The subject line is the only part of your email 95% of recipients judge before deciding to open. A useful subject-line prompt names the audience, the email type, and one promise, then returns 5-7 variants so you can A/B test instead of agonizing over a single guess.
These templates work in any current chat model: ChatGPT (GPT-5.5), Claude (Sonnet 4.6), or Gemini 3.1 Pro. None of them needs a paid tier — the free models write subject lines fine, since the task is short. The difference between a flat result and a sharp one is the brief, not the model.
TL;DR
- Length: aim for 30-50 characters. iPhone portrait shows roughly 41 characters; the Gmail mobile app cuts off near 30 (data as of June 2026). Front-load the verb, number, or surprise.
- Generate 5-7 variants, not 20. Past seven you are ranking noise, not concepts.
- Open rate is now a weak metric. Apple Mail Privacy Protection pre-loads tracking pixels and accounts for roughly half of all recorded opens, inflating reported open rates by an estimated 15-20 percentage points. Optimize for click-through and replies instead.
- Always pair the subject with preview text that extends it, never repeats it.
- The 12 prompts below cover newsletters, onboarding, re-engagement, cold sales, A/B briefs, and failure diagnosis.
Who this is for
Newsletter writers, lifecycle marketers, founders writing onboarding sequences, and indie developers who send a monthly product update.
When not to use these prompts
Skip them for transactional emails (receipts, password resets) — the body carries the value there and a clever subject only adds friction. And never run a curiosity subject the body cannot pay off; a one-time open bought with a misleading line trains subscribers to ignore you.
What changed: why open rate stopped being the target
Two shifts make 2026 subject-line strategy different from a few years ago:
- Apple Mail Privacy Protection (MPP) silently loads tracking pixels for Apple Mail users, marking emails as “opened” whether or not a human read them. As of June 2026 MPP accounts for close to half of all tracked opens, and the inflation runs an estimated 15-20 points above true human reads. A reported 40% open rate often maps to 25-30% real engagement.
- Gmail and Microsoft tightened bounce and spam thresholds. Crossing them now gets a sender suppressed, so deliverability — not a clever subject — is the first gate.
The practical takeaway: write subjects to drive a deliberate action you can actually measure. Across all industries the average campaign click rate sits near 1.7%, with the top 10% of senders around 3.4% and automated flows higher still (data as of June 2026). Click-through and reply rate are the numbers your prompts should optimize for.
Prompt anatomy: the six elements
Every strong subject-line prompt carries six things. Miss one and the output drifts to generic.
| Element | What it does | Weak vs. strong |
|---|---|---|
| Audience | One specific reader | ”marketers” → “solo SaaS founders with <50 customers” |
| Goal | One outcome | read / click / reply / share — pick one |
| Voice | 2-3 anchor adjectives + a sample line | ”be friendly” is noise; a real sample is signal |
| Constraints | Char cap, banned phrases, must-include facts | ”≤ 50 chars, no emoji, must mention the 12% number” |
| Format | How variants are returned | numbered list, table, subject-plus-preview pairs |
| Examples | 1-2 of your own best past subjects | the single strongest lever for matching voice |
12 copy-ready prompt templates
Swap the [bracketed] placeholders before running each prompt.
1. 7 variants from a brief
Email type: [type]. Audience: [audience]. Email promise: [promise].
Write 7 subject-line variants, max 60 characters each, one per angle:
(a) curiosity, (b) benefit-led, (c) specific number, (d) contrarian,
(e) name-drop, (f) question, (g) plain. No emojis unless I ask.
Label each line with its angle.Variables to swap: type — newsletter / onboarding / re-engage / sales; audience; promise
2. Subject + preview pair
For this subject "[subject]", write 3 preview-text pairings, max 90
characters each. Each preview must EXTEND the subject, not repeat it.
Read subject + preview aloud as one sentence; it should land like a
cliffhanger.Variables to swap: subject
3. Curiosity without clickbait
Write 5 curiosity-driven subjects that the email body fully pays off.
Forbidden: "You won't believe...", vague numbers, "this one trick".
Each subject must be 100% true given the email content I describe: [content].Variables to swap: content
4. Benefit-led subjects
Audience cares about: [outcome]. Write 5 benefit-led subjects, max 50
characters. Lead with a verb. Drop "how to" when a tighter line works
("Cut churn 12%" beats "How to cut churn").Variables to swap: outcome
5. Re-engagement subjects
Write 5 re-engagement subjects for a subscriber inactive 60+ days.
