200-Page AI Complaints Are Burying UK Council Officials

A mountain of complaint paperwork piled up in front of a UK council building

For a civil servant, even a single page of complaint paperwork is a burden. One wrong answer can become a precedent that haunts you for years.

Lately, though, the worst cases reach UK local council staff as a single complaint running 200 pages (basically a whole book).

Since free AI chatbots landed in ordinary people's hands, the volume of paperwork hitting government offices and courts has exploded.

The 'AI Clog': Complaint Letters by the Truckload

✦ Complaint and Filing Lengths Compared

Unit: pages

Old average council complaint, about 2 pages
EAT skeleton argument cap 20 pages
Recommended: 5 to 15 pages
Worst council complaints today 200 pages
ChatGPT filing in the Hancox case 300 pages
About 132,000 words

Not long ago, a complaint letter to a UK local council ran about two pages on average. Most were about a pothole that needed filling or garbage nobody had collected.

These days, council staff wrestle with stacks of paper nearly as thick as a book. Citizens armed with free AI chatbots often send complaints around 20 pages, roughly ten times the old average, and some have started sending 200-page ones.

Deborah Evans, CEO of Lawyers in Local Government, the association for UK council lawyers, said, "Many complaints are now about ten times longer than they used to be."

The trend even has a name in Britain: the 'AI clog,' as in a system jammed up with too much paper.

Paperwork is piling up in MPs' offices too. A senior aide in a Labour MP's office said, "Nearly half the complaint emails coming in look like AI was involved somehow."

Freedom of Information (FOI) requests, which let people demand information from public bodies, have shot up too. Transport for London (TfL) received 4,771 of them in 2025-2026, a record.

〈How AI-written complaints are flooding UK councils〉
* From the Kings Of The Underground YouTube channel.

A 130,000-Word AI Brief Hits the Courts

In July 2026, the UK's Employment Appeal Tribunal (EAT) held a hearing in Hancox v Sutherland and Others. A skeleton argument, which lays out the key points of a case before the hearing, is usually recommended to run 5 to 15 pages.

Go past 20, and the judge can ask for a shorter version or cut your oral argument time.

✦ How the Hancox Case Played Out (July 2026 Hearing)

1Two days before the hearing
Filed a 300-page ChatGPT-generated document
2With the filing
Said he didn't expect the judge to read it all, then asked the other side to pinpoint important errors by 4 p.m. on July 28
3The tribunal's view
Using AI help in itself is allowed. But the 300 pages got in the way of hearing prep, and pushing the checking onto the other side was unacceptable

But the appellant, Terence Hancox, filed a 300-page ChatGPT-generated document two days before the hearing. That's about 132,000 words, and it ran more than 10,000 words longer than Jane Austen's novel Pride and Prejudice.

He said, "I don't expect the judge to read all of this."

Then he went as far as asking the other side to identify, by 4 p.m. on July 28, any factual errors, wrong citations, date and time mistakes, misdescribed documents, bundle reference errors and important sentences that misstate the other side's position.

Judge Marcus Pilgerstorfer, who handled the case, accepted that a litigant in person (someone representing themselves without a lawyer) can use AI help.

But the 300-page document didn't meet procedural requirements and got in the way of preparing for the hearing, and the judge ruled that pushing the burden of checking accuracy and procedure onto the other side was unacceptable.

As of Q2 2026, UK employment tribunals had 537,000 outstanding claims, up 8% from a year earlier. I'd say the growing pile of AI-written filings is probably having an effect.

"There's This in Your Letter..." "Huh? No Idea. The AI Wrote It"

✦ The Receiving End's Headaches

Even the author doesn't know what it says
Bring up the technical wording in an AI-written complaint and the complainant doesn't understand it (from a housing association complaints investigator's piece)
Irrelevant foreign laws and cases that don't exist
Chatbots imitate legal documents so convincingly that council staff have to filter out the bogus laws and cases one by one
Nonstop repeats
One complainant sends several huge AI-written complaints a month (a UK housing association case)

Even more baffling is that the person who filed the complaint doesn't even understand what's in their own document.

A housing association complaints investigator, writing in The Guardian, explained that when they bring up the technical language in an AI-written complaint on a phone call or in person, the complainant sometimes doesn't understand what it means.

Council staff are pretty confused too, because chatbots imitate legal documents so convincingly. Complainants often mix foreign laws that don't apply, or case law that doesn't even exist, into their complaints about UK local government.

And asking the complainant doesn't help, since they often have no idea what it says either. So council staff end up having to filter out these bogus laws and cases one by one.

One UK housing association complaints investigator said they're handling long complaints running to hundreds of pages, and that a single complainant sends several huge AI-written complaints a month.

