Sales

How to Write Sales Emails with AI That Actually Get Replies

Most AI-written cold email reads like AI wrote it. Here's the structure we tested over 30 days that fixes that — and the prompt pattern that consistently produced replies.

Disclosure: We may earn a commission when you purchase through links. See our affiliate disclosure.

The short answer: The difference between a sales email that gets replies and one that doesn't is rarely the tool — it's the prompt structure and the edit. A role + context + task + constraints prompt beats 'write a cold email' every time. Short-form emails are exactly where budget tools like Rytr earn their keep; just budget time for the rewrite pass.

Why most AI cold emails fail

We analyzed why AI-generated sales emails get ignored. The pattern is consistent: they open with flattery ('I hope this finds you well'), they're about the sender, not the reader, and they bury the one relevant line. That's not an AI problem — it's a prompting problem. The same model that produces generic garbage with a one-line prompt produces genuinely sharp email copy with a structured one.

The prompt structure that works

In our 30-day test, the single biggest lever was the prompt. The structure we settled on:

You are a cold email writer. The recipient is [role] at [company], who [one pain point]. My offer: [one sentence]. Write a 90-word email that (1) opens with a specific observation about them, (2) states the one result I've delivered for someone similar, (3) asks a single low-commitment question. No flattery, no filler.

Three specific constraints — recipient's pain, one result, one question — are what turn generic output into something usable. ChatGPT free tier handled this fine in testing; the structure does the heavy lifting, not the model.

Subject lines: where AI genuinely saves time

Subject lines are the highest-leverage 50 characters in cold email, and AI is excellent at generating options. We tested asking each tool for 10 subject line variations per email. The best results came from a separate prompt: 'Write 10 subject lines for this email. Vary the style: curiosity, specific number, question, social proof. Mark which you'd send first.'

The cheap tools did this as well as the expensive ones — it's a short, discrete task, exactly where our testing showed price differences vanish.

The follow-up sequence

Most replies come from the follow-up, not the first email. We used AI to draft a 3-touch sequence per prospect: the original, a value-add bump ('saw your post on X, one thought'), and a breakup email. The key: each touch was prompted separately with the previous email's text included, so the model didn't repeat itself. For the full sequence structure, see how to write email sequences with AI.

Tools for the job

Sales emails are short-form, so the price gap we documented elsewhere mostly disappears. Our rankings from 30 days of testing:

The mistakes that kill reply rates

Frequently Asked Questions

Is AI good at writing cold emails?

Yes, when you structure the prompt — role, recipient's pain, one result, one question. With a one-line prompt, AI writes generic cold email that gets ignored. Our 30-day test showed the prompt structure matters more than which tool you use.

How do I choose an AI tool for sales emails?

For short-form sales emails, budget tools like Rytr ($7.50) produced output about 90% as good as Jasper ($39) in our tests. ChatGPT's free tier works too. Claude Pro is worth it when you need tone and multi-touch follow-up consistency.

How do I get AI to write subject lines that get opened?

Ask for 10 variations with a range of styles (curiosity, number, question, social proof) and have the model mark its top pick. Every tool we tested did this well — it's a short, discrete task.

Should I tell the prospect an AI wrote the email?

That's a relationship and disclosure decision, not a technical one. If you're drafting with AI and editing hard, you own the final email either way. See our guide on whether to tell clients you use AI.