ChatGPT Prompts for Cold Email Outreach: 40 Templates That Get Replies

The short version: ChatGPT can write you a technically correct cold email in ten seconds, but a technically correct cold email is exactly the kind prospects delete without reading. Below are 40 prompts, split by subject line, opener, body, follow-up, breakup and reply-handling, plus the one input most people leave out that decides whether any of this works.

Why most ChatGPT cold emails get ignored

I want to say something uncomfortable before I hand you 40 prompts, because it changes how you use every single one of them: the prompt is not what gets you a reply. The research you feed into the prompt is what gets you a reply. ChatGPT cannot invent a genuine reason to contact someone that you have not given it. It can only dress up what you already know in confident, well-punctuated sentences. Fed nothing, it produces something fast and fluent and completely generic, and generic is the one thing a cold email cannot afford to be.

I see this constantly with small business owners I work with. They paste in "write me a cold email to a marketing director," ChatGPT produces something that reads like every LinkedIn outreach message ever written, and they wonder why it gets a 1 percent reply rate. The tool did its job. The person using it skipped theirs. If you already use ChatGPT for other parts of running your business, the same rule applies everywhere, and I go into it more in ChatGPT prompts for small business owners: the output is only as sharp as the input you give it.

The test that changed how I write these prompts

In January 2026 I ran a straight test on my own list building, not for a client, so I could be honest about the numbers with myself. I sent 200 cold emails using a generic ChatGPT prompt with no research fed in beyond name, company and job title. Then I sent another 200 using the same underlying offer, but I fed ChatGPT one specific detail per prospect: a recent LinkedIn post, a funding announcement, a job listing they'd posted, something true and current about them.

Batch one: 6 replies. A 3 percent reply rate.

Batch two: 22 replies. An 11 percent reply rate.

Same offer. Same subject line structure. Same send time. The only variable was the specific detail I fed the prompt before I asked it to write anything. That gap, nearly four times the reply rate, is the whole argument for how you should use everything below. Treat these 40 prompts as containers you fill with real detail, not as finished emails.

There's a second uncomfortable bit worth saying out loud. Twice in the past year a prospect has replied to one of my emails with some version of "is this AI?" Both times I'd been lazy with the research step and it showed in the phrasing, that slightly too-smooth, slightly too-symmetrical sentence structure ChatGPT defaults to. Both times I lost the reply thread. People are getting better at spotting the tell, so the fix isn't to hide that you used ChatGPT, it's to make sure what you fed it was specific enough that the output couldn't have been sent to anyone else.

How to use these prompts (the process, not just the list)

Five steps, every time, no shortcuts:

• Step 1: Spend three minutes on the prospect. LinkedIn post, company news, a review they left, a podcast they were on. One real detail, not five.
• Step 2: Paste that detail into the prompt along with your product or service in one sentence.
• Step 3: Ask ChatGPT for three variations, not one. The first draft is almost always the safest and least interesting.
• Step 4: Cut it to under 90 words. In my own campaigns, emails over 120 words see reply rates drop by roughly half, because nobody reads a paragraph from a stranger.
• Step 5: Read it out loud before you send it. If it sounds like a press release, rewrite the first line yourself. ChatGPT is good at structure, weaker at sounding like an actual human who typed something quickly between meetings.

If you're building a system around this rather than doing it once, it's worth setting these up as saved prompts you reuse and refine, which is the whole idea behind building a personal library of ChatGPT prompts rather than starting from scratch every time you sit down to write outreach.

40 ChatGPT prompts for cold email outreach

Copy these as they are or adapt the brackets. If you want more general-purpose prompts beyond cold email, I've laid out 100 more by task in this list of copy-paste ChatGPT prompts.

Subject line prompts (1 to 8)

• 1. "Write 10 subject lines under 6 words for a cold email to [JOB TITLE] at [INDUSTRY] companies. No punctuation, no emojis, sound like a colleague not a marketer."
• 2. "Turn this fact about [PROSPECT COMPANY] into a curiosity-based subject line under 40 characters: [SPECIFIC FACT]."
• 3. "Write 5 subject lines that reference [SHARED CONNECTION OR MUTUAL CONTACT] without sounding like a favour is being called in."
• 4. "Rewrite this subject line to sound like it came from a person, not a company: [YOUR DRAFT SUBJECT LINE]."
• 5. "Write a subject line based on a question [PROSPECT'S ROLE] would ask themselves this quarter, related to [PROBLEM YOU SOLVE]."
• 6. "Give me 8 subject line options that would work as the first line of a text message, not an email."
• 7. "Write a subject line that references [THEIR RECENT LINKEDIN POST TOPIC] in under 5 words."
• 8. "Take this subject line and make it 30 percent more specific by adding a number or a name: [YOUR DRAFT]."

Opening line and icebreaker prompts (9 to 16)

• 9. "Write an opening line for a cold email that references this specific detail without flattering them: [SPECIFIC DETAIL ABOUT PROSPECT]."
• 10. "I'm emailing [NAME], [JOB TITLE] at [COMPANY]. They recently [RECENT EVENT]. Write 3 opening lines that connect this event to a problem, not a compliment."
• 11. "Write an opener that starts with a specific number or result relevant to their industry, not a greeting."
• 12. "Take this generic opener and rewrite it so it could not be sent to anyone but this exact person: [GENERIC OPENER] + [PROSPECT DETAILS]."
• 13. "Write an opening line based on a job posting this company put up recently: [JOB POSTING DETAILS]."
• 14. "Write 3 openers that reference a competitor's recent move without naming the competitor directly."
• 15. "Write an opener for someone who has probably received 20 cold emails this week. Make it the opposite of what those 20 probably said."
• 16. "Rewrite this opener to remove any word a spam filter would flag: [DRAFT OPENER]."