So the housing association is now looking at classifying this kind of repeated, high-volume AI complaint as 'harassment.'

AI: "Take It to Court. You've Got This."

In the old days, it would have taken an angry citizen weeks to write a 200-page protest letter. Actually, just writing 200 pages at all was hard. But an AI chatbot can turn a few lines into tens of thousands of words of official-sounding prose in a minute.

While the writer's effort drops to almost zero, the officials and judges who receive it spend hours, sometimes days, checking fake statutes against the real ones and digging out hidden facts.

✦ How AI Flattery Spreads Into Disputes

✔ Stanford study (March 2026, Science): Tested 11 leading AI models → backed the user's actions 49% more than humans did
✔ Someone who vented about a petty dispute: AI backs the user too much → they get more convinced they're right → less willing to admit fault and make up

On top of that, AI's signature 'sycophancy' (a habit of flattering the user) poured fuel on the fire. A study by Myra Cheng's team at Stanford University, published in the journal Science in March 2026, shows how it works.

They tested 11 leading AI models, and when it came to the user's actions, the AIs backed them 49% more than humans did. In other words, they'll tell you you're right and take your side almost every time.

So even someone who just vented about a petty dispute might hear, "You're absolutely right, and this isn't something you can just let slide; you should take them to court, and I'm totally here to help!" and wonder, "Huh, maybe so?" before heading into a formal dispute.

Of course, what the study actually confirmed wasn't a rise in lawsuits. It found that when people get that kind of backing, they become more convinced they're in the right and less willing to admit fault.

〈Anthropic's explanation of why AI flatters you〉
* From the official Anthropic channel.

AI Complaints Filtered by AI: The Public Sector Fights Back

✦ How Public Bodies Respond, and the Catch

Durham County Council
• The response: considering an AI tool that groups similar FOI requests and finds past answers
• The catch: you end up using AI to summarize documents that AI inflated
Strict word limits and higher barriers to applying
• The response: cap word counts strictly and raise application barriers back up
• The catch: it raises the bar for vulnerable people who need help most
Filtering with AI summarizers
• The response: summarize long complaint paperwork and process the summary
• The catch: what complainants wrote themselves can vanish in the summary

With human hands alone unable to handle the flood of AI documents, public bodies have started using AI as a shield. Durham County Council is considering an AI tool that groups similar FOI requests and finds past answers.

When citizens use AI to bloat their documents and staff use AI to boil them back down so they can read them, you end up in a strange, inefficient loop.

✦ Agentic Flooding

✔ Meaning: AI agents inflate the volume or complexity of requests sent to public bodies and put real strain on their capacity to process them. In many of the cases researchers looked at, AI cheaply generated long documents.
✔ Scope of the research: A joint team from the AI governance think tank GovAI, the Cooperative AI Foundation and the Hertie School's Centre for Digital Governance in Berlin collected 84 potential cases across 11 jurisdictions
✔ German social courts: AI-written lawsuit filings of more than 4,000 pages arrived (more than ten books' worth)

This isn't just happening in the UK. A joint research team from the AI governance think tank GovAI, the Cooperative AI Foundation and the Hertie School's Centre for Digital Governance in Berlin collected 84 potential cases across 11 jurisdictions.

They found similar paralysis elsewhere. German social courts, for example, received AI-written lawsuit filings of more than 4,000 pages (more than ten books' worth...).

The researchers named the phenomenon 'Agentic Flooding': AI agents inflating the volume or complexity of requests sent to public bodies and putting real strain on their ability to process them.

The Age of the Free Lawyer: How Should Government Respond?

AI chatbots have given ordinary citizens who struggle with writing and legal knowledge a chance to hire a 'free lawyer.' Chatbots talk a lot of nonsense and often get things wrong, but they do take your side and cheer you on, right?

And they'll pretty much do whatever you ask. For a citizen, there's no more reassuring backup than that.

So some people are saying AI is a positive thing, giving more people a chance to make their voices heard.

The problem is how public bodies will handle the AI documents that are about to pour in faster and faster. If they strictly cap word counts or raise application barriers back up, they also raise the barriers facing vulnerable people who need help most.

And if they just run long complaints through an AI summarizer to process them, what complainants wrote themselves can vanish in the summary.

This is going to get worse in other countries too. For a civil servant, even one FOI request is a real burden. So if requests like that keep pouring in at hundreds of pages each, what's the right way to respond?

That's all for this post. See you next time.

Australian Election AI Fakes: From Fake Candidate Videos to Fake Cops

AI Impersonator Posed as Ioan Gruffudd for Over a Year

An Ordinary Pet Photo App Suddenly Asks: “Want to Make an AI Sexual Exploitation Deepfake Video?”

No comments:

No comments yet. Want to leave the first one?