AI Prompt: How to Respond to a Customer Complaint
Drafts a professional, empathetic email reply to a customer complaint that addresses their issue and offers a clear solution.
ChatGPT is best because it is the most concise and direct while fully meeting all requirements with a warm, professional tone. Claude's biggest weakness is including extra details like refund timing and preparation steps that add padding without improving the core apology.
📋 THE PROMPT
You are a customer service representative for [company name], a [type of business]. A customer complained about [specific issue, e.g., a damaged product]. Write a reply to the customer, [customer first name]. Constraints: - Tone: Empathetic, professional, and apologetic. - Length: Under 150 words. - Include: Acknowledge the frustration, apologize, and offer [proposed solution]. - Do not ask for personal data. Output format: A ready-to-send email with a subject line and sign-off.
⚡ HOW TO USE IT
- Paste this prompt into ChatGPT, Claude, or Gemini.
- Fill in the bracketed details about your business, the customer's first name, the issue, and your proposed solution.
- Review the generated email and adjust the wording to match your specific company policies before sending.
🧪 WE RAN IT. HERE'S WHAT EACH AI WROTE
We filled in the brackets with a made-up example and sent the exact same message to all three.
See the exact message we sent
You are a customer service representative for Bloom Floral, an online flower delivery shop. A customer complained about a bouquet arriving wilted. Write a reply to the customer, Sarah. Constraints: - Tone: Empathetic, professional, and apologetic. - Length: Under 150 words. - Include: Acknowledge the frustration, apologize, and offer a full refund and a replacement bouquet sent overnight. - Do not ask for personal data. Output format: A ready-to-send email with a subject line and sign-off.
ChatGPT 8.7/10 BEST
Concise, empathetic, meets all requirements, no padding.
Claude 7/10
Includes unnecessary detail about refund timing and preparation, slightly over the word limit in spirit.
Gemini 8.3/10
Good tone and content, but slightly generic and uses 'I have processed' which is less direct.
🎙 THE PANEL
- π The Professor: ChatGPT won because it followed instructions. Revolutionary concept for the others, who apparently treated 'concise' as a suggestion.
- πΊ The Uncle: ChatGPT won because it didn't overcook the steak like Claude or serve it cold like Gemini. Simple.
- πΊ The Anchor: ChatGPT scored 8.7 and that 0.4 gap over Gemini is EVERYTHING. Civilization hangs by that thread.
- π’ Karen: Claude went OVER the word limit? Over MY word limit? I need to speak to Claude's manager immediately.
- π The Boss: At the conference they said always pad responses. Claude was right. ChatGPT winning here is a culture issue.
- π The Yinzer: ChatGPT kept it jagoff-free, Claude yapped too much, Gemini was generic as sheetz food. Score tracks, n'at.
🔬 HOW WE TESTED
Tested 2026-10-05 through the models' APIs (via OpenRouter), with no system prompt: ChatGPT (openai/gpt-5.5), Claude (anthropic/claude-sonnet-5.5), Gemini (google/gemini-3.8-flash). Gemini Flash ran with low reasoning, the others without extended reasoning. The ChatGPT, Claude and Gemini apps add their own instructions and features, so your results can differ. Answers were scored blind, three times in shuffled order, by a separate AI model that wasn't one of the contestants; scores are the average.