Customer Complaint with AI: Getting a Refund

You’re waiting for a refund that never comes, a defective product hasn’t been replaced, or a service wasn’t delivered as promised. Writing an effective complaint requires precision: the right facts, the right legal references, the right tone. AI can do this work in minutes, provided you give it a well-constructed prompt. Here’s the method, copy-paste prompts, and how to compare multiple AI models to keep the most convincing version.

Why AI changes the game for complaints

A complaint that gets you a refund is neither aggressive nor vague. It’s factual, it cites precise legal grounds, and it sets a clear deadline. That’s exactly what AI knows how to produce quickly, as long as you feed it the right information. It also avoids the two classic pitfalls: too soft a tone that doesn’t motivate action, and too aggressive a tone that puts the other party on the defensive and slows down your case.

The method in 5 steps

1. Gather evidence before writing

Before launching a prompt, collect: the order or contract number, purchase date, amount paid, previous exchanges with customer service (screenshots, emails), and a factual description of the problem (defective product, undelivered shipment, service not rendered, cancelled order not refunded). The more precise your facts, the stronger your AI-generated complaint will be.

2. Identify the applicable legal framework

In France, several legal bases may apply depending on your situation:

• Legal warranty of conformity (articles L217-3 to L217-14 of the Consumer Code): a defective or non-conforming product must be repaired, replaced, or refunded.
• Right of withdrawal (article L221-18) for a distance purchase: full refund within 14 days of withdrawal.
• Breach of contract (article 1231-1 of the Civil Code): if a paid service wasn’t delivered, you can demand a refund or compensation.

You don’t need to be a lawyer: AI can identify the relevant article if you describe your situation precisely.

3. Write with a structured prompt

This is the key step. A vague prompt produces a vague letter. A precise prompt produces a letter that hits the mark.

4. Adjust the tone: firm but courteous

Once the first draft is generated, ask for tone variations. The goal is to stay professional while showing you won’t drop the matter.

5. Proofread, verify facts, then send

AI can get wrong a law article or a deadline. Always verify the legal references cited before sending, especially if the amount at stake is significant. It’s the same logic as verifying your sources with Perplexity: AI speeds up writing, not verification.

The copy-paste prompt

Here’s a typical prompt to adapt with your information:

“You are an expert in French consumer law. Write a complaint letter to the customer service of [company name], firm but courteous, to request reimbursement of [product/service] ordered on [date] for an amount of [amount] €. Problem encountered: [factual and precise description]. Available evidence: [order number, screenshots, previous exchanges]. Cite the relevant articles of the Consumer Code or Civil Code depending on the situation. Request a refund within 14 days, with wording indicating possible next steps (mediation, DGCCRF) in case of no response. Structure the letter as: subject line, recall of facts, clear request, response deadline, firm closing formula.”

To refine, then add: “Make the tone firmer without being aggressive” or conversely “Soften the tone slightly while keeping the refund request explicit”.

Compare multiple AI models to keep the most convincing version

Not all AI models have the same writing style. ChatGPT tends to produce structured and neutral letters. Claude often refines tone nuances. Gemini gets straight to the point. Running the same prompt on multiple models allows you to spot the most persuasive wording, the one that balances firmness and courtesy without losing clarity.

It’s the same reflex as reviewing a freelance contract with Claude or Gemini: you don’t stop at the first response, you compare. Using multiple AI at the same time becomes a real advantage to get the most effective text, not just the fastest to generate.

With try noov.ai for free, you run the same prompt on ChatGPT, Claude, Gemini, DeepSeek, and Perplexity from a single interface, and compare versions side by side in seconds, without switching tabs or creating a new account each time.

Example of structure to aim for

An effective complaint typically follows this outline:

1. Subject line: “Refund request — Order No. XXXX”
2. Recall of facts: date, amount, nature of problem, in 3-4 sentences maximum.
3. Legal basis: the article cited, in one sentence, without excessive jargon.
4. Precise request: the exact amount to refund and your preferred method.
5. Deadline and consequences: “Without your response within 14 days, I reserve the right to contact the consumer ombudsman.”
6. Firm closing formula: professional, without familiarity or aggression.

Mistakes to avoid

• Being too vague about facts: AI can’t invent solid evidence, give it exact details.
• Citing a random article: always verify the legal reference before sending.
• Threatening without grounds: mentioning legal action without foundation weakens your credibility.
• Sending without proofreading: even AI-generated, read the letter aloud before sending to verify it sounds right.

The logic is close to that of a well-written customer follow-up email: clarity and firmness do more to get a quick response than any threat.

In summary

An effective complaint rests on three pillars: precise facts, correct legal grounds, and a firm yet courteous tone. AI speeds up writing if you provide these three elements in your prompt. Comparing multiple models allows you to refine the final text before sending it and increase your chances of getting a quick refund.