How to Write Cold Emails With ChatGPT That Get Replies

Key takeaways

• Research one real detail about the person before you write a word.
• Keep the email short, with a single, easy-to-answer request.
• Send a polite follow-up, and respect anyone who says no.

Most cold emails fail in the first line. "I hope this email finds you well. My company offers innovative solutions…" is deleted before the second sentence. If you've ever stared at a blank draft, then pasted a prompt into ChatGPT and gotten something that sounds like a brochure, you know the feeling. The tool isn't the problem. It writes whatever you feed it, and a vague prompt gets a vague email. Here's a practical method: research one real detail, write a short message around it, test a few subject lines, and follow up without being annoying.

1. Find one real reason to write to this person

Before using AI, spend five minutes on the recipient. Look at their LinkedIn, company blog, recent posts, or product page. Find one specific, true detail: a new product launch, a talk they gave, a hiring push, or a problem their industry faces. Write it down in a sentence. That detail is what separates your email from a mass blast.

If you can't find any reason to contact this person, that's useful information. Maybe they're the wrong target. Cold email works best when the connection is obvious to the reader. Ask yourself: why would this person care about my message today, and not on any other day? If you can't answer, keep researching. Relevance is the whole game here.

2. Give ChatGPT the facts and tight limits

Paste your research detail, who you are, what you offer, and what you want. Then set limits: under 90 words, no flattery, no "I hope you're well," one question at the end. Ask for plain language a busy person can read in fifteen seconds. Tell it not to invent mutual connections, results, or claims about the recipient.

Request two versions with different angles, such as one that leads with their recent news and one that leads with a shared problem. Compare them. Usually one reads as more natural. The goal is a short note a real person might send to a colleague, not a polished pitch. Read both aloud before you choose, since awkward phrasing is easy to hear.

3. Test subject lines and fix the ask

Subject lines should be short and honest, ideally under six words, and look like something a coworker might send. Ask ChatGPT for ten options, then pick two or three. Avoid anything misleading like "Re: our meeting" when there was no meeting. Tricks might raise opens but they lose trust and can break anti-spam rules. Honest and plain usually works best.

Next, check the ask. A request like "Can we schedule a 30-minute demo?" is heavy for a stranger. Try something smaller: "Would it be useful if I sent a two-page example?" or "Who on your team handles this?" Ask AI to rewrite your closing line into three lower-pressure options and pick the most natural. Low pressure makes a reply feel easy.

4. Plan follow-ups and respect the answer

Most replies come after a follow-up, so plan one or two. Ask ChatGPT to write a short second email, three or four days later, that adds something new, such as a relevant example or a different question, instead of "just bumping this." Keep it under 60 words. Space them out and stop after two or three attempts.

If someone says no or asks to be removed, thank them and remove them right away. In many places, rules like CAN-SPAM in the US and GDPR in Europe govern commercial email, including opt-outs and identifying who you are. Include your real name, business details, and an easy way to opt out. Doing this properly protects both your reputation and your inbox.

A weak prompt vs. a better one

You run a small bookkeeping service and want to contact a local dentist who just opened a second clinic.

Weak prompt:

Better prompt:

The weak prompt returns a generic sales pitch. The better prompt anchors the email on a real event, limits length, avoids made-up claims, and makes a small ask. Expect a short, relevant message that's easy to reply to with a simple yes or no.

A prompt you can copy and use

Paste this into ChatGPT, Claude or Gemini and replace the parts in brackets with your own details:

Mistakes and limits to know about

AI can write a smooth email that's still irrelevant, so research comes first. Check every fact it mentions about the recipient, because it can get names, titles, and companies wrong. Don't scrape or email huge lists of strangers. Follow the email laws that apply to you, including honest sender details and opt-out options. Avoid fake urgency, false familiarity, and misleading subject lines. And be realistic: even good cold emails often get no reply. Silence usually isn't personal, so test small batches and improve one thing at a time. Keep your tone respectful.

Keep learning

If you are new to this, start with the 30-day plan to start using AI in a small business. These related articles may help too: How to Write Customer Service Replies With AI and How to Write a Simple Business Plan With ChatGPT.

Frequently asked questions

Can ChatGPT write good cold emails?

It can write a solid first draft if you give it specific research, a clear offer, and strict limits. Without those, it produces generic pitches. Always edit for your voice, check facts about the recipient, and keep the email short with one simple request.

How long should a cold email be?

Aim for roughly 50 to 100 words. That's long enough to show relevance, say who you are, and make one request, but short enough to read in seconds. If you need more space to explain, the email probably has too many ideas.

How many follow-ups should I send after a cold email?

One or two follow-ups is reasonable, spaced a few days apart, with each adding something useful instead of just asking if they saw your message. If there's still no reply after two or three tries, stop. If someone asks you to stop, do so immediately.

Is cold emailing legal?

In many places, yes, if you follow the rules. In the US, CAN-SPAM requires honest sender information, no deceptive subject lines, and a working opt-out. The EU and other regions have stricter consent rules. Laws vary, so check the rules for your country and audience.

Last updated: October 2026. Written by the Learn AI Smart team. Always check AI output before you use it.