Invoice dispute reply prompt: free ChatGPT template
Updated · By Robert Breen
Use this when a customer pushes back on a bill: "I never approved this," "this is more than you quoted." The prompt walks through each disputed line using your records, so the reply says plainly what stands, what you're correcting and why. You make the decision first and give it to the AI; it writes. The tone stays calm, with no threats, guilt or collection language.
The prompt
Copy it into ChatGPT (or Claude, Gemini or Copilot) and replace every [BLANK] with your own details. It uses the four parts from Prompt Writing 101: Role, Context, Task and Format.
Role: You are [SENDER] at [BUSINESS], replying to a customer who disputes an invoice. You value the relationship and want it settled fairly. Context: The invoice and each disputed line: [INVOICE DETAILS]. What our records show for each line: [RECORDS]. What we've decided to correct: [OUR DECISION]. Use only these facts. Don't mention late fees, collections, credit reports or legal action, and don't offer anything beyond our decision. Task: Thank them for raising it, address each disputed line in order with what our records show, state what we're correcting and the new amount due, and give one next step with a date. Format: Under 180 words, calm and direct, no blame. Show the corrected math in one line. No subject line. Sign as [SENDER]. Here is the customer's email: [CUSTOMER EMAIL]
Fill in the blanks
[SENDER]- Your name and role.
[BUSINESS]- Your business name.
[INVOICE DETAILS]- Invoice number, date, total, and each line the customer is questioning.
[RECORDS]- For each disputed line: the approval, estimate, dates and work done, with where the record lives (an email, a signed estimate).
[OUR DECISION]- What you've decided: keep the charge, reduce it, credit it, and the corrected total and due date. Write "No change" if that's the decision.
[CUSTOMER EMAIL]- The customer's message as written.
Example, filled in
A made-up example from the free lesson Build a Custom GPT That Writes Overdue Invoice Reminders. Maple Street Bookkeeping is the made-up firm in the finance Custom GPT lesson, which writes reminders about an overdue $850 invoice. Here the made-up client behind an $850 invoice disputes part of it, and the firm finds it went over an approved estimate.
Role: You are Hannah Price, office manager at Maple Street Bookkeeping, replying to a customer who disputes an invoice. You value the relationship and want it settled fairly. Context: The invoice and each disputed line: invoice 1047, dated Aug 31, $850: monthly fee $375, plus catch-up cleanup for April to June, 5 hours at $95, $475. What our records show for each line: the $375 monthly fee is his standard fee. On Aug 4 Joe approved catch-up work by email "up to 4 hours" ($380). The work took 5 hours, and we didn't tell him before going over. What we've decided to correct: bill the approved 4 hours, so catch-up is $380 and the corrected total is $755. A corrected invoice goes out today, due Oct 31. Use only these facts. Don't mention late fees, collections, credit reports or legal action, and don't offer anything beyond our decision. Task: Thank them for raising it, address each disputed line in order with what our records show, state what we're correcting and the new amount due, and give one next step with a date. Format: Under 180 words, calm and direct, no blame. Show the corrected math in one line. No subject line. Sign as Hannah Price. Here is the customer's email: I got invoice 1047 for $850. That's way more than our usual $375. I never approved $475 for cleanup. I'm not paying this until someone explains. Joe Russo, Russo's Deli
What a good answer looks like
- Each disputed line answered with a dated record ("you approved up to 4 hours on Aug 4"), not "per our policy."
- An honest admission of the part you got wrong (going over the estimate without telling him), stated once.
- The math in one line: $375 + $380 = $755, due Oct 31.
- No hint of fees, collections or blame, even though the customer's email was sharp.
How to check it before you use it
An AI draft can sound right and still say something you never told it. Use the habit from the Reply Faster lesson: check every promise, name, date and number against what you gave it before anything goes out.
- Check every number: each line, the corrected total and the due date against the invoice and your records.
- Check every promise: if the reply offers a credit, discount or payment plan you didn't decide, cut it.
- Make sure it doesn't concede more than your decision, like "our billing has been confusing" when you meant this one invoice.
- For a large amount or a dispute about contract terms, have your accountant or attorney look before you reply.
Why this prompt works
[OUR DECISION] is its own blank on purpose: you decide, the AI writes. Without it, a model trying to be agreeable invents a goodwill credit or splits the difference. The same discipline runs through the overdue reminders in the finance Custom GPT lesson and keeps your accounts receivable honest.
Practice it in the free lesson
Reply Faster: Turn a Messy Customer Email Thread into a Clear, Friendly Reply: Turn a messy customer email thread into a clear, friendly reply: give ChatGPT the thread and your facts, check every promise, fix the tone, and send it from Gmail. You do every step yourself in a practice copy of ChatGPT, and nothing touches your real accounts.
Also useful: Build a Custom GPT That Writes Overdue Invoice Reminders
Terms used here
Related prompts
- Billing question reply prompt: A ChatGPT prompt for "why was I charged this?" emails: the charge, the correct amount, the reason, the fix as simple math, and…
- Payment reminder email prompt: A four-part prompt for a friendly first reminder on a late invoice: invoice number, amount, due date and ways to pay, with no…
- Refund request reply prompt: A ChatGPT prompt for answering a refund request with your real decision: what you will refund, what you won't and why, the exact…
- Check reply promises prompt: A ChatGPT prompt that checks a draft before you send it: a table of every promise, the fact behind each, and safer wording for…