humanflow

How to humanize an AI-written email

Delete the first paragraph. AI emails almost always open with a sentence restating why they are writing, which the subject line already said. Start at the actual point, keep sentences short, and end with one clear request. Most AI email drafts improve by losing a third of their length.

Typical length · 80–200 words. Well inside the free plan. · Last reviewed 16 August 2026

Before and after

Before

I hope this email finds you well. I am reaching out to you today regarding the upcoming project deadline that we discussed in our previous meeting. I wanted to touch base with you to see if there have been any updates.

After

Quick check on the Thursday deadline — are we still on track? If the API work has slipped, I would rather move the date now than the day before.

Forty-four words became twenty-nine. Two sentences of preamble disappeared, and the email now contains an actual decision the reader can respond to.

The first paragraph is almost always deletable

AI-written email has one dominant tell and it is structural: an opening sentence that restates the reason for writing. The subject line already carried that, so the reader has been told the same thing twice before reaching anything actionable.

In the example above, two sentences of preamble disappear and forty-four words become twenty-nine. Nothing was lost — the deleted portion contained no information, only the shape of an email.

Try it as a habit rather than an edit: write the message, then delete the first paragraph and see whether anything is missing. It usually is not.

Direct is not blunt, and the difference is cultural

This is the one place where the general advice on this site needs qualifying. Cutting courtesy is not automatically an improvement, and expectations vary sharply between workplaces, industries and countries — an opening that reads as efficient in one context reads as curt in another.

The distinction worth holding is between courtesy that carries meaning and courtesy that fills space. "I know this is short notice" acknowledges something real. "I hope this email finds you well" acknowledges nothing, which is why it survives translation into every AI draft.

If you write across cultures, calibrate to your reader rather than to the shortest possible version.

Nobody is running a detector on your email

Worth saying plainly, because the anxiety transfers from academic contexts where it belongs. Ordinary workplace email is not screened for AI authorship, and the reason to edit a draft is that the unedited version wastes the reader's time.

That makes this the one format on this site where the argument is purely about quality. The rewrite is worth doing because a twenty-nine-word email containing a decision gets answered faster than a forty-four-word email containing a preamble.

Formats with the same problem

Cover letters longer, higher stakes, and the same trap of sounding like everyone else.

Press releases what happens when the reader is a professional scanning for the news.

Questions

How do I make an AI-written email sound less like AI?
Delete the first paragraph. The dominant tell is an opening sentence restating why you are writing, which the subject line already said. Then check that the message contains something the reader can respond to — a decision, a question, or a date.
Is 'I hope this email finds you well' really a problem?
It is filler rather than a problem, and it is the single most reliable marker of an unedited draft. The useful distinction is between courtesy that carries meaning — acknowledging short notice, for instance — and courtesy that only fills space.
Does anyone check work emails for AI writing?
Not in any ordinary workplace. The anxiety comes across from academic contexts where detection is real. Here the only reason to edit is that a shorter, clearer email gets a faster answer.

What to watch for

  • Direct is not blunt. Keep the courtesy that carries meaning and cut the kind that does not.
  • Some cultures and workplaces expect a warmer opening — adjust rather than apply this everywhere.

If your writing gets flagged

Rewriting for rhythm and specificity tends to lower detection scores, because that is what detectors read as human. It is not a guarantee — detectors disagree with each other and change without notice, and we do not promise a result from any of them.

Why human writing gets flagged →

Other use cases