ChatGPT Cold Email Prompt: 7 Prompts That Produce Better Outreach

If you have ever asked ChatGPT to "write a cold email," you already know the problem.

The result is usually clean. It is also usually useless.

It sounds polished in the way airport hotel art looks polished. Nothing is technically broken, but nobody is rearranging their day because they saw it.

That is not really ChatGPT's fault. The model can only work with the prompt you give it. If the prompt is vague, the output gets vague. If the prompt asks for "a professional cold email," you tend to get the exact kind of professional sludge that busy prospects delete on sight.

The fix is not mystical. You need better inputs, a tighter job for the model, and a prompt structure that forces relevance instead of filler.

This guide gives you seven practical ChatGPT cold email prompts, plus the rules that make them produce something you might actually send.

---

Why Most ChatGPT Cold Email Prompts Fail

Most people prompt for cold email the same way they would prompt for a school assignment:

Write a cold email for my product.

That prompt fails for three reasons.

1. It gives the model no commercial context

Who is the buyer? What changed in their world? What is the real problem? What is the offer? What is the right ask?

If you leave those blank, the model fills the space with defaults. That is where all the "hope this finds you well" energy comes from.

2. It asks for one answer instead of useful options

Cold email is not a single-perfect-draft game. Good outbound usually comes from testing angles:

• direct
• curiosity
• outcome
• teardown
• timing-based

One prompt, one output is a bad bargain.

3. It rewards tone before relevance

A lot of prompts ask the model to sound persuasive, engaging, professional, or human. Those are not useless instructions, but they come after the real job.

First make the email relevant. Then make it sound good.

---

What to Put in a Cold Email Prompt

Before the actual templates, here is the core input stack that matters:

1. What you sell
   One sentence. No feature parade.

2. Who the buyer is
   Role, company type, or segment.

3. Why you are reaching out now
   A trigger, timing event, or known pressure.

4. What problem you solve
   A problem they can recognize fast.

5. What proof or constraint matters
   Case-study signal, price angle, speed, stage, or credibility limit.

6. What ask you want
   Short demo, reply, teardown offer, or permission to send details.

The model does better when you constrain the job. It does not need freedom. It needs a lane.

---

Prompt 1: The Fast First-Touch Prompt

Use this when you want a general first-touch draft for B2B outreach.

Write 3 cold email variants for this offer.

Product: [one-sentence product description]
Target buyer: [role + company type]
Prospect context: [name, company, and one relevant observation]
Problem we solve: [specific pain point]
Desired outcome: [credible result]
CTA: [small ask]

Rules:
- Keep each email under 140 words
- Write one direct version, one curiosity version, and one outcome-led version
- Do not use fake flattery
- Do not use generic claims like "streamline" or "optimize"
- Make the email sound like a person, not a sales sequence template

Why this works: it forces angle variation and stops the model from hiding inside one bland draft.

Good use case: founders, SDRs, consultants, and anyone who wants three directions instead of one shot at mediocrity.

---

Prompt 2: The Founder-Led Outreach Prompt

This is better when the sender is a founder and the email should feel more direct.

Write 3 founder-led cold emails to this prospect.

My company: [what you do]
Prospect: [name, role, company]
Reason for outreach: [specific trigger or observation]
What makes us relevant: [clear business fit]
What I want: [reply, quick look, teardown offer, short intro call]

Rules:
- Sound commercially sharp, not corporate
- Avoid hype, grand claims, and startup jargon
- Keep it under 120 words
- Make the CTA feel easy to answer
- Do not mention AI unless it actually helps the case

This is the right prompt when you want the email to sound like somebody who understands the problem, not like a junior AE in borrowed loafers.

If that is your main motion, the full playbook in cold email for startups goes deeper on what early teams should sound like.

---

Prompt 3: The Agency or Consultant Prompt

Service businesses have a different problem. Generic AI output makes them sound interchangeable with every other shop in the inbox.

Write 3 cold emails for a service business.

Offer: [agency or consulting offer]
Target buyer: [role + company type]
Observed issue: [growth gap, messaging issue, hiring move, pricing problem, etc.]
Why we are credible: [specific experience, specialization, or result]
CTA: [quick look, audit, teardown, short call]

Rules:
- No buzzwords
- No broad claims like "help businesses scale"
- Make each email specific to the observed issue
- Keep the ask low-friction
- Keep each version under 150 words

Use this if you sell a service and need the first line to sound chosen, not sprayed.

For more on that angle, see cold email for consultants and cold email for agencies.

---

Prompt 4: The Recruiter Outreach Prompt

Candidate outreach is cold email too. It just breaks faster when it sounds automated.

Write 3 recruiter outreach emails for this candidate.

