How do I write cold emails with AI without sounding like a bot?
Two rules: have AI critique the email first, then rewrite. And keep it under 90 words with one specific observation about the recipient.
The reason most AI-written cold emails sound like a bot is not that AI cannot write — it is that the writer asked for the wrong thing. They asked for a "compelling cold email" and got a generic template, because that is what the request pattern-matched on.
Two upgrades make a sharp difference.
Upgrade one: critique first, write second. Paste the email you have and ask the model to read it as the busy recipient — what would make them archive it? You will get sharper, more specific feedback than any "make this better" prompt. Then ask for the rewrite informed by that critique.
Upgrade two: write tight on purpose. The constraints that produce good cold emails are well known. Under 90 words. Open with one specific observation about the recipient that proves you looked them up. End with a single yes-or-no question. AI is good at honoring constraints — give it real ones.
The last move is to actually send it from your account, with your real name and identity. AI is a co-writer, not a hiding place. Recipients can sense when the sender is not on the line, and they respond to humans, not pipelines.