How to stop asking AI before every email
Before you open the AI, write the point of the email in one sentence. Then give yourself three minutes to draft it. If it is five sentences or fewer, send it as it is. Keep the AI for the few emails that carry real risk, and use it there as a reader of your draft, not as the writer.
Why it became a reflex
Email is where the habit usually starts. The stakes feel social, the blank reply box is uncomfortable, and the AI is one click away. So a two-line reply goes through a chatbot "just to check the tone". After a few months, sending anything unchecked feels risky.
It is rarely the writing that is hard. It is deciding what you want to say. The AI cannot decide that for you. It can only guess, in polished sentences. That is why AI-drafted emails so often need three rounds of "make it shorter" and "less formal".
There is no study of email habits in our research list, so we will not quote one. The nearest finding is a survey of 319 knowledge workers (Lee and others, Microsoft Research and Carnegie Mellon, CHI 2025, peer reviewed): people with more confidence in their own skill reported thinking more critically about AI output, and people with more confidence in the AI reported less. It is self-reported and shows a link, not a cause. But it fits: sending your own emails is how you keep the confidence to judge them.
The method
Write the point in one sentence
"I need the figures by Thursday." "I am saying no to the meeting." "I am sorry the order was late and here is what we will do." If you cannot write this sentence, no tool will write the email well. Work this out first.
Draft for three minutes
Set a timer. Point first, then the one or two things the reader needs, then what you want them to do. Plain words. Stop when the timer stops.
Apply the five-sentence rule
Five sentences or fewer: read it once and send it. No AI. Most emails are in this group, and nobody remembers their wording by lunchtime.
Sort the rest by risk
Ask: would a clumsy sentence here cost money, a relationship or my reputation? If not, send it. If so, go to step 5.
For risky emails, ask for a reader
Paste your draft with: "I wrote the draft below myself. Please do not rewrite it and do not give me replacement sentences." Then ask: "Read this as the person receiving it. What might they misunderstand or take badly?" Make the changes yourself.
Count for one week
Keep a tally on paper: emails sent, emails that went through an AI. You are not aiming for zero. You are aiming for the number to mean something.
Small things that help
- Reply in the email window. If you never open the chat tab, you never paste the thread into it.
- Write like you speak. Say the reply out loud, then type what you said. It is nearly always fine.
- Keep your own templates. Three or four replies you send often, in your own words, saved as drafts or signatures. You get the speed without asking anyone.
- Sleep on the hard ones. A difficult email written at 5pm and reread at 9am needs less help than you think.
When it is fine to let the AI write
Plenty of email does not need to be yours: a form reply, a booking confirmation, a long thread that needs summarising, a message in a language you are still learning. Hand those over without guilt. The point is to choose, not to ask by reflex.
Take care with other people's information. Check your employer's policy before pasting customer or colleague emails into any AI tool.
If the reflex keeps winning
Add a small step in front of it. In your AI's settings, a saved instruction can make it ask for your own draft first. This is the work line from ours:
I use you at work. For thinking tasks (plans, analysis, decisions, messages that matter) ask what I think first. For routine chores, just help.
The free coach setup builds the full instruction and shows where to paste it. The think pad gives you a box for your draft and sends it along with your question to the AI you pick.
Questions
Is it bad to use ChatGPT for emails?
No. For routine and form emails it saves time and costs you nothing. The habit worth changing is asking before you have decided what you want to say, on emails you could write in a minute.
How do I trust my own tone again?
By sending. Start with short, low-risk emails written by you alone, and notice that nothing goes wrong. Confidence follows evidence.
English is not my first language. Should I still stop?
Use what helps. Write your own draft first, then ask for a list of errors and unclear phrases, not a rewrite. You keep your voice and still catch the mistakes.
Can people tell when an email was written by AI?
Sometimes, and nobody can tell reliably. The better question is whether the email says what you mean. A short email in your own words usually does that best.