Personalised subject lines can lift open rates dramatically, yet the vast majority of marketing emails still skip personalisation entirely. That gap is your opportunity. This is a copy-paste AI workflow to produce 20 strong subject lines in minutes, filter them down to the best two, and A/B test like the data-driven marketer you are.
Why the subject line is 80% of the battle
Let’s be blunt about the maths. If your subject line doesn’t earn the open, nothing else you wrote matters. The body copy, the offer, the beautiful design: all of it dies unread in the inbox.
The numbers back this up hard. Emails with personalised subject lines see open rates around 46%, compared with roughly 35% without, and simply using the recipient’s first name can lift opens by about 26%. [1] Segmented campaigns drive around 30% more opens and 50% more clicks than unsegmented blasts. [2]
Now the opportunity. Despite all that, the overwhelming majority of marketing emails still use no subject-line personalisation at all, with one analysis finding 97.7% skipped it entirely. [3] That’s the gap. Most of your competitors are sending the same flat subject line to everyone and hoping. You’re about to test twenty.
And here’s why AI is the right tool: writing one good subject line is creative work, but writing twenty good ones across different angles is a volume game, and volume is exactly what AI does well. You bring the judgment. It brings the options.
Step 1: Brief the AI properly (not just 'write subject lines')
If you type “write me some email subject lines,” you’ll get generic mush with too many exclamation marks. Garbage in, garbage out. The brief is everything.
A good brief gives the AI four things: who the email is for, what the email actually offers, the one action you want, and the voice. Here’s the structure:
“You’re helping me write email subject lines. Audience: [who they are and what they care about]. The email offers: [the specific value or news]. The one action I want: [open and click to X]. Our brand voice: [e.g. warm, direct, a little playful, no hype]. Constraints: under 50 characters, no spammy words, no exclamation marks. Don’t write anything yet, just confirm you understand the brief.”
That last line (“don’t write anything yet”) is a small trick that works. It makes the AI lock in the context before it starts generating, which noticeably improves what comes next.
If you’ve already done the work of teaching your AI tool how your brand sounds, this step gets even faster. Our guide on training AI on your brand voice is worth ten minutes if your AI output keeps coming out sounding like a press release.
Step 2: Generate 20 across five proven angles
Here’s the move that separates this from “ask AI for ideas.” Don’t ask for twenty subject lines. Ask for four each across five different psychological angles. Variety is what makes an A/B test meaningful.
“Now write 20 subject lines, exactly 4 for each of these 5 angles: (1) curiosity gap, (2) clear benefit, (3) urgency or timeliness, (4) a question, (5) personalised or segment-specific. Label each by angle. Keep every line under 50 characters and in our brand voice.”
You’ll get a labelled menu. Curiosity lines for the intrigued, benefit lines for the practical, urgency for the deadline-driven. Different people open for different reasons, and now you have coverage across all of them.
Twenty random subject lines are usually twenty versions of the same idea. Four lines across five distinct angles forces genuine range, which is the whole point of testing: you’re not picking the best wording of one idea, you’re discovering which idea your audience responds to.
Step 3: Score them before you send
You now have 20 options. You’re not testing all 20. Use AI one more time as a critic, then make the final call yourself.
“Score each of these 20 subject lines from 1 to 10 on three things: clarity (is the value obvious?), curiosity (does it make me want to open?), and authenticity (does it sound human, not like marketing?). Show the scores in a table, then recommend the top 3 and explain why in one line each. Be a tough critic.”
The table does two jobs. It kills the weak lines fast, and it surfaces the one or two that score high on every dimension. Those are your test candidates.
Then apply human judgment, because AI scoring is a guide, not a verdict. Read the top three aloud. Does it sound like something you’d actually send? Would it embarrass you in a screenshot? Cut anything that feels even slightly off-brand. You know your audience better than the model does.
Step 4: Run the actual A/B test
Now the part too many marketers skip: actually testing. Pick your two strongest lines, ideally from different angles (say, one curiosity and one clear-benefit), and pit them against each other.
Every major email platform (Mailchimp, Klaviyo, HubSpot, Brevo) has built-in A/B testing for subject lines. The setup is the same everywhere: send version A to one slice of your list, version B to another, and the platform sends the winner to everyone else once it has enough opens to call it.
Three rules to make the test mean something:
- Test one variable at a time. If you change the subject line, keep the send time and content identical. Otherwise you won’t know what moved the numbers.
- Give it a big enough sample. A test on 40 people tells you nothing. You need enough volume for the difference to be real, not noise. For small lists, that’s the limitation to respect.
- Log the winner and why. Over a few campaigns you’ll learn which angle consistently wins for your audience. That pattern is worth more than any single test.