Body and value proposition prompts (17 to 24)

• 17. "Write a 2-sentence value proposition for [YOUR PRODUCT] aimed at [JOB TITLE], focused on the outcome they get, not the features."
• 18. "Turn this list of features into one sentence about the result a customer sees: [FEATURE LIST]."
• 19. "Write the body of a cold email under 80 words: 1 line connecting to them, 1 line on the problem, 1 line on the result, 1 question."
• 20. "Write a version of this email that includes one specific number or statistic as proof: [YOUR DRAFT] + [YOUR REAL DATA]."
• 21. "Rewrite this email so the call to action is a question, not a meeting request."
• 22. "Write a cold email body that mentions a client result without naming the client, only the industry and the outcome."
• 23. "Take this email and cut it by a third without losing the main point: [YOUR DRAFT]."
• 24. "Write this same email for a [SMALLER OR LARGER] company, adjusting the tone and the size of the claim."

Follow-up prompts (25 to 32)

• 25. "Write a 3-line follow-up to this email that adds new information, not just a reminder: [ORIGINAL EMAIL]."
• 26. "Write a follow-up that references something that has happened since the first email, such as [NEW EVENT OR NEWS]."
• 27. "Write a follow-up that asks a different, smaller question than the original email did."
• 28. "Write a follow-up that includes a short case study or result in one sentence, aimed at [JOB TITLE]."
• 29. "Write a follow-up that assumes they're busy, not that they ignored me on purpose."
• 30. "Turn this follow-up into a version with no question mark, ending on a statement instead."
• 31. "Write a follow-up that offers a specific resource, such as a short guide or template, relevant to [THEIR PROBLEM]."
• 32. "Write a follow-up sequence of 3 short emails, each under 50 words, spaced a week apart, each with a slightly different angle on the same offer."

Breakup and last-touch prompts (33 to 36)

• 33. "Write a breakup email that gives them a graceful way to say no, under 40 words."
• 34. "Write a last-touch email that asks who the right person to speak to would be, if not them."
• 35. "Write a breakup email with a light, self-aware tone that does not guilt-trip the reader."
• 36. "Write a final email that includes one useful piece of free information related to [THEIR INDUSTRY], with no ask at the end."

Reply-handling prompts (37 to 40)

• 37. "They replied with '[THEIR REPLY]'. Write a short response that moves the conversation toward a call without sounding pushy."
• 38. "They said 'not right now.' Write a reply that leaves the door open and suggests a specific future check-in date."
• 39. "They asked for pricing before a call. Write a short reply that gives a range and still books the call."
• 40. "They said 'we already use [COMPETITOR].' Write a reply that asks one genuine question about how that's working, without criticising the competitor."

What separates a prompt worth keeping from one you'll never use again

Out of these 40, you'll probably end up reusing about six or seven for most of your outreach. That's normal, and it's worth being deliberate about which ones earn a permanent spot in your process rather than treating all 40 as equally important. I wrote a whole piece on what makes a ChatGPT prompt worth saving, and the short version applies here too: a prompt is worth keeping if it still produces something useful after you've fed it 20 different sets of details, not just the one you tested it with.

If you're doing this for recruiting rather than sales, the mechanics barely change, you're still writing to a stranger about an opportunity they didn't ask about, and I've broken out the specific version of this for hiring outreach in these prompts for recruiters.

The bit that decides whether any of this works

None of these 40 prompts will fix a bad list. If you're sending to people who have no reason to need what you sell, the best-written email in the world gets a 2 percent reply rate, and a bad email to a well-targeted list still beats a great email to a random one. I've watched founders spend three hours perfecting a ChatGPT prompt and ten minutes building the prospect list, when the ratio should run the other way. Get the list right first. Then let ChatGPT help you say something specific to the people on it, one detail at a time, and check your reply rate against a benchmark that means something, not a vanity metric. A 5 percent reply rate on cold email is solid. Above 10 percent, as I saw in that January test, means your targeting and your personalization are both doing their job at the same time, which is rarer than most cold email advice admits.

Related guides live in the ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps. To write your own, try the free tool that writes the prompt for you.

I covered the follow-on problem in How can a cold email be written to get a response?.

Frequently asked questions

What is the best ChatGPT prompt for cold email subject lines?

The one that forces specificity: ask ChatGPT to turn one real fact about the prospect's company into a subject line under 6 words, rather than asking for "catchy subject lines," which produces generic clickbait that hurts deliverability.

How many cold emails should I send before changing my ChatGPT prompt?

Send at least 50 to 100 with the same structure before judging it. Anything smaller and you're reacting to noise, not a real pattern, and you'll change a working prompt for the wrong reason.

Does using ChatGPT for cold email hurt reply rates?

It hurts reply rates only when the input is generic. Fed specific research on each prospect, ChatGPT-assisted emails performed nearly four times better than generic ones in my own 400-email test in January 2026, 11 percent versus 3 percent.

Can prospects tell when a cold email was written with ChatGPT?

Sometimes, yes, especially when sentences are too smooth or symmetrical. The fix isn't hiding the tool, it's feeding it detail specific enough that the email couldn't have been sent to anyone else, which is what makes it read as genuine rather than templated.

Sources worth reading

• OpenAI's own platform documentation
• the UK Information Commissioner's guidance on direct marketing