In June, Mr Justice Chamberlain, the Administrative Court’s judge in charge, told the court’s user group that the court suspected litigants in person were now using AI to help file claims. Writing on the UK Constitutional Law Association blog on 30 October, Sam Guy of the University of York and Joe Tomlinson of King’s College London went looking in the court’s own statistics for the trace that theory should leave. Judicial review applications are rising sharply, they found, but the share of claims granted permission, and the share certified as hopeless, are in line with earlier years. Our news report sets out their figures. This analysis takes up the wider question their piece raises. AI-drafted claims are now being reported at several tiers of UK redress, from council complaints desks to the Administrative Court, and on our reading the evidence for them gets thinner the further up the system you look. What should the institutions on the receiving end prepare for?

Strategic Insight: Guy and Tomlinson suggest one might expect judicial review, a comparatively legalistic procedure, to feel any AI effect less than complaints systems that cost little or nothing to enter. In our view, that makes the court data a hard test of the AI theory, not a verdict on the wider redress system.

Why are institutions asking about AI and caseloads now?

Caseloads appear to be rising across much of administrative justice. The blog points to record backlogs in tribunals hearing administrative appeals and to similar trends at public sector ombudsmen. The Parliamentary and Health Service Ombudsman (PHSO) put numbers on the second point in a December 2025 briefing covering six UK schemes. Complaints considered in 2024-25 were up 25% on 2022-23 at both the PHSO and the Local Government and Social Care Ombudsman, 39% at the Scottish Public Services Ombudsman and 42% at the Housing Ombudsman.

The same briefing lists three factors among the likely drivers of that demand. One is growing use of generative AI by complainants, a trend the PHSO, the Scottish ombudsman and the Public Services Ombudsman for Wales have all noted. Another is pressure on public services themselves, and the third is ombudsmen’s own success in becoming more accessible. The briefing also says the full set of drivers “remains insufficiently understood”.

The real story is a measurement gap

Each of these accounts is a judgement, a sample or a suspicion rather than a count.

  • Rhondda Cynon Taf Council says in its social services complaints report that “it is considered that AI tool availability has influenced a higher number of stage two complaints”, according to WalesOnline’s account of the report.
  • The Financial Ombudsman Service found that as many as a third of replies to its initial assessments, in a small sample checked earlier this year, looked AI-generated or heavily AI-assisted, Insurance Business reported.
  • Caseworkers in MPs’ offices told The Independent that about half their caseload now arrives at least partly AI-written.
  • The Administrative Court’s minutes record that “the Court suspects” AI-assisted litigants in person explain the rise in claims.

Critical Context: The PHSO’s briefing says AI-enabled submissions can raise volume, raise expectations of a quick resolution and add complexity when the content is less tailored or wrong. It lists AI as one driver among several and does not put a figure on its share.

MeasureFigureSource
Judicial review applications, January to June 20262,630Ministry of Justice data, via Guy and Tomlinson
’Civil (other)’ judicial review applications, January to June 20261,670Ministry of Justice data, via Guy and Tomlinson
Claims received by the Administrative Court Office over the year to June 2026About 5,500, up from just under 4,000Administrative Court User Group minutes, 24 June 2026
Housing Ombudsman complaints considered, 2024-257,082, up 42% since 2022-23PHSO briefing, December 2025
Rhondda Cynon Taf social services stage two complaints, 2025/2620, up from fourCouncil report, via WalesOnline
Financial Ombudsman cases, first half of 2026/27110,500Insurance Business

What is really happening at the court door?

The user group minutes add two details the blog does not quote. The figure Mr Justice Chamberlain cited counts claims received by the Administrative Court Office, about 5,500 over the previous year against just under 4,000 before. That is not the Ministry of Justice series the academics use, which put 2025 applications at 3,647, but both point the same way. And the minutes record that the court would consider whether its administration could use AI itself, with discussions under way.

The academics’ test is a fair one. Litigants in person who draft grounds with AI would likely lack a lawyer’s advice on whether the case is any good, so one possible effect of a wave of such claims might be more claims failing at permission. Among claims reaching that stage, grant rates in 2025 and so far in 2026 sit a little below the 2021-2024 level, but not meaningfully different from earlier trends. ‘Totally Without Merit’ certifications, the label kept for claims bound to fail, are consistent with the last decade.

In our view, two readings survive this. AI may not be driving the rise. Or it may be, and the claims it helps produce are no weaker than the rest. The measures the authors use cannot separate the two.

Reality Check: The permission figures leave out claims withdrawn after reaching the permission stage. Guy and Tomlinson note that many of these end with the public body agreeing a consent order. Claims withdrawn after the public body agrees a consent order are not counted in those permission figures.

The growth is concentrated in ‘civil (other)’ cases, the non-immigration, non-criminal claims covering social security, planning, housing and similar areas. The authors caution that this caseload regularly passed 2,000 a year through the 1990s and 2000s. Even so, they describe the recent shift as “a substantial change over a short period”.

Who is affected?

Public bodies handling first-tier complaints. Councillor Richard Clewer of the Local Government Association told The Independent that longer and more complex AI-generated complaints are “increasing administrative burdens at a time of significant resource pressure”. Rhondda Cynon Taf’s social services customer feedback team handled 814 interactions in 2025/26, up from 735, and the council is preparing guidance that will cover the use of AI applications.

Ombudsmen. James Dipple-Johnstone, chief ombudsman at the Financial Ombudsman Service, told the Financial Times about the worst example he had come across: a consumer’s submission in which 37 of the 39 references were incorrect. The service is now phoning more consumers to find the real issue in a dispute.

