Written complaints and claims across several British public services have risen since ChatGPT launched in November 2022, according to a Sky News report published October 20. The report says the pattern appears across unrelated channels, raising questions about how public bodies can handle a larger volume of correspondence without making it harder for people to seek help or challenge decisions.

Sky News technology correspondent Rowland Manthorpe compared figures from multiple channels. Since ChatGPT’s launch, household complaints about water companies rose 144%; complaints to English local authorities increased 78%; and written complaints to GP practices rose 71%. The report also recorded increases in complaints to the Department for Work and Pensions, individual employment tribunal claims, legal services, the NHS, police forces and public services.

The figures cover different services and types of correspondence, each with its own possible causes. Sky News says what stands out is a similar shape across many of the series: relatively little movement for years, followed by rises that began around late 2022 and then accelerated. The timing is consistent with wider public access to generative AI, but the figures alone do not show that AI produced the increases.

To examine whether requests might have been written with AI, Manthorpe submitted 100 recent Freedom of Information requests to an AI-text detector. It labelled 89% fully AI-written and 9% human; a smaller sample from 2019 was labelled entirely human. The report cautions that detectors can be unreliable and that this small test cannot establish who wrote the requests or whether AI caused the wider trend.

Pressure on Public Service Inboxes

More correspondence can add work for public employees and extend the time needed to respond. A survey of civil servants cited by Sky News found that 58% said public use of AI-generated content significantly affects their work. Some described receiving correspondence at speed, feeling they were arguing with ChatGPT, and finding the work exhausting or anxiety-inducing. These are reported experiences from surveyed staff, not a measure of the effect across every public body.

The policy choice matters to both residents and institutions. If public bodies respond to heavier workloads by adding barriers to complaints or information requests, people with legitimate concerns could find it harder to be heard. If they instead adapt their processes, they will still need to distinguish poorly formed or repetitive submissions from valid requests and claims. The report’s figures do not settle which response will work or how much extra correspondence is actually AI-generated.

Some of the data cited in the report suggests that increased volume does not make every complaint less meritorious. The source material says about 40% of DWP complaints are still upheld, down from a level in the 50s; it gives no precise period or full methodology for that comparison. It also says roughly one-third of Financial Ombudsman complaints were found to be AI-generated, without specifying the sample or period in the supplied account. Those figures are reminders that authorship and validity are separate questions.

A Rise Across Different Channels

The report places the increase against a longer period in which the number of Freedom of Information requests sent to British public bodies changed little. Its central comparison is not a single surge at one agency but increases across forms of written contact, including complaints about water companies, local authorities, GP practices and government departments, alongside employment tribunal claims.

Each channel has its own circumstances. Water companies, for example, have faced difficult years, and that could contribute to complaints about them. The figures supplied by Sky News do not apportion the rise among service problems, changes in public awareness, shifts in how bodies record cases, population changes or AI-assisted writing. The repeated timing is a reported pattern, not proof of a shared cause.

The issue is also being discussed outside Britain. Researcher Chris Schmitz’s paper, covered by TechCrunch in September, documents 84 possible cases of “agentic flooding” across 11 jurisdictions. The term refers to high-volume activity involving AI agents; the cases are described as possible, and most were roughly flat before 2022 before rising. TechCrunch also reported that the UK Housing Ombudsman received about 2,600 complaints in 2022 and just over 7,000 last year, while complaints to the US Consumer Financial Protection Bureau increased fivefold over the same period. These comparisons indicate a wider concern, but do not show that every increase has the same cause.

““It’s like being bombarded with correspondence at speed.””

— Civil servant quoted in the survey cited by Sky News

What the Figures Cannot Show

The available figures do not establish that generative AI caused the rise in complaints, requests or claims. They show changes in volume and timing; they do not isolate AI use from other explanations, including service problems or changes in how people submit and record complaints. The different channels may also use different definitions, periods and collection methods, limiting direct comparisons.

The detector test is particularly limited. It covered 100 recent FOI requests, and the report warns that AI-text detectors are unreliable. Its result cannot reliably identify each author or establish the role AI played in the increase. The supplied information also does not give the survey’s sample size or detailed methodology, or explain the periods and methods behind every channel’s growth figure. How much of the new volume is AI-assisted, and how much represents valid but newly expressed demand, remains unclear.

The May guidance from the Information Commissioner’s Office does not resolve those questions. It responds to reports of more numerous and complex FOI requests, including some that misquote legislation or require clarification, while setting out how authorities should treat AI-written requests.

How Authorities Handle Requests

In guidance issued in May 2026, the Information Commissioner’s Office said AI-written FOI requests remain valid. Public authorities do not need to determine whether a person or an AI system wrote a request. Where a request is unclear or badly formed, they may ask the requester to clarify it. The guidance followed reports from authorities of higher request volumes and more complex submissions.

The next test for public bodies is whether they can manage the workload while preserving access to complaints and information. The report does not identify a national policy change or a timetable for one. It leaves open whether authorities will change how they process correspondence, what additional resources they may need, and how they will track the effect of AI use without treating authorship as a proxy for whether a complaint or request deserves a response.

Further figures and research may help clarify those questions, but the supplied report does not specify a next release or government review. For now, the reported increases and staff accounts describe pressure on public service inboxes; the extent to which AI is responsible, and what response will be adopted, remain unsettled.

Key Questions

What did the Sky News report find?

It reported increases since ChatGPT launched in November 2022 across several kinds of written contact with British public services, including a 144% rise in household complaints about water companies and a 78% rise in complaints to English local authorities.

Do the figures prove AI caused the rise?

No. They show increases and a similar timing across different channels, but do not isolate AI from other possible causes. The report also says AI-text detectors are unreliable.

Are AI-written Freedom of Information requests valid?

According to guidance from the Information Commissioner’s Office issued in May 2026, AI-written requests are valid. Authorities may ask for clarification when a request is unclear or badly formed.

Does AI use mean a complaint is without merit?

No. Authorship does not establish whether a claim is valid. Researcher Chris Schmitz said most cases in his research involved people claiming something they were entitled to claim.

Source: Thorsten Meyer AI