AI disclosure

What the AI does

We use a large language model from Anthropic (Claude) to write the text of your letters and guides from your answers. It also powers two helpers in the form: “Make this clearer”, an optional button that suggests neutral wording for your descriptions, and a suggested summary of each uploaded file, written from the file’s text when you upload it. Both are labelled as AI suggestions, and you can edit or replace them.

What our software does instead of the AI

• Names, addresses, dates and the letter layout come straight from your answers, not from the AI.
• Deadline dates, statutory interest and collection costs are calculated by our code from the rules of your jurisdiction. The AI never does arithmetic.
• Legal references come only from our jurisdiction rules, and only those a person has verified against official sources may appear in a document.

The automated checks

• A second AI pass reviews every letter against your answers: facts you did not give, legal references that are not verified, unlawful threats, the wrong tone, missing required elements.
• Rule-based checks in our code look for the same problems independently, such as threats of criminal prosecution or contacting an employer.
• If a problem is found, the letter is rewritten once with the problems pointed out. If a serious problem remains, we do not deliver the letter and refund you in full.

What we do not do

• No lawyer reviews your documents.
• We do not give legal advice or tell you whether to sue.
• We do not send anything to the other party.

Before paying you confirm: “I understand that my documents are generated by an AI system with automated checks, that no lawyer reviews them, and that they are document drafts and general information, not legal advice.”

Mistakes

AI systems can make mistakes. Where a fact is missing, the letter says [MISSING: …] instead of guessing. Read every document before you send it. If you find an error we should know about, write to support@example.com.