Tones: (a) self-aware ("We get it..."), (b) plain ("Still want X?"),
(c) value-led ("New feature you missed"), (d) farewell ("Should we
say goodbye?"), (e) reset ("Want fewer emails?"). Add a one-line
preview for each.6. Onboarding sequence subjects
For a 5-email onboarding sequence, write subjects that build a narrative
arc: Day 1 welcome, Day 2 quick win, Day 4 dig deeper, Day 7 advanced,
Day 14 community. Each max 50 characters. Numbered prefix optional.7. Sales outreach subjects
Cold sales subject. Audience: [persona] at [companySize]. We solve:
[problem]. Write 5 subjects, max 40 characters, that don't pattern-match
to spam. Skip "Quick question" and "[first name], are you the right
person?" — both overused.Variables to swap: persona, companySize, problem
8. A/B test brief
I want to A/B test subjects. Output a brief: (1) Hypothesis, (2) Variant A,
(3) Variant B, (4) Metric — pick click or reply, not open (Apple MPP
inflates opens), (5) Sample size needed, (6) Stop conditions. Test a
concept, not capitalisation.9. Mobile-truncation safe
Rewrite these subjects to survive 30-character mobile truncation (Gmail
app, iPhone portrait). Move the verb, number, or surprise into the first
30 characters. Output a table: original | mobile-safe | what survived.
Subjects: [subjectList].Variables to swap: subjectList
10. Brand-voice subject check
Audit these subjects against our brand voice: [subjectList]. Score each
in-voice (5) to off-voice (1). For anything below 4, suggest one rewrite.
Flag any line that sounds like every other newsletter.Variables to swap: subjectList
11. Localised subject line
Localise these subjects from English to [lang] without literal
translation — use the local idiom. Re-test each for 30-character mobile
truncation. Skip emoji if that locale's inbox renders them oddly.
Subjects: [subjectList].Variables to swap: lang, subjectList
12. Subject-line failure forensics
My last 5 subjects had open rates 18 / 22 / 14 / 28 / 16% (avg 20%).
Diagnose: (1) which subjects drove the spread, (2) hypotheses — audience
fatigue, topic mismatch, send timing, (3) one experiment to run next.
Note: opens are MPP-inflated, so weight click data if I give it. Max 200 words.Common mistakes
- Vague audience (“anyone who…”) — the output reads generic every time.
- No tone anchor — every variant comes back the same flavour.
- No constraints — without a char cap and banned-phrase list, the model defaults to the safe middle.
- Skipping your own examples — they are the strongest signal for matching voice.
- Trusting the first draft — AI lands on the average; your job is to break it.
- Asking the model to optimize “opens” — that target is now distorted by Apple MPP.
- No fact-check pass — the model will occasionally invent a confident, wrong number.
How to push results further
- Supply 1-2 real past subjects that performed; “be friendly” alone tells the model nothing.
- Constrain hard: word count, banned phrases, the one fact that must appear.
- Read every finalist aloud — if you stumble, rewrite.
- Cut adjectives and adverbs that carry no weight.
- Use AI for drafts one and two, then edit by hand for draft three. Draft three ships.
- Anchor the brief to one real person from your customer list, not a persona.
- Test the subject without the body: does the message still survive on its own?
FAQ
- What subject length should I target? 30-50 characters for most inboxes. iPhone portrait shows about 41 characters and the Gmail mobile app truncates near 30 (as of June 2026), so front-load the verb or number.
- Do emojis help? It depends on the list. In B2C, an emoji has lifted open rate meaningfully; in B2B it tends to flatten or slightly lower it. Across tested campaigns emoji subjects showed higher click-through in most cases, but limit yourself to one or two — more reads as spam. Always A/B it.
- Should I use first-name personalisation? Sparingly. It is so overused in cold email that “[first name], quick question” now signals automation. Personalise on behaviour (what they did) over identity (their name).
- How many variants should I generate? 5-7. Past seven you are evaluating noise rather than distinct concepts.
- Does capitalisation matter? Sentence case usually feels less promotional than Title Case for newsletters. ALL CAPS and excess punctuation can also trip spam filters.
- What is a good open rate now? It is category-dependent and unreliable since Apple MPP inflates the number by roughly 15-20 points. Watch click-through (industry average near 1.7%, as of June 2026) and reply rate instead.
Related
- Newsletter prompts
- Email writing prompts
- Sales copy prompts
- Tone rewrite prompts
- Writing & Copywriting Prompts hub
Tags: #Prompt #Writing #Email #Subject line