A new cost asymmetry
Before generative AI, a long, polished legal letter imposed meaningful costs on its author. Someone had to investigate the facts, organise the chronology, choose the arguments, and write the document. Today, a party can produce the appearance of that work in minutes.
The recipient cannot safely assume that the result is irrelevant simply because AI helped write it. A lawyer, employer, supplier, insurer, or former business partner may still have to identify every allegation, verify every date, check every cited authority, and decide which point changes the legal or commercial position. The cost of producing an argument collapses; the cost of responsibly evaluating it does not.
From workslop to dispute correspondence
BetterUp and the Stanford Social Media Lab use the term “workslop” for AI-generated work that looks finished but leaves the real thinking and clean-up to someone else. Their September 2025 survey of 1,150 full-time US desk workers reported that 40% had received workslop in the previous month, with respondents estimating an average of two hours to resolve an incident and a monthly cost of $186 per employee.
Those figures describe the workplace generally, not legal disputes specifically. But the transfer mechanism is the same. A polished surface can conceal missing context, invented detail, duplicated arguments, or conclusions that no person has checked. In an ordinary workflow that creates rework. In a dispute it creates review risk, defensive cost, and pressure to respond in kind.
Why disputes amplify the burden
Conflict makes recipients cautious. A six-page complaint may contain only three material facts, but the recipient does not know that until someone has read all six pages. A fictitious case citation may still require a database search before it can be rejected. A vague accusation may force a search through months of messages and invoices.
Industry reporting now describes clients sending lawyers extensive AI-generated questions, strategies, and documents that require more review rather than less. The Law Society Gazette has described an emerging “AI slop” problem in courts, particularly where plausible but unreliable submissions still have to be processed by judges and lawyers.
- More words create more propositions to verify.
- Formal tone can make weak claims appear more authoritative.
- Every answer becomes fresh input for the next generated counter-answer.
- Human attention remains the scarce and expensive resource.
The AI-versus-AI escalation loop
Once both sides use AI, the marginal cost of another round approaches zero. One party generates a comprehensive response. The other asks its model to find omissions, contradictions, and stronger language. That output returns to the first model, which discovers another set of objections. Each round appears rational in isolation; together they form a negative-sum loop.
Research on difficult workplace emails shows that LLM-assisted writing can be more formal and empathetic, and that hybrid human-plus-AI drafting can outperform either alone in some scenarios. The problem is therefore not “AI writing is always bad”. The problem is using generation without responsibility for relevance, accuracy, proportionality, and closure.
Not every long AI-assisted message is abuse
A consumer, employee, or small business may use AI because professional help is unaffordable, because they are writing in a second language, or because they need assistance organising a complex history. Length alone does not prove bad faith, and the use of AI should never automatically decide a dispute against its user.
The fair question is functional: does the submission identify verifiable facts, relevant evidence, and a concrete requested outcome? If apparent manipulation, fabricated material, or deliberate volume tactics are alleged, attribution and context must be established and the affected party must have an opportunity to explain.
A better response than another twenty pages
The answer to correspondence inflation is not to forbid useful drafting tools. It is to change the communication architecture. A dispute process should extract a chronology, separate allegations from evidence, identify what each party actually wants, and close factual issues when the record is complete.
That is the category din.org calls the AI email war: conflict in which generating more language is cheap but reaching a reliable decision remains expensive. The purpose of structured online dispute resolution is to stop rewarding volume and redirect both sides toward a finite, reviewable record and a concrete path to settlement.