Let AI write your emails (without sounding like a robot)

You’ve tried having AI write an email. The result sounded like a corporate press release crossed with a LinkedIn guru post. Transition words everywhere, no personality, and that characteristic “AI-smooth” tone that makes the recipient close the email after 3 seconds.

That’s not AI’s fault. It’s the prompt’s.

AI is a writing tool — it’s only as good as the instructions you give it. With the right approach, AI can produce emails that sound like you, save you hours per week, and actually convert.

This guide shows you how. From prompt engineering to editing to the situations where you should close the AI and write it yourself.

Why AI emails sound robotic

Before we fix the problem, let’s understand it. AI-generated text has characteristic flaws:

Too polished. AI writes grammatically perfect, well-structured text. But humans don’t write perfectly — we use incomplete sentences, start with “but,” and have personal quirks. Perfection feels fake.

Too many transition words. “Furthermore,” “moreover,” “in addition to this,” “that being said.” AI loves transition words because they create flow. But in short emails, they feel bloated.

Too generic. AI draws on average text from across the internet. The result is average text. It lacks the specific details, personal anecdotes, and industry-specific references that make text interesting.

Too long. AI wants to be thorough. You asked for an email — you get 500 words. A good email is often 100-200 words.

No opinion. AI doesn’t have opinions. It hedges its statements. “It might be a good idea to consider…” instead of “Do this.” Recipients sense the absence of conviction.

Prompt engineering for email: The right foundation

Your prompt is everything. A good prompt produces text requiring minimal editing. A bad prompt produces text you might as well write yourself.

The basic email prompt

Here’s a prompt structure that works:

Write an email with the following specifications:

SENDER: [Who you are, your role, your company]
RECIPIENT: [Who they are, their situation, their relationship to you]
GOAL: [What the email should achieve — one specific goal]
TONE: [Casual/professional/warm/direct — give an example]
LENGTH: [Word count or sentence count]
AVOID: [What the AI must not do]

CONTEXT:
[All relevant background information]

EXAMPLE OF MY TONE:
[Insert 1-2 examples of emails you've written]

Example: Welcome email

Write a welcome email with the following specifications:

SENDER: Alex, founder of an email marketing platform
RECIPIENT: New subscriber who just signed up for the newsletter via a blog about email automation
GOAL: Get them to reply to the email with what their biggest challenge is
TONE: Direct, informal, like an email from a friend who knows a lot about email marketing. No corporate speak.
LENGTH: Max 150 words
AVOID: Transition words, "furthermore"/"moreover", bulleted lists, emojis, capitalized headings

CONTEXT:
The newsletter covers email marketing tactics. We send weekly. Subscribers are primarily small businesses and solopreneurs.

EXAMPLE OF MY TONE:
"Hey — I'm Alex. I've spent 8 years helping businesses write emails that actually get opened. The short version: most emails are too long, too generic, and sent at the wrong time. The long version is what this newsletter is about."

What makes the difference

The tone example is key. Without it, AI writes generic text. With a concrete example, it matches your style 80% of the way.

The length constraint forces precision. AI given 500 words writes 500 words of mediocre quality. AI given 150 words writes 150 words of high quality.

The “AVOID” section removes AI fingerprints. The most robotic elements can be specified away.

Tone of voice: Give AI your voice

The most effective method for avoiding generic AI text is giving it a detailed description of your voice.

Create your tone of voice guide

Describe your voice with these parameters:

Formality: From “boardroom presentation” to “message to a friend” Example: “Informal but knowledgeable. Like an expert talking over a cup of coffee.”

Sentence length: Short, punchy? Or flowing, narrative? Example: “Mix short sentences (3-7 words) with medium sentences (10-15 words). Never over 20 words.”

Words you use: Specific terminology, slang, industry language Example: “Say ‘open rate’ not ‘email opening frequency.’ Say ‘emails’ not ‘e-mails.’ Say ‘works’ not ‘functions effectively.’”

Words you never use: Corporate jargon, cliches Example: “Never: ‘in this regard,’ ‘we are proud to,’ ‘state-of-the-art,’ ‘leverage,’ ‘synergy.’”

Attitude: Do you have strong opinions? Are you diplomatic? Example: “Strong opinions, expressed directly. ‘Never buy an email list’ not ‘Purchasing email lists can be problematic.’”

Use the tone guide in every prompt

Insert your tone guide at the start of every prompt, or save it as a “system prompt” in your AI tool. The more consistently you use it, the more consistent the output becomes.

