How to write cold emails with AI that don't read as AI
The problem isn't AI. It's generic AI. Here's how to use it to sound more like you, not less.
Why most AI cold emails fail
Buyers can smell a generic AI email in the first line. It's polished, it's about the sender, and it could have gone to anyone, so it gets deleted. The failure isn't the tool. It's using it to skip the thinking.
Used well, AI does the reverse. It frees up the time you'd spend formatting so you can spend it on relevance.
Feed it something specific first
AI can only be as specific as what you give it. Before you ask for a draft, spend a few minutes on the prospect: recent company news, a role change, a priority you can infer. Hand AI those raw facts.
'Write a cold email to a CFO' gets you sludge. 'Here are three facts about this company and this person, write a two-sentence opener that references the most relevant one' gets you something worth sending.
Short, one idea, in your voice
Put the limits in the prompt: under 90 words, one clear ask, no adjectives, and a specific reference to the prospect. Then give AI a sample of your own writing and tell it to match your tone.
The best AI outreach reads like a sharp version of you on a good day, not a template.
Edit like it's yours, because it is
Never send the first draft untouched. Cut the vague opening line, make sure the personal details are actually true, and set any timeframe or commitment yourself. The final edit is where a good email becomes a real one.
A quick test: if a prospect replied 'why are you emailing me?', would your first line answer it? If not, rewrite it.
A prompt structure that works
Keep a simple structure you can reuse: here's the prospect and two or three real facts, here's my offer in one line, write a sub-90-word email with one ask in this voice, referencing the most relevant fact. Keep refining it with follow-up instructions instead of rewriting from scratch.
Once it's dialed in, you're doing in five minutes what used to take thirty, and getting more replies for it.