Why Subject Lines Deserve Their Own AI Workflow

Your subject line is the only part of your email that competes in the inbox. Everything else — the copy, the design, the offer — only gets seen if the subject line wins.

Despite that, most email marketers spend the least time on it. Subject lines get written in the last five minutes before a send, with whatever energy is left after building the campaign.

AI changes that calculation. You can now generate ten strong subject line variations in under a minute, test psychological frameworks you'd never have time to think through manually, and iterate on tone and angle with almost zero friction.

This guide covers how to actually do it — with real prompts, before/after examples, and a workflow you can repeat for every campaign.

---

What Makes a Subject Line Work (Before You Involve AI)

AI is a force multiplier. But it multiplies whatever direction you point it in. Before you write a single prompt, it helps to know what you're aiming for.

The research on high-performing subject lines points to a few consistent factors:

Specificity beats cleverness. "Your cart expires tonight" outperforms "Don't miss out" almost every time. The more concrete the subject line, the less cognitive effort it requires to decide whether to open.

The first 40 characters carry the most weight. Mobile clients truncate subject lines aggressively. The value proposition — or the hook — needs to be front-loaded.

Curiosity works, but only when it's honest. Subject lines that tease without misleading earn the open and build trust. Subject lines that mislead earn the open and lose the subscriber.

Personalisation beyond first names. Using someone's name is table stakes. Behavioural personalisation — referencing a purchase, a category they browse, or a milestone — performs significantly better.

Keep these principles in front of you when you're reviewing AI output. They're your filter.

---

The Anatomy of a Good AI Prompt for Subject Lines

Generic prompt in, generic output out. The quality of your subject lines depends almost entirely on how much context you give the model.

Here's the information every strong subject line prompt should include:

• Who the audience is: shapes tone, vocabulary, and relevance.
• What the email is about: clarifies the offer, topic, or hook.
• The goal: tells the model whether to optimize for opens, clicks, or re-engagement.
• The desired tone: keeps output aligned with your brand voice.
• Character constraints: avoids mobile truncation and improves scanability.
• What to avoid: prevents spammy phrasing and off-brand wording.

Compare these two prompts:

"Write 5 subject lines for a promotional email."

"Write 5 subject lines for a promotional email going to lapsed subscribers who haven't opened in 60 days. The email offers 20% off their next order. Tone: warm and direct, not pushy. Keep each under 45 characters. Avoid using the word 'exclusive'."

The second prompt takes 30 seconds longer to write and produces dramatically different output.

---

Copy-Paste Prompts for Common Email Types

These prompts are ready to use in ChatGPT, Claude, or any large language model. Adjust the bracketed fields to your campaign.

---

Promotional / Sale Email

Write 8 email subject lines for a [DISCOUNT]% off sale running until [DATE] 
for [AUDIENCE]. The email leads with [KEY PRODUCT OR CATEGORY]. 
Tone: [urgent / conversational / playful]. 
Keep each under 50 characters. 
Produce two versions of each: one with the discount in the subject line, 
one without. Label them clearly.

Example output (20% off, skincare, ending Sunday):

• With discount
  • 20% off ends Sunday
  • Final hours: 20% off everything
  • Today: 20% off skincare
  • Save 20% before midnight Sunday
• Without discount
  • Your skin. This Sunday. Done.
  • Three days left (then full price)
  • We hate to see you pay full price
  • This one's for the Sunday scrollers

---

Re-engagement / Win-Back Email

Write 6 subject lines for a win-back email targeting subscribers 
who haven't opened in [X DAYS]. 
The email acknowledges the gap without being guilt-tripping. 
It offers [INCENTIVE / VALUE]. 
Tone: warm, self-aware, slightly playful. 
Under 50 characters.

