Skip to content
September 2026

How Do You Write Cold Email With Claude? Four Prompts

The research prompt does most of the work, the drafting prompt does less than you expect, and the constraint that matters most is telling it what not to write.

Here's the trap. You ask an assistant to write a cold email, it produces something fluent and plausible in four seconds, and you send it. It reads exactly like the other nine emails that person received that morning, because they were all written the same way by the same models from the same kind of instruction.

The drafting step was never the bottleneck. Writing a competent paragraph is the thing these tools are best at and it is the least valuable part of the job. What determines whether the email works is what you knew before you started writing, and what you forbade.

So here are four prompts, in the order you use them, with the constraint block that matters more than any of them.

Prompt one: research, before any drafting

This is where the value is. You are asking for raw material and specifically not for prose, because the moment it starts writing it stops looking.

I am about to contact [NAME], [TITLE] at [COMPANY]. Here is what I found on their site and their LinkedIn: [PASTE]. Give me: (1) three things that changed at this company in the last six months, with the evidence for each; (2) what this person's job is likely measured on; (3) two things they probably do NOT care about that a vendor like me would wrongly assume they do. Do not write any email. If the material does not support a point, say so rather than inferring it.

The third item is the one worth keeping. A list of what your prospect does not care about is more useful than a list of what they might, because it removes the two sentences you were going to waste. The final instruction matters too: without it you get confident inference presented as research.

Prompt two: the first touch

Using only the research above, write a first cold email to [NAME]. Under 90 words. One specific observation about their situation that could not apply to another company. One sentence on what we do, in plain language. One question that is easy to answer in a sentence. No compliments, no "I noticed", no "hope this finds you well", no claim about their priorities that the research does not support. British English. Do not use the word solution.

Ninety words is doing real work as a constraint. Almost every bad cold email is bad because it is long enough to include the parts that make it generic, and a hard ceiling forces the model to spend its budget on the specific observation.

Prompt three: the follow-up that is not a bump

Write a follow-up to the email above, sent five working days later, assuming no reply. It must contain one piece of information that was not in the first email and would be useful to them even if they never reply to me. Under 60 words. Do not reference the previous email, do not apologise for following up, do not ask whether they saw it.

The instruction not to reference the previous email is the whole prompt. A follow-up that says "just bumping this" transfers the cost of your persistence onto the recipient. One that arrives with something new does not.

Prompt four: the breakup, written to be answerable

Write a final email closing the loop after three unanswered attempts. Under 40 words. It should make it easy and consequence-free to say no, and easy to say "not now, ask me in Q2". No guilt, no false scarcity, no "I will assume you are not interested". Do not ask a question that requires them to explain anything.

This one produces replies more often than the other three combined, in most people's experience, and the reason is that it is the only email in the sequence that costs the recipient nothing to answer.

The constraint block that does the actual work

Everything above works because of what it forbids rather than what it asks for. Keep a standing block and append it to every drafting prompt:

  • No em dashes. They are the single clearest tell of unedited model output in 2026, and this site bans them in its own editorial rules for that reason.
  • No "not just X, it's Y" constructions, and no sentence that begins by naming a problem the recipient has not told you they have.
  • No adjectives in front of your own product. Powerful, seamless, game-changing and unlock all mean the sender could not think of a specific benefit.
  • No claim the research does not support. This is the one that prevents the most damage, because a confidently wrong sentence about someone's business ends the conversation faster than a boring one.
  • Name the register. British or American English, and one real person's writing to imitate if you have one. Otherwise you get the average of everything, which is what generic means.

Where this stops working

At volume. These prompts assume you are pasting real research about a real person, and that is roughly a five-minute job per prospect even with an assistant doing the reading. Fifty a week is achievable; five hundred is not, and pretending otherwise is how the output goes back to being generic.

That ceiling is the same one we hit costing the manual approach: a complete outbound motion runs on $17 a month and tops out around fifty messages a week. Past that you are buying sending capacity rather than better writing, and the honest comparison of what that costs is in the itemised sending model.

The short version: spend your prompt budget on the research step and on the list of forbidden moves. The part where it writes the paragraph will take care of itself, and it was never the part that decided whether anyone replied.

Questions

What is the best prompt for writing cold emails?
The research prompt rather than the drafting one. Ask for three recent changes at the company with evidence, what the person is measured on, and two things they probably do not care about that you would wrongly assume they do. Then draft from that, under a hard word limit.
Why do AI-written cold emails all sound the same?
Because they are written from similar instructions by similar models, and most prompts specify what to include without specifying what to avoid. The constraint block does more work than the request: no em dashes, no adjectives in front of your own product, no claim the research does not support, and a named register.
How long should a cold email be?
Under 90 words for a first touch, under 60 for a follow-up and under 40 for a breakup, as hard constraints in the prompt. Length is where generic material accumulates, so a ceiling forces the model to spend its budget on the one specific observation that makes the email worth reading.
Should a follow-up reference the first email?
No. A follow-up that says it is bumping a previous message transfers the cost of your persistence to the recipient. Instruct the model to include one piece of information that was not in the first email and would be useful even if they never reply.
How many personalised emails can you actually send this way?
Around fifty a week, because each one needs real research pasted in and that is roughly five minutes per prospect even with an assistant reading for you. Past that the personalisation degrades and you are buying sending capacity rather than better writing.

Tools mentioned

All tools

Sources

Source interests are labelled. Almost everything published about this subject is written by someone selling into it.

More from the blog

All posts