The short version: ChatGPT is brilliant at cold email drafts if you feed it specifics about the prospect and keep it away from generic sales language, and useless if you just type "write me a cold email." The prompts below give you the inputs that matter and a worked example you can copy. The uncomfortable bit nobody tells you: the prompt isn't your biggest problem, your targeting is.
Why most ChatGPT cold emails get deleted
I've read hundreds of these over the years, from people proud of the AI tool they'd just discovered, and 90% of them have the same fault. They're well written and completely irrelevant to the person receiving them. ChatGPT will happily produce a polished, grammatically perfect email that sells to nobody in particular, because that's what you asked it to do.
The fix isn't a cleverer prompt. It's giving the model something true and specific about the prospect before you ask it to write anything. A cold email prompt without research attached is just a template generator with better punctuation.
The five inputs ChatGPT needs before it can write well
- Who the prospect is: name, company, role, and one fact about what they do or recently did
- What problem you solve and for whom, in one sentence
- One proof point: a number, a result, a method (not "we help businesses grow")
- The one action you want: a reply, a 15-minute call, a yes/no answer
- The tone: formal, casual, blunt, warm
Paste those five things into the prompt every time and your output improves instantly. Skip them and you get the same beige email everyone else is sending.
Prompt 1: the first-touch email
"Write a cold email to [name], [role] at [company]. They recently [specific fact, e.g. launched a new product line / posted about hiring issues / opened a second location]. I help [type of business] solve [specific problem] and my proof point is [number or result]. Keep it under 90 words, no flattery, no 'I hope this finds you well', end with a low-pressure ask for a reply, not a meeting. Tone: direct and human, like a note from a person, not a brand."
Prompt 2: the follow-up that isn't annoying
"Write a follow-up to the email below, sent 5 days ago with no reply. Don't apologise for following up, don't say 'just checking in', add one new piece of value (a stat, a short insight, or a different angle), and keep it under 60 words. Here's the original email: [paste]."
Prompt 3: the break-up email
"Write a final follow-up for a prospect who hasn't replied to two emails. Give them a clean, no-pressure exit, make it clear I won't keep emailing, and leave the door open for them to come back later. Under 50 words."
Prompt 4: the industry-specific opener
"I'm emailing [number] [type of business, e.g. independent dental practices] in [location]. Write 3 different opening lines I can test, each referencing a different common pain point for this industry: [list two or three pain points]. No generic business speak."
Prompt 5: the research-first prompt
"Here's everything I know about this prospect: [paste LinkedIn bio, company website snippet, recent news]. Pull out the one detail most worth mentioning in a cold email opener, and tell me why."
A worked example
Say you run a small bookkeeping firm and want to email the operations manager at a 40-person logistics company. You know, because you checked their LinkedIn and company blog, that they switched software providers three months ago and posted about "finally having clean numbers." You feed ChatGPT prompt 1 with that fact, your proof point (you cut month-end close from 12 days to 4 for a similar-sized client), and ask for a direct tone under 90 words.
A decent output looks like this:
"Hi Sarah, saw your post about finally having clean numbers after the software switch, that kind of overhaul is never small. I work with logistics firms your size cutting month-end close from around 12 days to 4, usually without touching the software itself. If getting the close down further would be useful, happy to send over the three changes that made the biggest difference. No pitch, just the list. Worth a quick reply?"
That's specific, short, and asks for almost nothing. Compare that to the generic version ChatGPT produces with no inputs ("I hope this email finds you well, I wanted to reach out because I noticed your company might benefit from our services...") and the difference is obvious. Same tool, completely different result, because of what you put in before you hit enter.
The step-by-step for a batch of cold emails
- Pick 10 to 20 prospects, not 200. Quality of research beats volume every time with cold email.
- Spend 5 minutes per prospect finding one specific, recent, true detail about them.
- Run prompt 1 for each, pasting in their detail and your proof point.
- Edit every single output by hand. ChatGPT writes "I" too often and can sound slightly too pleased with itself, strip that out.
- Send in batches of 5 to 10 and track replies before you send the rest.
- Use prompt 2 for anyone who hasn't replied after 4 to 6 working days.
The uncomfortable truth about cold email and AI
Here's what most posts on this topic won't say plainly: if your reply rate is under 2%, the problem is very rarely the wording of the email. It's the list. ChatGPT cannot fix a list of 500 people who have no reason to need what you sell. I've seen business owners spend hours refining prompts to squeeze an extra half a percent out of open rates, when the real lever was sending to 50 well-researched prospects instead of 500 random ones. A rough email to the right person outperforms a polished one to the wrong person, every time. Fix your list before you fix your prompt.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The other thing people avoid saying: AI-written cold email, even good AI-written cold email, is getting easier for recipients to spot. The rhythm gives it away, that slightly-too-balanced sentence structure, the em-dash habit (which I'd tell you to strip out anyway), the "I wanted to reach out" openers. Use ChatGPT to draft, then rewrite the first and last sentence yourself in your own words. Those two sentences do most of the work of sounding human.
What to do once you get a reply
Cold email is only the opener. Once someone replies, the quality of what comes next matters more. If you're also sending ongoing email marketing campaigns to warm leads, keep the tone consistent with your cold outreach so people recognise your voice. And if part of your pipeline comes from people already on your list, the same research-first approach works for newsletters people open, not just cold prospects. For seasonal pushes, the same specificity rule applies, which is why Black Friday email prompts work the same way: facts in, generic language out.
If you run a service business and send a lot of outbound, it's worth browsing a wider set of small business email prompt ideas so you're not reusing the same three structures on every prospect. And if cold email isn't your only channel, the research habit transfers directly, the same prompts that pull one sharp detail from a LinkedIn bio work for LinkedIn posts that get noticed too.
If you'd rather have someone build and test the system with you instead of trial and error, that's the kind of work an AI consultant for small business does: setting up the research process, the prompts, and the follow-up cadence so it isn't rebuilt from scratch every time.
Related: the logistics page.
For the rest of the series, see the ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps. To write your own, try the free prompt generator.
Frequently asked questions
Can ChatGPT write a whole cold email campaign on its own?
It can draft the emails, but it can't do your prospect research or pick your list, and those two things decide most of your reply rate. Treat it as a drafting tool, not a strategist.
How long should a cold email to a client be?
Aim for under 100 words for a first email. Busy people decide whether to reply within the first two sentences, so front-load the relevant detail and keep the ask small.
How many follow-ups should I send after a cold email?
Two is usually enough: one around day 5 with a new piece of value, and a short break-up email around day 10 to 12 if there's still no reply.
Why do my ChatGPT cold emails sound generic even with a good prompt?
Because the prompt is missing a specific, true detail about the prospect. Without that, ChatGPT fills the gap with safe, broad language that could apply to anyone, which is exactly why it gets ignored.
Run a business, a brand or a blog? I feature tools, products and expert guest posts for an audience of 15,000. See how it works