Example output (90-day lapse, no incentive, SaaS product):

• "Still there? (No pressure)"
• "We cleaned up while you were gone"
• "Your account is waiting. So are we."
• "Things have changed (for the better)"
• "Pick up where you left off?"
• "Miss us? We definitely missed you"

---

Newsletter / Content Email

Write 7 subject lines for a weekly newsletter email. 
Main story: [TOPIC IN ONE SENTENCE]. 
Secondary stories: [BRIEF BULLETS].
Audience: [WHO THEY ARE AND WHAT THEY CARE ABOUT].
Tone: smart, not academic. Curious, not clickbaity.
Under 55 characters.

Example output (email marketing analytics newsletter):

• "The metric your dashboard is hiding"
• "Why your open rate is lying to you"
• "Three numbers that actually matter"
• "What good looks like in 2026"
• "The benchmark problem no one talks about"
• "Your campaign worked. Did it though?"
• "Reading the gap between opens and revenue"

---

Abandoned Cart Email

Write 6 subject lines for an abandoned cart email. 
The product abandoned: [PRODUCT NAME / CATEGORY]. 
Price point: [PRICE].
Time since abandonment: [X HOURS].
Tone: [helpful / light urgency / conversational].
No countdown language unless specified.
Under 45 characters.

Example output (running shoes, £85, 4 hours):

• "Your trainers are still here"
• "Left something behind?"
• "These won't last (neither will your cart)"
• "Still thinking about them?"
• "Your basket hasn't forgotten you"
• "One click from new trainers"

---

Transactional / Post-Purchase Email

Write 5 subject lines for a post-purchase email sent immediately after checkout. 
Product: [PRODUCT]. 
Goal: confirm the order and set up the next action (review, referral, related product).
Tone: warm and efficient. Not over-excited.
Under 50 characters.

Example output (online course purchase):

• "You're in. Here's what's next."
• "Order confirmed — let's get started"
• "Your course is ready"
• "Welcome. Here's your access link."
• "Done. Now the good part."

---

A Workflow That Actually Improves Over Time

Generating subject lines with AI is useful. Building a system around it is where the real leverage comes from.

Here's a repeatable workflow that compounds with every campaign:

Step 1: Brief the model properly

Use the prompt framework above. The 60 seconds it takes to write a proper prompt pays for itself in the quality of the first output.

Step 2: Generate more than you need

Ask for eight to ten options. You'll discard most of them, but the volume forces the model to explore different angles — urgency, curiosity, specificity, humour, directness. One of those angles usually turns out to be better than what you'd have written manually.

Step 3: Apply your own filter

Review the output against what you know about your audience. Ask:

• Is this honest about what's in the email?
• Does it sound like us?
• Would this work in the first 40 characters on mobile?
• Does it avoid the patterns our list has seen too many times?

Shortlist two or three.

Step 4: A/B test — always

Pick your top two and split your send. Even a 70/30 split gives you meaningful data over time. The goal isn't to find one winning subject line. It's to build a mental model of what your audience responds to.

Step 5: Log what wins

Keep a running record of subject lines that outperformed and underperformed, with context: the list segment, the type of email, the open rate, the date. Feed this back into your AI prompts.

"Based on these past high-performing subject lines from our newsletter: [LIST]. Write 8 new subject lines for an email about [TOPIC] in a similar style."

This is how AI output gets progressively better — not through the model improving, but through the quality of your prompts improving.

---

Prompts for A/B Testing Frameworks

AI is especially useful for generating systematic A/B test variations. Rather than testing random ideas, you can test specific hypotheses.

Test: Curiosity vs specificity

Write two versions of a subject line for the same email about [TOPIC].
Version A: curiosity-driven — hints at the content without revealing it.
Version B: specific — states exactly what's in the email.
Both under 50 characters.

Test: With vs without personalisation

Write the same subject line twice.
Version A: no personalisation.
Version B: includes [FIRST NAME] or references [BEHAVIOUR / PURCHASE / SEGMENT].
Email is about [TOPIC]. Tone: [TONE].

Test: Urgency vs no urgency

Write one subject line for [EMAIL TYPE] that includes a deadline or scarcity signal.
Write a second version of the same subject line without any urgency language.
Both should feel natural for a [TONE] brand. Under 45 characters.