For context, average open rates hover around 21% across industries, so even a few points of lift from a better subject line compounds into real revenue over a year. [4]
The full prompt stack (copy these)
Here’s the whole workflow in one place, so you can run it start to finish. Paste these in order into ChatGPT, Claude, or Gemini.
“You’re helping me write email subject lines. Audience: […]. The email offers: […]. Action I want: […]. Brand voice: […]. Constraints: under 50 characters, no spam words, no exclamation marks. Confirm you understand before writing.”
“Write 20 subject lines, 4 each across: curiosity, clear benefit, urgency, question, personalised. Label by angle, under 50 characters, in our voice.”
“Score all 20 from 1-10 on clarity, curiosity, and authenticity in a table. Recommend the top 3 with one line each. Be a tough critic.”
Ten minutes, twenty options, a scored shortlist, and two lines ready to test. Compare that to staring at a blank subject field hoping inspiration strikes.
One efficiency tip: save this three-prompt sequence somewhere you can reuse it, or build it into a saved project in your AI tool. Subject lines are a weekly job for most marketers, so the workflow pays back every single send.
Where AI subject lines go wrong
This workflow is genuinely good, but AI subject lines fail in predictable ways. Watch for these.
The clickbait trap. AI loves a curiosity gap, sometimes too much. If the subject line promises drama the email doesn’t deliver, you’ll win the open and lose the trust. Every open you earn dishonestly costs you the next ten. Keep it intriguing, keep it honest.
The sameness problem. Run this enough times and you’ll notice AI has favourite patterns (“The one thing…”, “Here’s why…”). If every subject line starts to sound the same across campaigns, your audience tunes out. Push the AI for fresh structures, and trust your own ear.
Forgetting the preview text. The subject line doesn’t work alone. The preview snippet next to it in the inbox is part of the pitch. Ask the AI to write a matching preview line for your top picks, so the two work together instead of repeating each other.
Ignoring deliverability. Spammy words, all caps, and too many emojis can land you in the promotions tab or worse. Your subject line strategy matters less if the email never reaches the inbox. This is also why owned channels like email matter so much now that search traffic is getting harder to win: when you have someone’s inbox, protect it.
Start with your next campaign. Run the three prompts, test two lines, and write down which angle won. Do that for a month and you’ll know more about what your audience opens than most marketers learn in a year of guessing.
Frequently asked questions
How long does this AI subject line workflow take?
About 10 minutes once you've done it once. Roughly two minutes to write the brief, two to generate 20 lines, three to score and shortlist them, and a few to set up the A/B test in your email platform. Saving the prompt sequence makes every future run faster.
Which AI tool is best for writing email subject lines?
ChatGPT, Claude, and Gemini all work well for this. The tool matters less than the brief you give it. The single biggest quality improvement comes from telling the AI your audience, your offer, and your brand voice up front, rather than just asking for 'some subject lines'.
Do personalised subject lines really improve open rates?
Yes. Personalised subject lines have been shown to achieve open rates around 46% versus roughly 35% without, and using a recipient's first name can lift opens by about 26%. Yet the large majority of marketing emails still use no personalisation, which makes it an easy edge to claim.
How many subject lines should I A/B test at once?
Just two for a clean test. Pick your two strongest options, ideally from different angles like curiosity versus clear benefit, and keep everything else (content, send time) identical so you know the subject line caused any difference. Make sure your sample is big enough that the result isn't random noise.
Will AI-written subject lines sound generic?
They will if you brief poorly or accept the first draft. Give the AI your real brand voice, push it across five different angles, score the results, and cut anything that sounds off. Used that way, AI gives you range and speed while you keep the final judgment on what actually sounds like you.
About this guide
This is a hands-on workflow for marketers who want better email open rates without spending an afternoon on subject lines. It includes the exact AI prompts to generate, score, and A/B test 20 options, plus the common mistakes that make AI subject lines fall flat. Email statistics are drawn from published industry benchmark studies.
Related reading
Hina brings 10+ years of marketing strategy and brand growth experience to the AI conversation. She helps businesses and teams cut through the noise and apply AI where it actually matters. Future Factors offers AI Bootcamps, Corporate Workshops, and Speaking & Consulting for businesses ready to adopt AI without the overwhelm.
More about Hina →- [1] Marketing Dive. Study: Personalized email subject lines increase open rates by 50%. 2025.
- [2] HubSpot. Subject line statistics: open rates and best practices. 2025.
- [3] Belkins. B2B Email Subject Line and Engagement Statistics. 2025.
- [4] Superhuman. Email subject line statistics for better open rates. 2025.
- [5] HubSpot. 2026 State of Marketing Report. 2026.
