How to Break Bad News Over Text (And Whether You Should Let AI Help)

Originally published April 2016. Updated September 29, 2026 with new messaging data and guidance on using AI.

Does anyone really know how to break bad news over a text? Ten years after I first wrote about this, the question has a new wrinkle: should you let AI write it for you?

Let me start where I started then. A call from a certain person who tends to be the bearer of bad news. I saw the missed call, felt my stomach drop, and sent a text: “Saw you called. Is everything alright?” What came back was the bad news, cold, with no warm up, no formality, nothing to soften it. It was accurate. It was also careless, and it stayed with me longer than the news itself.

Texting is convenient, and it is a permanent record. A screenshot travels farther than any phone call ever will.

When I first wrote this in 2016, Americans sent about six billion texts a day and I thought that was a lot. Ten years on, the world sends around 25 billion SMS messages a day, and that is the small number. WhatsApp alone moves more than 100 billion messages every day, over a million a second, and it passed the entire global text network back in 2019. Between WhatsApp, Messenger, and WeChat, humanity now fires off close to 185 billion messages a day. That is nearly one message an hour for every person alive, and a fair share of those are now being drafted by an AI before a human hits send. More words than ever, moving faster than ever, and every one of them a permanent record. Grandma Agnes would have had thoughts.

The rules my grandmother Agnes read to me from Amy Vanderbilt’s Complete Book of Etiquette have not expired. They have become more important, because the message now lives forever.

Here is how I handle it, updated for a world where the draft might come from a machine.

1. If you can, pick up the phone

Text loses tone. Bad news delivered by voice gives the other person a chance to ask a question and hear that you care. Open with “Is this a good time?” It is a small courtesy that changes everything after it.

2. If it has to be text, warm up first

Never lead with the blow. “John, I have great respect for the work you are doing. I have been trying to reach you and want to let you know that Richard has moved to another firm. Can we find time to talk this week?” Respect, then the news, then a next step.

3. Let AI read it before you send it, not write it for you

This is the new rule. I use AI every day, and here is the honest truth: an AI written text about hard news reads like an AI wrote it. People can feel it, and it makes the news land colder. What AI is brilliant at is reading tone. Before I send anything difficult, I paste my draft in and ask: “Here is a message I am about to send about [situation]. How will this person feel reading it? What could be misread? Suggest one change that makes it kinder without hiding the truth.” Thirty seconds, and it has saved me from more regret than I can count. Your words, checked by a machine, sent by a human.

4. Never let AI break the news on your behalf

If the message matters, the other person deserves to know it came from you. Auto replies, AI agents, and scheduled messages are for logistics. They are not for telling someone they lost a client, a role, or your trust.

5. Follow up with an email

Email is still the record of business. Anything corrective goes in writing, kept and backed up. A text is where the news lands; the email is where it lives.

A prompt you can use today

Copy this into ChatGPT or Claude before your next hard message:

Here is a text I am about to send: [paste it]. The situation is [describe it]. Tell me how the recipient is likely to feel, what could be misread, and rewrite it once in my voice so that it is honest, respectful, and leaves the relationship intact.

That prompt is one of ten I use to run my businesses. If you want the full set, the Top Ten Prompts are free on my homepage.

The machines bring the speed. You bring the manners. Nobody automates that.