SuperAPI AI Directory

subject-line-generator

Marketing & growth

View repository

Updated 10-19 18:39

About this skill

Subject Line Generator

Generate 10+ subject line options for a given email, score each one, and return a ranked shortlist for testing.

What you need

  • Email body (step 1 at minimum)
  • ICP and persona context
  • Signal the email is built around
  • Channel: cold email or LinkedIn InMail

Principles

  • Subject lines under 6 words perform best in cold outbound
  • Never use the company name in the subject line — it reads as a template
  • Curiosity beats clarity when it comes to cold email subject lines
  • Questions outperform statements when the question is specific, not generic
  • Personalization tokens work only when they're signal-derived, not just [First Name]

Subject line patterns to generate

Generate at least 2 from each pattern:

  1. Signal reference: directly references what just happened at their company ("new SDR team, new pipeline problem")
  2. Problem statement: names the specific pain without a solution ("pipeline that depends on one channel")
  3. Reframe: presents an assumption they hold and challenges it ("outbound isn't broken — the playbook is")
  4. Direct question: asks the exact question the email answers ("is your outbound tied to one channel?")
  5. Unexpected angle: something they wouldn't expect from a sales email ("your new CRO's first 90 days")
  6. Ultra short: 1–3 words, creates pattern interrupt ("quick thought", "SDR question")

Process

  1. Generate 2 subject lines from each of the 6 patterns above (12 total)
  2. Score each on: curiosity (0–2) + specificity (0–2) + length penalty if over 6 words (−1)
  3. Rank by score
  4. Flag top 3 for A/B testing

Output format

Subject lines: [hypothesis / email step]
Email: [step number and first line]

RANKED OPTIONS
Score | Subject line | Pattern
[n]   | [subject]    | [pattern type]
...

TOP 3 FOR TESTING
A: [subject]
B: [subject]
C: [subject]

Recommended default: [A/B/C] — [one-line reason]

Notes

  • Vary the top 3 by ANGLE, not by swapping synonyms — token-level randomisation changes nothing a filter reads and nothing a prospect notices. Different question, different fact, different frame.
  • Test subject lines systematically — track open rate by subject line variant, not campaign overall
  • If open rate is below 40%, subject line is the first thing to change before touching copy
  • Avoid: "quick question", "following up", "[First Name]" as opener — overused and filtered mentally