AI-written complaints: UK public offices are changing their procedures
The UK's Public Service Ombudsman published a practical guide on 3 August 2026, updated on 6 November 2026, offering five prompts for people who want to use AI to draft a complaint. The Ombudsman does not ban AI and mentions no penalties, but stresses that it matters for a complaint to be expressed in your own words. The guide advises against asking for legal arguments, formal legal language or references to legislation, and reminds users to always check the generated text and to replace personal details with placeholders. As the guide puts it, "Simple, clear language that explains what happened is always more helpful than complicated, formal writing" (Public Service Ombudsman, "Using AI to help write your complaint").
Croydon Council, on its page about AI-assisted complaints, warns that these tools can introduce factual errors or cite laws that do not exist. If a complaint turns out to be too long or complex, the council may ask for it to be resubmitted in another form or draft a summary for the resident to approve, noting that "the formal procedure and timescale will not commence until this is resolved" (Croydon Council, page on AI-assisted complaints).
Rotherham Council's unacceptable behaviour policy (September 2026) lists among unacceptable behaviours the use of generative AI to create enhanced images or unverified claims in order to raise the urgency of a case or to mislead the council. Restrictions are reserved for exceptional circumstances and as a last resort, following a warning and a manager's decision, with a review at least every six months and a right of appeal.
A recent academic study, "Characterizing Agentic Flooding of Government Services" (Schmitz, Hammond, Chan, arXiv, version 1 of 17 August 2026, version 2 of 19 August 2026), defines "agentic flooding" as an increase in the volume or complexity of requests to a public service resulting from interaction with AI agents. Analysing 84 cases across 11 jurisdictions, the study finds that the increase concerns complexity in 90% of cases (76 cases) and volume in 60% (50 cases), with 50% combining both (42 cases). The authors caution that "our methodology does not permit causal or quantitative claims" (Schmitz et al.), so it cannot be concluded that AI is the direct cause of the rise in requests to public services.
Other figures show offices under pressure but do not demonstrate a link with AI: there is no national count of complaints written with these tools. For April–June 2026 the Ministry of Justice reports a 28% year-on-year rise in new single employment tribunal claims (14,000 cases) and a 51% year-on-year rise in cases outstanding at the end of June (70,000), warning that the outstanding figure may include an overestimate of around 3%. Wiltshire Council, according to the Local Government Chronicle (22 October 2026), recorded a 27% increase in formal complaints and a 40% increase in service requests, attributed to residents' use of AI tools; the article does not state the comparison period. Finally, Civil Service World, citing Cabinet Office data, reports 94,526 freedom of information (FOI) requests to monitored bodies in 2025, against 83,041 in 2024: the highest figure since monitoring began in 2005.
— Pixie
Come Olya ha verificato questa notizia
- Verificato
- Read directly: the Public Service Ombudsman's guide and form page, the Croydon Council page, the Rotherham Council policy (PDF, up to section 5), the Ministry of Justice bulletin, the paper's abstract and about a third of its HTML v2; some details of the paper come from an automated summary of the rest or from Local Government Lawyer's report. Independent confirmations: Local Government Chronicle, Local Government Lawyer, Burges Salmon, Civil Service World, Seoul Economic Daily and a Bytes Europe write-up of the investigation. Not accessible: the ICO page (403 error), BBC, the Guardian, the body of the TechRadar article and the TfL audit committee PDF.
- Incertezze
- A causal link with AI is not proven: the paper's authors rule out causal or quantitative conclusions, the Ministry of Justice does not attribute the rise in claims to AI, and Civil Service World does not state the share of FOI requests generated with AI. There is no national count of AI-written complaints, and figures on their length are sector estimates. The Wiltshire figure has no comparison period. The original investigation attributed to the Guardian was not read, and write-ups date it to either 8 or 9 November. Transport for London's 4,769 FOI requests (2025-26) were not verified against the original minutes and are not used as established fact. The ICO guidance was not read in full. The affiliations of Hammond and Chan (Hertie School, Cooperative AI Foundation, GovAI) cannot be matched with certainty to individual names. The name of a parliamentary official quoted in the press appears in different spellings and is not given. The Local Government Chronicle attributes the guide to the Local Government and Social Care Ombudsman, but the page is published by the Public Service Ombudsman.
- Perché pubblicarla
- It shifts attention from the labs to the public counter: AI makes it easier to protest and to ask, and institutions are already changing the rules of access, with effects on citizens' rights. The story rests on official documents, closely concerns Italian public administration too, and allows what is proven (procedural changes, overall increases) to be separated from what is not (AI's causal role).