Is an AI-written birthday email cheating? How to use it and still sound like you

Here is the uncomfortable truth about most human-written birthday messages: they are templates too. 'Happy birthday! Hope you have a great day!' has been typed a few billion times. The question is not human or machine. It is specific or generic.

Where AI writing goes wrong

Give a model nothing and it gives you clichés back. 'Another trip around the sun.' 'Words cannot express.' Em dashes everywhere. It sounds like a greeting card because that is the average of everything it read. A generic prompt produces a generic email, and people can feel it.

The three inputs that make it yours

• Facts. Real, small, true things only you would know. 'Drove 200 miles for a lobster roll.' 'Still can't read a map.' Two or three per message is enough.
• Relationship and tone. Your brother gets roasted. Your mother gets sincerity. The same facts read completely differently through each.
• Your sign-off and your review. It goes out under your name, so you read it first. Change one line and it is yours in every sense that matters.

How DontRemember uses it

The model is given your facts, the relationship, the occasion with the number of years, the tone you chose, and a set of rules: first person, no clichés, no mention of apps or reminders, concrete over general, no dashes. It returns a subject and a few paragraphs. You get the draft the night before and can edit or regenerate with a hint. Then it sends from your Gmail. The recipient never sees a service. They see you.

The honest position

If you would otherwise send nothing, or 'hbd!!' at midnight, a warm specific email written from your memories and reviewed by you is better for the person receiving it. If you would otherwise write three heartfelt paragraphs by hand every time, keep doing that. Most of us live in between, and the between is where this helps.