The courts. The Lady Chief Justice’s annual report, as the Law Society Gazette reported, says growing numbers of litigants in person are using AI to build their claims in the Administrative List and in the King’s Bench Division’s Civil List. The report says this improves access to justice but places “significant demands on both court staff and the judiciary”, with extra case management often needed to work out what remedy the litigant actually wants.

People with legal problems. JUSTICE and the Administrative Fairness Lab surveyed 3,287 people for a study. Of the 1,428 who reported a legal dispute in the previous two years, 16.3% had turned to an AI chatbot for help, rising to 26% among 18 to 24-year-olds. The 77 conversations participants shared suggest people turn to chatbots for common problems, such as consumer, housing and employment disputes, that largely fall outside legal aid.

Strategic Reality: Rising volume is not the same as rising noise. The PHSO says uphold rates “remain high”, which it says suggests complaints are also “raising substantive issues where failings have been found”. At Rhondda Cynon Taf, 14 stage two complaints were upheld or partly upheld and five were not.

What this means for public bodies that face judicial review

The fastest growth is in claims about social security, planning, housing and other civil matters. Guy and Tomlinson warn that, in the past, prolonged caseload pressure has ended in procedures being reorganised, and that judicial review is particularly exposed to caseload shocks. The PHSO’s briefing points to where pressure can be relieved earlier. Among the longer-term structural changes it lists is better first-tier complaint handling in public bodies. In our view, a complaint resolved properly at the council desk is one that never needs an AI-drafted claim form.

The open challenges

Separating AI-assisted from AI-generated

According to The Independent, AI has empowered genuinely aggrieved people to complain, whilst the complaints themselves are often lengthy and inaccurate. Sam Freedman, the journalist quoted by The Independent, put the bind plainly: “Some AI-written complaints will be genuine so you can’t just filter out AI, you have to look at them.” He added that once bodies use AI to sort them, “it becomes a situation where it is just chatbots talking to each other”.

Paying for capacity

Lord Briggs, a Justice of the UK Supreme Court, argued in a May 2026 speech that the taxpayer seems unlikely to fund a big increase in civil court staff and judges. His conclusion was that “AI is going to have to come to the rescue of court staff and judges”. The Administrative Court’s minutes suggest the same thinking within the court itself.

Writing rules before the wave arrives

Lord Briggs asked whether procedural rules should require litigants to disclose their use of AI, and warned that the rules adapted slowly to digitisation. The Civil Justice Council has since reported on AI in court documents. Whether any disclosure rule would also give courts the data to answer Mr Justice Chamberlain’s question is a separate matter.

Protecting access while filtering demand

Freedman told The Independent that if the trend continues, “local authorities will have to change the way that their systems work”, perhaps by making it harder to complain, which “may have negative effects for everybody”. Lord Briggs warned that the court door “may very well burst off the hinges” without mechanisms to cope.

⚠️ Warning: Caseload pressure has reshaped redress before, according to Guy and Tomlinson. If access rules are tightened on the belief that AI is flooding the system with weak claims, whilst the court’s own merit data shows no rise in weak claims yet, the people turned away may include claimants who would have won.

Implementation Note: The PHSO lists the measures ombudsmen are already prioritising: enhanced triage, “improved prioritisation of complex or urgent cases”, streamlined initial assessment and more use of technology. In our reading, none of them depends on knowing which complaints were AI-assisted.

What to watch

Four signposts would show whether AI is reshaping UK redress or being blamed for a rise with other causes:

  1. The next quarterly judicial review statistics. The 5,260 projection for 2026 assumes the first half’s rate holds. If permission grant rates fall or ‘Totally Without Merit’ certifications rise as volumes climb, that would be the first sign in the data of Mr Justice Chamberlain’s theory. Steady rates would keep the question open.
  2. The PHSO’s research on drivers of demand. The ombudsman is commissioning external research. Whether it can separate AI use from service failures and improved accessibility will show whether redress bodies can plan on evidence rather than anecdote.
  3. The Administrative Court’s own use of AI. The minutes say discussions are under way. Any tool the court adopts to manage incoming claims could also give it a way to record how many show signs of AI drafting.
  4. First-tier intake changes. Rhondda Cynon Taf plans a new online complaints form built to draw out what the complaint is about and what the complainant wants. Forms like this, if adopted more widely, could ask complainants whether they used AI and start a count.

Guy and Tomlinson close on a mixed picture: the caseload is clearly rising, whilst AI’s part in it is less clear. That is an honest place to stop, and an uncomfortable one for institutions that have to plan budgets and procedures now. The bodies best placed to cope will be those that know what is arriving at their door, rather than inferring it from the length and tone of the paperwork.


Source: Sam Guy and Joe Tomlinson, Rising Caseloads in Judicial Review: A Question of AI?, UK Constitutional Law Association blog, 30 October 2026. Guy lectures in law at York; Tomlinson is a professor of administrative law at King’s. Also drawn on: the Administrative Court User Group minutes of 24 June 2026; the PHSO’s briefing Demand across the public service ombudsman landscape (December 2025); Lord Briggs, AI and Civil Justice: Preparing for the Tsunami (May 2026); JUSTICE and the Administrative Fairness Lab, What AI chatbots can teach us about unmet legal needs; and reporting by WalesOnline, Insurance Business, The Independent and the Law Society Gazette.

This strategic analysis was written by Resultsense, a UK-focused AI news and analysis publication. We will be watching whether any UK redress body starts counting AI-assisted claims rather than inferring them. Read more analysis at Insights, or get in touch.