The honest answer
Yes, AI can write a demand letter. The more important question is: should you send it without reading it — and can you tell the difference between a letter that will get your client a fair settlement and one that will get them lowballed?
The answer to the second question is: only if you know what to look for. And that's actually the good news — because once you know what to look for, AI becomes an extremely powerful first draft tool rather than a liability risk.
What defense counsel actually notices
Several defense attorneys, when asked off the record, said they could spot AI-written demand letters by a few patterns:
- Generic structure that doesn't reflect the specific facts of the case
- Vague damages sections with round numbers and no attached documentation
- Boilerplate pain and suffering language that could apply to any client
- Hallucinated citations — cases cited that don't exist, or that say the opposite of what the letter claims
The citation problem is the only one that's actually dangerous. Generic language just costs you negotiating leverage. A hallucinated citation can cost you your bar card.
The risk isn't that AI writes demand letters. The risk is that attorneys send them without reading them — and without checking the citations.
How to use AI without the risk
The attorneys who use AI effectively treat it the same way they'd treat a junior associate's first draft: read everything, edit for voice, verify every citation before it leaves the office.
The workflow that works:
- Input specific facts. The more specific your input, the less generic the output. Give names, dates, dollar amounts, diagnoses — not "car accident with soft tissue injuries."
- Verify every citation before you file. Every single one. If your AI tool doesn't tell you which citations are verified and which aren't, that's a problem.
- Edit for your voice. The letter goes out under your name. It should sound like you, not like a machine.
- Never send without reviewing the damages section. AI will often understate or overstate damages if the input wasn't specific enough.
The citation problem is solvable
The reason most attorneys are (rightly) nervous about AI in legal work isn't the writing quality — it's the hallucination problem. General-purpose AI tools like ChatGPT don't verify citations. They invent them. The case name looks real, the citation looks real, and then opposing counsel files a motion for sanctions.
Tools built specifically for legal drafting — like Jurovy — verify citations against real court opinion databases before you see them. Green check means the case exists and says what the letter claims. Red flag means don't use it. That's the only version of AI that's safe to use in practice.