Editing: From AI draft to your email

AI output is never finished. It’s a first draft. Here’s your editing process:

Step 1: Remove AI gimmicks (2 minutes)

Search for and remove:

• Transition words at the start of sentences (“Furthermore,” “Moreover,” “In addition”)
• Filler sentences (“It’s important to note,” “It’s worth mentioning”)
• Over-explanations (AI explains things the recipient already knows)
• Passive phrasing (“It could be considered” becomes “Do this”)

Step 2: Add yourself (3 minutes)

Insert:

• A personal reference (“As I mentioned when we talked last week…”)
• A specific example from your reality (“We tested this for [Client] and…”)
• Your opinion (“I believe that…” / “My experience is that…”)
• A detail only you would know (“Most people forget that…”)

Step 3: Cut it down (2 minutes)

AI always writes too much. Cut:

• The opening sentence (AI’s “summary” is rarely the best opening)
• Redundant points (did AI say the same thing twice with different words?)
• The closing summary (the email doesn’t need a conclusion)

Step 4: Read it out loud (1 minute)

Read the email out loud. If you stumble on a sentence, rewrite it. If it sounds like something you’d never say, remove it. The email should sound like you talking.

Email types AI is good at

Routine emails (save 70% time)

• Order confirmations
• Shipping updates
• Meeting invitations
• Onboarding guides
• FAQ responses

These emails have low creative freedom and high repetition. Perfect for AI.

Newsletters (save 50% time)

AI can produce a solid first draft of your newsletter based on your key points. You edit, add your personality, and send.

Subject lines (save 60% time)

Ask AI for 20 variants of a subject line. Choose the 3 best and A/B test them. AI is surprisingly good at subject lines because they’re short and testable.

Sequences (save 60% time)

Welcome sequences, drip campaigns, re-engagement flows. AI can produce the entire sequence based on a detailed brief. You edit each email individually.

Email types AI is bad at

Crisis communication

“We’ve had a data breach.” AI lacks the empathy, timing, and legal care for this type of communication. Write it yourself (or let your lawyer do it).

Personal sales emails

A real 1-to-1 sales email to a specific prospect requires research, context, and personal insight that AI doesn’t have. AI can help with structure, but the core message should come from you.

Apologies and error correction

“We sent the wrong price in our newsletter yesterday.” Authenticity is crucial. AI-generated apologies feel hollow.

Emails to VIP customers

Your most important customers deserve a personal email. Not because AI can’t write one — but because personal effort signals value.

Advanced techniques

Chain-of-thought prompting

Instead of asking AI to write the email directly, ask it to think first. This technique, described in OpenAI’s prompt engineering guide, produces significantly better output:

Before you write the email, answer these questions:
1. What is the recipient's biggest concern right now?
2. What is the single most important point in this email?
3. What action do I want them to take?
4. What objection will they likely have?

Then write the email based on your answers.

Few-shot prompting

Give AI 3-5 examples of good emails you’ve written. Ask it to write in the same style. The more examples, the better it matches your tone.

Iterative refinement

First prompt gives a draft. Then:

• “Make it 30% shorter”
• “Remove all transition words”
• “Make the tone more direct”
• “Add a specific example in paragraph 2”

Three iterations typically give better results than one long prompt.

Your AI email workflow

Here’s the workflow that gives the best results:

1. Brief (5 min) — Write a detailed prompt with goal, recipient, tone, and examples
2. Generate (1 min) — Let AI produce the draft
3. Edit (5-10 min) — Remove AI gimmicks, add your personality, cut down
4. Review (2 min) — Read aloud, check it sounds like you
5. Send (1 min) — Approve and send

Total time: 15-20 minutes for an email that normally takes 30-45 minutes.

For sequences and campaigns, multiply by the number of emails — but generation and editing get faster the more you write in the same session, because AI (and you) get into flow.

The golden rules

Rule 1: AI writes the first draft. You write the final one.

Rule 2: Always provide examples of your tone. Without them, AI is blind.

Rule 3: Shorter is always better. Ask AI for half of what you need.

Rule 4: A human always approves. No automatic sending of AI-generated content.

Rule 5: Use AI for speed, not creativity. The best ideas still come from you.

AI is the best writing assistant you’ve ever had. Use it right, and you write twice as many emails in half the time — without anyone noticing. Use it wrong, and you send soulless emails that drown in the inbox alongside everyone else’s AI-generated spam.

The difference is 15 minutes of editing.