How to draft a polite complaint email with AI, without oversharing

In one minute

  • You will be able to: turn an angry draft or a few notes into a short, polite complaint email that says what went wrong and what you want.
  • You need: any AI chatbot (a free version is enough) and your receipt, order number and dates to hand.
  • Time: about 15 minutes, most of it checking the draft.

Why it is worth doing

A good complaint is short, calm and specific, which is hard to write when you are annoyed. An AI chatbot is good at exactly this kind of rewrite: taking messy notes or a cross first draft and turning it into something polite and well ordered.

What it is not good at is knowing your facts or your rights. It can get dates wrong, invent a law, or promise a refund you are not owed. So you stay in charge of the facts, and the chatbot only handles the wording.

You can use any general chatbot for this, for example ChatGPT, Claude, Google Gemini or Microsoft Copilot. All four can be used without paying, though free versions have usage limits, so check each maker’s pricing page.

The steps

  1. Write your facts down first, without the chatbot. In a note, list: what you bought or booked, the date, the price, what went wrong (one line per problem), what you have already tried, and what you want (refund, repair, replacement or an apology). The US Federal Trade Commission’s sample complaint letter suggests the same core facts: the product, the date and place you bought it, the problem, and “what you want, like a refund, repair, exchange, or store credit”.
  2. Take out anything private. Before you paste, remove or replace your full name, home address, phone number, card or bank details, account passwords and any order number you would not want stored. Use placeholders such as MY NAME and ORDER NUMBER in capitals.
  3. Consider switching off training on your chats. Most makers let you stop your conversations being used to improve their models. According to OpenAI’s Data Controls FAQ, as of 16 Nov 2026, you go to Settings, then Data Controls, and turn off “Improve the model for everyone”. According to Anthropic’s privacy page (updated 3 Aug 2026), the toggle is “Help Improve our AI models” under Settings, then Privacy. Google says in its Gemini Apps Privacy Hub, as of 16 Nov 2026, that with Keep Activity off your future chats “won’t be used to train our AI models”, but are still kept for 72 hours. For Copilot, check Microsoft’s privacy controls page, as the steps differ between versions of the app.
  4. Ask for a polite draft. Paste your cleaned-up notes with a short instruction, for example: “Write a polite, firm complaint email from these notes. Under 200 words. Say what went wrong, list the problems as short points, say clearly what I want, and ask for a reply within 14 days. Do not add any facts that are not in my notes.”
  5. Check every fact against your notes. Read the draft line by line. Correct any date, price or detail the chatbot changed or added. Delete any sentence that claims something you cannot prove.
  6. Check any mention of the law yourself. If the draft names a law or says you are “entitled” to something, check it with an official source before you send it. In the UK, Citizens Advice has free template letters that name the right law, such as the Consumer Rights Act 2015. In the US, the FTC’s sample letter is a good starting point. If unsure, leave the legal claim out.
  7. Finish it in your email app. Copy the text into a new email, put your real name and order number back in, attach copies (not originals) of receipts or photos, and keep a copy of what you sent. The FTC advises: “Keep the original documents.”

A worked example

Notes pasted into the chatbot:

Bought a kettle from SHOP online 2 Nov, £34.99, order ORDER NUMBER. Stopped working 12 Nov. Emailed once on 13 Nov, no reply. Want a refund or a replacement.

Instruction: “Write a polite, firm complaint email from these notes. Under 150 words. Ask for a reply within 14 days. Do not add facts.”

A good draft will open with the order and date, say the kettle stopped working after ten days, mention the unanswered email of 13 November, ask for a refund or replacement, and ask for a reply within 14 days. If it adds a line such as “as required by law, you must refund me in full”, check that against an official source or cut it.

Watch out for

  • Invented facts. Chatbots sometimes fill gaps with confident guesses (often called hallucination). Your notes are the source of truth, not the draft.
  • Made-up legal claims. A wrong legal claim can weaken your complaint. Check it or remove it.
  • Too much politeness or too much anger. Ask for “polite but firm”. If the draft grovels or threatens, ask it to rewrite.
  • Your data goes to the company that runs the chatbot. Everything you paste is sent to the maker’s servers, even when training is off. Google, for example, says chats may be reviewed “with help from human reviewers” and asks you not to enter “confidential information that you wouldn’t want a reviewer to see”.
  • Money, legal and medical complaints. An AI draft is not legal advice. For a large sum, a dispute that is going nowhere, or anything involving your health or a contract, talk to a free advice service (such as Citizens Advice in the UK or your state consumer protection office in the US) or a qualified professional.

Go deeper

Sources


Related