Role we are hiring for: [job title + short context]
Candidate: [name + relevant background]
Why they may care: [specific fit, problem, stack, or growth stage]
What we can honestly say: [comp range, team quality, mission, technical challenge, etc.]
CTA: [open to hearing more, worth sending details, quick reply]

Rules:
- Sound specific without sounding invasive
- Do not oversell the company
- Avoid stiff HR language
- Keep each email under 130 words
- Make one version more direct, one warmer, and one more curiosity-led

This matters because candidate inboxes are full of messages that feel copied from the last engineer with the same title.

If recruiter outreach is your lane, cold email for recruiters covers the structure in more detail.

---

Prompt 5: The Follow-Up Prompt

Most people use ChatGPT for first-touch emails and forget the follow-up, which is where a lot of replies actually happen.

Write 3 follow-up emails to this cold outbound message.

Original email: [paste original]
Prospect: [name, role, company]
What we offer: [short offer]
What new angle we can add: [case study, example, timing reason, teardown, alternate benefit]
Goal: [get a reply, offer more detail, make the CTA easier]

Rules:
- Do not guilt the prospect
- Do not say "just bumping this to the top of your inbox"
- Each follow-up should introduce a new angle
- Keep each one under 90 words

That last rule matters. A follow-up that says nothing new is just a second chance to annoy somebody.

For sequence structure, timing, and what to say across touches, read cold email follow-up.

---

Prompt 6: The Teardown Offer Prompt

This one is useful when you want the email to offer value up front instead of leading with a pitch.

Write 3 cold emails offering a short teardown.

Prospect: [name, role, company]
Public asset we reviewed: [site, onboarding, outbound flow, pricing page, job post, etc.]
What we noticed: [1-2 plausible issues]
What we can offer: [short teardown, 3 notes, brief audit]
CTA: [want me to send it?]

Rules:
- Keep it concrete
- Do not pretend we did deep research if we did not
- Avoid aggressive meeting asks
- Under 120 words
- Make the value clear in the first half of the email

This prompt is good because it gives the model a reason to write with substance. You are not asking it to invent chemistry. You are asking it to package a useful observation.

---

Prompt 7: The Subject Line Prompt

People spend too much time asking ChatGPT for body copy and not enough time making the open worthwhile.

Generate 20 subject lines for a cold email.

Offer: [what we sell]
Target buyer: [role + company type]
Reason for outreach: [trigger or pain point]
Tone: [direct, curious, plainspoken]

Rules:
- No clickbait
- No fake familiarity
- No all lowercase gimmicks unless the line earns it
- Mix direct, curiosity, and problem-led subject lines
- Keep most under 50 characters

Then do the obvious thing. Throw out most of them.

The model will still produce some junk. That is normal. You are looking for a few lines worth testing, not divine revelation from the autocomplete gods.

If you want examples and categories that actually hold up, cold email subject lines is the longer guide.

---

How to Get Better Output From These Prompts

The prompt matters, but the second pass matters too.

Here is the editing checklist I would actually use after ChatGPT gives you a draft:

Cut the first sentence if it is generic

If the opener could be pasted into a thousand emails, it probably should not survive.

Replace category language with observed detail

"We help companies improve outbound performance" is a sentence built to be forgotten.

"You are hiring AEs while moving upmarket" is specific enough to earn a second look.

Lower the ask

If the draft ends with "Do you have 30 minutes next week?" it is probably asking for too much too early.

Small asks win more replies:

• worth sending the short version?
• want the teardown?
• should I send a few example angles?

Keep only one idea per email

ChatGPT loves stacking claims. Cut until the email has one clean reason to exist.

Test multiple angles

This is the biggest advantage of using AI in the first place. Do not waste it by treating the first draft as the final one.

---

When ChatGPT Is Enough, and When It Is Not

ChatGPT is fine when you already know:

• the market
• the prospect
• the angle
• the ask

At that point, it can speed up drafting.

It is weaker when you want a purpose-built workflow:

• multiple usable variants by default
• a structure tuned for cold email instead of general writing
• fewer prompt gymnastics
• faster movement from blank page to sendable draft

That is the real reason products like ColdCraft exist. Not because ChatGPT cannot write. Because most people do not want to keep rebuilding the same prompt scaffolding every time they need three cold emails before lunch.

ColdCraft takes a short product brief, a target customer, and a prospect, then gives you 3 personalized cold email variants in about 30 seconds. No prompt engineering spiral. No pretending the first generic draft is secretly good if you squint.

---

Final Take

The best ChatGPT cold email prompt is not the cleverest one. It is the one that gives the model enough commercial reality to stop improvising brochure copy.

That means:

• a real buyer
• a real problem
• a real reason now
• a small ask
• more than one angle

Do that, and ChatGPT becomes useful.

Skip it, and you get clean nonsense.

If you want a faster way to go from rough context to multiple usable first drafts, try ColdCraft. It was built for exactly that job.