Across many parts of the administrative justice system, caseloads appear to be rising. Tribunals handling administrative appeals are currently experiencing record backlogs. Public sector ombudsmen are observing similar trends. This has renewed focus on a perennial question in these institutions: what drives these caseloads? Increasingly, institutions must now ask a follow-up question: is use of AI part of the explanation?
Rises in caseload can have important consequences. While they may be interpreted positively (such as an increase in accessibility), significant and unexpected caseload increases in particular can create challenges for institutions. As the Public Service Ombudsman recently put it, ‘[w]hen demand rises unpredictably or quickly, the impact can be significant: growing operational pressures, increased backlogs, longer waiting times and, ultimately, reduced public confidence in the redress system.’ In the longer term, sustained caseload pressures often drive structural reforms to redress systems, which can, in turn, shape the quality of justice individuals receive.
Thus far, recent debates on rising caseloads have largely focused on tribunal and ombudsman institutions, as well as first-tier complaint processes within government bodies. However, early signs now suggest the Administrative Court’s judicial review caseload is rising.
The Ministry of Justice publishes judicial review caseload statistics every quarter, most recently covering April-June 2026. From April-June of 2025 onwards, each successive quarter has seen a notably higher rate of judicial review applications relative to levels in recent years. Figure 1 below shows these trends. In particular, the six months of 2026 for which we have data (January-June) have seen 2,630 applications. Were this rate to be replicated across the second half of the year (an admittedly predictive exercise), there would be 5,260 judicial review applications in 2026. This would represent a 36% increase compared with 2025 (3,647 applications) and a 56% increase compared with 2024 (2,953 applications). This recent increase in applications is stark compared to earlier years in the 2020s too, though less so when compared with the mid-2010s.
However, where much attention during the 2010s focused on the immigration and asylum judicial review caseload, there has in recent months been a sharp rise in the rate of ‘civil (other)’ applications – non-immigration or criminal judicial reviews, such as social security, planning, housing, etc. There were 2,204 civil (other) applications issued in 2025, higher than each previous year in the period captured in Figure 1. Meanwhile, in the first half of 2026 there were 1,670 civil (other) applications. If this pattern held across the entire year, there would be 3,340 civil (other) applications in 2026, 41% higher than in 2025 and 61% higher than in 2024. These trends should be viewed in historical context – for instance, the civil caseload regularly exceeded 2,000 applications annually in the 1990s and 2000s – but nonetheless represent a substantial change over a short period. In sum, across the last year, there has been a sharp and sustained increase in judicial review applications compared to the rest of the 2020s. Based on the data so far, 2026 may see comfortably the highest number of annual applications in well over a decade.
What explains these trends? Judicial review caseloads rise and fall over time. Many different factors contribute, and the literature already documents many of them well. However, Mr Justice Chamberlain, Judge in Charge of the Administrative Court, ventured an interesting theory in June’s Administrative Court User Group meeting. The minutes note that he ‘drew the meeting’s attention to the Court statistics which showed a significant increase in the amount of claims received’ and ‘suspects that the explanation behind this is that Litigants in Person are now filing claims with the assistance of artificial intelligence.’
There is growing recognition globally that public use of AI may catalyse increases in demand for justice institutions. One UK Supreme Court justice recently noted that institutions should ready themselves for a ‘tsunami’ of claims. In August, The Economist magazine claimed that ‘AI is breaking the British state’ as ‘citizens frustrated by the poor deal they get from the authorities are turning to artificial intelligence to file objections and appeals, and claim their dues.’ A recent report by JUSTICE and the Administrative Fairness Lab presented a range of examples from real chatbot exchanges in which people used AI chatbots for legal problems. In short, we know people use chatbots to address legal issues, even if there is contestation over how far this can be evidenced to be shaping caseloads. However, if we are seeing some effect of AI in judicial review caseloads, it is a particularly interesting example of this phenomenon. One might reasonably expect this AI effect to impact procedures like judicial review less than complaints systems closer to everyday administrative decisions and interactions; such systems often have no, or very little, cost to enter. It would be striking if judicial review was being meaningfully impacted.
If AI were having an effect of increasing caseloads, one possible effect on the judicial review caseload might be an increase in cases the courts regard as unmeritorious, since AI-generated cases may be of poorer quality. The User Group minutes specifically highlight AI use by Litigants in Person – any such AI-generated grounds of claim would likely therefore be submitted without the filter on poor-quality arguments provided by professional legal advice on the merits.
One way to examine this question is to look at whether the permission stage of judicial review is filtering out a higher rate of claims, which may indicate that some of the increase in applications is due to more meritless claims (potentially generated by AI) being issued. Among claims which proceeded to permission each year, both across all claims and isolating civil (other) claims specifically, the percentage rate of claims being granted permission on the papers or at oral renewal can be seen in Figure 2 below. These figures do not take into account claims which are withdrawn having proceeded to permission (many due to the public body agreeing to a consent order). While the grant rate in 2025 and 2026 (so far) has been slightly lower than in 2021-2024, it does not appear to meaningfully differ from the trends of previous years. There does not therefore appear to be a notable increase in claims which are filtered out by the permission stage as lacking merit—at least not one which is yet visible in the data.
Another indicator could be the rate of claims being certified at the permission stage as ‘Totally Without Merit’ (TWM). This is a designation applied to a smaller subset of claims refused permission which are truly hopeless and bound to fail. As is shown in Figure 3 below, however, among claims reaching the permission stage, the certification of claims as TWM in 2025 and 2026 is (so far) also entirely consistent with broader trends over the last decade. Accordingly, there is no indication yet that an increase in judicial review applications in 2025 and 2026 has increased the rate of claims warranting a TWM designation.
Ultimately, this leaves us with a mixed picture. We can clearly see the judicial review caseload rising, but it is less clear whether AI is playing a role. The relevant aspects of the Administrative Court’s data do not show the signs we might reasonably expect if this were the case, or at least not yet. However, while the minutes of the User Group document appropriate caution in linking AI use by litigants to this recent caseload shift, Mr Justice Chamberlain’s expert insights, informed by his familiarity with the Court’s day-to-day work, deserve considerable weight. This developing issue deserves continued, careful study and attention, not only because institutions must separate narrative from reality, but also because sustained caseload pressure has historically led to procedural reorganisation. Judicial review—a relatively legalistic form of adjudication—is particularly vulnerable to caseload shocks in this respect.
Many thanks to the editors for helpful comments on a previous draft of the blog.
Sam Guy is Lecturer in Law at the University of York.
Joe Tomlinson is Professor of Administrative Law at King’s College London.
(Suggested citation: S. Guy and J. Tomlinson, ‘Rising Caseloads in Judicial Review – A Question of AI?’, U.K. Const. L. Blog (29th October 2026) (available at https://ukconstitutionallaw.org/))
Leave a Reply