"Write a cold email" — why it reads like every other cold email
The cold outreach prompt that produces templated openers, and a rewritten version built around a specific reason for contact.
The prompt, as typed
Write a cold email to potential customers about our services. Make it personalized.
What it leaves the model to guess
- Who specifically, and why them right now
- What you know about them that is actually specific
- The ask — a meeting, a reply, a click
- Length, and what the email must not claim
Graded before and after
Both versions below were scored by the same rubric the free prompt grader runs.
As typed
18/100 · grade F
| Clarity of context | 0 |
|---|---|
| Structure | 3 |
| Specificity | 2 |
| Reusability | 0 |
| Decision logic | 0 |
Rewritten
90/100 · grade A-
| Clarity of context | 2 +2 |
|---|---|
| Structure | 7 +4 |
| Specificity | 8 +6 |
| Reusability | 6 +6 |
| Decision logic | 5 +5 |
The rewritten prompt
Write a first-touch cold email.
Recipient: {{role}} at {{company_type}}
Trigger — the specific reason for contacting them now: {{trigger}}
What we do, in one plain sentence: {{value_line}}
Ask: {{ask}}
Output: subject line under 45 characters, body under 90 words.
Structure: open with the trigger, not with us. One sentence on relevance.
One specific ask with a low-effort yes.
Tone: {{tone}} — plain, not clever. Channel: cold email, first touch.
Constraints: no "I hope this finds you well", no "quick question", no claims
about their business we cannot support from the trigger above, no more than
one link.
Write 3 variants and rank them by how specific the opening line is to this
recipient. Explain the ranking in one line.
This email succeeds if the recipient could not receive the same text from a
competitor. If the trigger is weak enough that the email would read as
generic, say so instead of writing it. Why the rewrite works
The last instruction matters more than it looks. Telling the model it may refuse gives you a signal about your own inputs — if it declines, the problem is that you had no real reason to contact this person, which no amount of rewriting fixes.
Common questions
- Why does this prompt score badly?
- It scores 18 out of 100 (F) because it leaves the model to decide 4 things you already know: who specifically, and why them right now; what you know about them that is actually specific; the ask — a meeting, a reply, a click; length, and what the email must not claim.
- Does a longer prompt always score higher?
- No. The rubric rewards named inputs, a fixed output format and an explicit rule for choosing between options. Extra prose without those adds length and no score.
- Will this rewritten prompt work in ChatGPT and Claude?
- Yes. It is plain text with placeholders you fill in before running, so nothing in it is specific to one model or vendor.