, it also carries risks that remain only partially understood.
When AI gets it wrong: From embarrassment to sanction
The difficulty with AI is not that it occasionally makes errors. It is that it does so convincingly.
Courts across multiple jurisdictions have now encountered cases where AI has produced entirely fictitious legal material complete with case names, citations, and judicial reasoning. In Harber v HMRC [2023], a UK tribunal found that a self-represented litigant had relied on case authorities that did not exist at all, despite appearing superficially authentic.
Similarly, in a Manchester civil matter, a litigant-in-person presented four authorities generated by ChatGPT. Upon scrutiny, one case had been wholly invented, and the remaining citations bore no relation to the judgments they purported to quote.
These are not isolated incidents. Hundreds of cases globally have now been identified as containing AI-generated errors, including fabricated citations submitted to courts.
The consequences can be severe. In the United States, lawyers have been sanctioned and fined after relying on AI-generated material without verification, with courts criticising submissions “replete with citation-related deficiencies”.
More recently, a federal court went further, pausing proceedings entirely, removing legal representatives from the case, and imposing financial penalties after both sides filed AI-generated material containing inaccuracies.
In England and Wales, the judiciary has also taken a firm stance. The High Court has made clear that presenting false authorities, whether generated by AI or otherwise, may amount to contempt of court and, in serious cases, conduct capable of constituting a criminal offence.
The hidden risk for litigants in person
For litigants in person, the risks are particularly acute.
Unlike regulated professionals, litigants in person are not trained to interrogate legal sources, assess evidential weight, or verify procedural compliance. AI fills that gap with apparent confidence, providing answers that sound authoritative, structured and legally fluent. The difficulty is that those answers may be only partially correct, or even entirely incorrect and fabricated.
The potential consequences are many. Misleading the court (even inadvertently) by submitting incorrect authorities or legal arguments can undermine credibility at the earliest stage. Procedural mistakes are common, with AI often failing to reflect the nuances of the Civil Procedure Rules, leading to defective pleadings or missed deadlines.
There are costs at stake – financial and otherwise. Even where a party is unrepresented, a reliance on flawed material may increase costs, delay proceedings or result in sanctions. Then there’s the cost of losing an otherwise viable claim or defence: a case may fail not on its merits, but on the way it is presented. And at what cost does the loss of privacy come? Disclosable documents typed or uploaded into AI are not subject to privilege in the way that solicitor-client communications are and AI chat history may be disclosable to the other party.
Research suggests that a significant number of AI-related errors in court filings originate from self-represented parties. Courts may show a degree of understanding towards litigants in person, but this does not insulate them from the practical realities: time lost, credibility damaged, and opportunities missed.
Regulation, responsibility and the limits of technology
The regulatory position is clear. AI is a tool, not a substitute for professional judgment.
Guidance from the Law Society and Bar Council emphasises that legal professionals must verify all AI-generated content, maintain oversight, and exercise independent judgment at all times. Blind reliance on AI risks falling below the required standard of competence.
The UK’s broader approach to AI regulation is grounded in principles of safety, transparency and accountability principles that apply with particular force in a legal context.
At its core, the issue is simple: AI does not understand truth. It does not distinguish between binding authority and persuasive analogy. It does not appreciate nuance, context or judicial discretion. It produces language, not legal reasoning.
A great example of this is an AI chat history that recently came across our desk, which fundamentally misunderstood the trust law principles that apply to property ownership disputes. The AI confidently concluded that a party had no claim because the parties’ agreement had not been recorded in writing. In doing so, it completely overlooked the potential application of a common intention constructive trust, which may arise even where the agreement to share a property was made orally.
Why legal representation still matters
The growing body of case law on AI misuse reveals a consistent theme: the problem is rarely the tool itself, but the absence of professional oversight.
A qualified lawyer does more than produce words. They assess the strength of a case, not just its presentation; interpret law in context, rather than in isolation; anticipate procedural pitfalls and strategic risks; apply judgment informed by experience, not probability.
AI, by contrast, offers the form of legal advice without the substance of it.
For individuals tempted to rely solely on AI in the course of a dispute, the risk is not merely that something may go wrong. It is that it may appear to go entirely right until it reaches a courtroom.
Conclusion: Proceed with caution
AI will undoubtedly continue to shape the future of legal practice. Its benefits are real and its potential significant. But so too are its limitations.
For litigants in person, the temptation to rely on AI as a substitute for legal advice is understandable. In many cases, however, it is a false economy that may cost far more in the long run.
In law, as in life, there remains no substitute for informed judgment. And that, for now at least, is something artificial intelligence cannot provide.
About the author
Gabriella Banham joined Birketts in March 2020 and has over eight years’ experience working within the legal industry, assisting solicitors in full-service law firms in Cambridge, gaining experience in property litigation, construction litigation, employment law and real estate.