Test: Question vs statement

Write two subject lines for an email about [TOPIC].
Version A: a question.
Version B: a declarative statement.
Same information, different format. Under 50 characters each.

Running these systematic tests over three to six months builds something much more valuable than any individual winning subject line — it builds an understanding of what your specific audience actually responds to.

---

What AI Gets Wrong (and How to Catch It)

AI subject lines fail in predictable ways. Knowing the failure modes makes them easy to spot and fix.

Over-reliance on urgency language. If you don't specify tone, most models default to "Don't miss out", "Last chance", "Hurry" — patterns that feel worn out to most subscribers. Explicitly tell the model to avoid urgency language unless the email genuinely has a deadline.

Generic hooks that don't reflect your brand. AI defaults to a neutral, marketing-speak register. If your brand voice is dry and direct, or warm and informal, you need to specify that — and ideally give examples.

Subject lines that overpromise. AI will sometimes write a compelling hook that your email can't actually deliver on. Always check: does the subject line match the content of the email? Mismatched expectations are the fastest way to increase unsubscribe rates.

Too long for mobile. Without character constraints in the prompt, models often write subject lines that are 60-80 characters — fine for desktop, truncated on mobile. Always specify a character limit.

Spam trigger patterns. Excessive capitalisation, certain financial terms, and some urgency phrases can affect deliverability. If you're generating at volume, it's worth running output through a subject line tester before sending.

---

Before and After: The Difference a Good Prompt Makes

These are real rewrites using AI, going from a vague first draft to a stronger version.

• Flash sale: "Big sale this weekend!" -> "48 hours. 30% off. Starts now."
• Re-engagement: "We miss you" -> "Still there? We kept the lights on."
• Content newsletter: "This week's email" -> "The benchmark that's misleading you"
• Product launch: "New feature alert" -> "You asked for this one. It's here."
• Abandoned cart: "You left something behind" -> "Still thinking about it?"
• Post-purchase: "Thanks for your order!" -> "You're in. Here's what's next."

The improvement isn't magical. It's the difference between writing a subject line in a hurry versus spending thirty seconds on a proper prompt and filtering the output.

---

Quick Reference: Subject Line Prompt Templates

Bookmark this section. These templates cover the most common email types and are ready to drop into any AI tool.

Promotional:

"Write 8 subject lines for a [%] off sale ending [DATE] for [AUDIENCE]. Tone: [TONE]. Under 50 characters. No exclamation marks."

Re-engagement:

"Write 6 win-back subject lines for subscribers inactive for [X] days. Warm and self-aware tone. No guilt language. Under 45 characters."

Newsletter:

"Write 7 subject lines for a newsletter about [TOPIC]. Smart, curious tone. Not clickbaity. Under 55 characters."

Abandoned cart:

"Write 6 subject lines for an abandoned cart containing [PRODUCT]. [X] hours since abandonment. Conversational, not urgent. Under 45 characters."

Post-purchase:

"Write 5 subject lines for an order confirmation email for [PRODUCT]. Warm and efficient. Set up the next action. Under 50 characters."

A/B test pair:

"Write one curiosity-based and one specificity-based subject line for the same email about [TOPIC]. Label each. Under 50 characters."

---

The Bottom Line

AI doesn't write great subject lines on its own. It writes a large volume of subject lines quickly — and some of them are great.

The marketer's job shifts from writing to directing and filtering. That's a better use of your time. A well-written prompt plus thirty seconds of human judgment produces better output than either alone.

The marketers who get the most from AI subject line generation aren't the ones who generate the most variations. They're the ones who pair good prompts with good testing — and use what they learn to write better prompts next time.

---

Related Articles

• Why Some Subject Lines Feel Impossible to Ignore
• Email A/B Testing Reporting: What Metrics Should You Compare?
• Why Email Open Rate Is a Misleading Metric (And What to Track Instead)
• The Hidden Cost of Resending Emails to Non-Openers
• Inbox 2.0: How AI Is Changing What Gets Opened, Clicked, and Ignored