Last updated:
Quick answer
Describe the situation, who it is for, and the tone you want. ChatGPT drafts the text. Then you rewrite it in your own words before sending, because an unedited AI apology reads like an AI apology.
At a glance
All outcome figures are estimates from cited sources, not independent verification by us.
- What it is allowed to do
- Drafts for you
Writes drafts (emails, posts, plans, replies). You review and send. Nothing goes out without you.
- Time saved (estimate)
- Estimate: 20 to 40 minutes of drafting and dread per message
- Difficulty
- Intermediate
- Tools needed
- ChatGPT, Claude
- Typical cost
- Free on ChatGPT and Claude free tiers.
Reviewed Oct 19, 2026
Basis (estimate): No study cited. Estimate based on replacing long blank-page sessions and multiple rewrites with a structured draft.
What does this setup do?
Draft hard messages: apologies that take real responsibility, boundary texts that are firm without being cruel, and difficult work messages you keep rewriting.
Who is it for?
Anyone staring at a blank text box dreading a conversation: apologizing to a friend, telling family no, or pushing back on a coworker politely.
How do you set it up?
Tools
- ChatGPT or Claude (free tier is fine)
- The facts of the situation (what happened, who is involved)
- Your honest goal for the message
Steps
- Write down what happened in plain facts, no spin. What you did, what they did, what you want now.
- Decide the message type: apology, boundary, feedback, or decline.
- Tell the AI the relationship and the tone: warm but firm, professional, brief.
- Get the draft, then read it out loud. If it does not sound like you, it is not done.
- Rewrite at least 30% in your own words. Add one specific detail only you would know.
- For serious apologies, consider calling instead. A text carries minor repairs, not major ones.
- Send it yourself. Never let an agent send personal messages for you.
Example output
Illustrative. Your results will differ.
- Your input turned into a clear, prioritized plan
- Next 3 actions with dates, in priority order
- You stay in charge. The agent just organizes.
Where to paste this prompt
ChatGPT: ChatGPT → Project instructions, or ChatGPT Tasks for scheduled reminders (once a day on Free)
Copy-paste prompt
Adapt bracketed fields to your situation.
Help me draft a difficult message. I will rewrite it in my own words before sending.
Situation: [what happened, plain facts]
Message type: [apology / boundary / feedback / declining]
Recipient: [e.g. "close friend" / "my manager" / "my mother-in-law"]
My goal: [e.g. "take responsibility without groveling" / "say no to hosting Thanksgiving firmly"]
Tone: [e.g. "warm but firm" / "professional and brief" / "honest, not harsh"]
Length: [text message / short email]
Rules:
- Apologies must name what I did wrong specifically. No "sorry if you felt" non-apologies.
- Boundaries must be clear about what I will and will not do, without over-explaining.
- Keep it short. Long messages dilute the point.
- Match the channel to the seriousness: tell me if this deserves a call instead.
- Draft only. I send everything myself after rewriting.Newsletter
Get the free starter pack
Free starter pack with five beginner copy-paste prompts after you confirm, plus three sourced setups in your inbox each week. Unsubscribe anytime.
What usually goes wrong?
Where agents tend to stall on this job. More patterns in Where AI gets stuck.
- Sending the draft as-is. Apology guides specifically warn that unedited AI apologies read as insincere.
- The AI may soften your boundary into meaninglessness. Check that the “no” survived the draft.
- Using text for something that needs a voice. If the harm is serious, the draft should become talking points for a call.
Which chat tools work with this prompt?
Paste the prompt into any of these assistants (free or paid tiers).
- ChatGPT
- Claude
Common questions
- Should I use AI to write an apology?
- As a starting draft, yes. As the final message, no. Use it for structure, then rewrite in your own words with specific details. Unedited AI apologies are easy to spot and read as insincere.
- What makes a good apology text?
- Name what you did specifically, acknowledge the impact on them, skip the excuses, and say what changes. “Sorry if you were offended” is a non-apology.
- How do I set a boundary over text without sounding harsh?
- State what you will do, not just what you will not. “I can’t host this year, but I’ll bring dessert” lands better than a bare no. Keep it short; over-explaining invites negotiation.
What tips make it smoother?
- Honest limitation: AI cannot capture your voice or your history with the person. Its draft is scaffolding. The final 30% rewrite is what makes it yours.
- Recipients can spot unedited AI text, especially in apologies. One specific personal detail fixes most of it.
- For workplace apologies, front-load responsibility and the fix. Nobody wants three paragraphs of context first.
- If you are still angry, draft now and send tomorrow. AI will not tell you to wait, so I will.
Where do the claims come from?
Summaries use public sources linked below, not independent verification. How setups are picked.
- Read Iienstitu’s write-up (www.iienstitu.com)
Iienstitu: Apology structure by severity and channel, with copy-ready scripts.
- Read Revivalist’s write-up (revivalist.com)
Revivalist: Sincere apology frameworks and the warning to edit AI drafts into your own voice.
Related setups
- Bot shared-computer security audit
Before you give Bot Bots more connectors or routines, run a read-only audit of what every Bot on your account can see on the shared cloud computer.
- Route Bot tasks to your Claude and ChatGPT subscriptions
Install Claude Code and OpenAI Codex CLIs on Bot’s shared cloud computer, sign in with your own Anthropic and OpenAI subscriptions, then set routing rules so writing, coding, and browser-heavy jobs hit the model you already pay for, without pasting the setup into Codex or Claude Code directly.
- How to add a shared Bot template when Add Bot fails
Install a public Bot template (for example Side Hustle–style giveaways people share on X) when the in-app “Add Bot” button errors, without copying proprietary template prompts, then onboard the new coworker with minimum connectors and approval gates.