Newsletters die at two specific moments, and most writers optimize neither: the subject line (where roughly 60% of readers decide whether to open) and the first sentence (open, scan, or close in about three seconds). These 12 prompts attack those moments first — 12 subject lines split across curiosity, contrarian, specific-number, and plain-promise, then five opening sentences that pull past the preview pane — before moving to body outlines (curation, deep-dive, personal essay), CTAs that earn the click, and a re-engagement flow for the dormant half of your list. Pair with the meta description prompts: subject-line craft and meta-craft are the same skill aimed at different inboxes.
TL;DR
- Spend half your time on the subject line. Keep it 30–50 characters; the first ~30 are all most mobile clients show. Two-to-four-word lines test best in 2026 data.
- Stop trusting open rate alone. Apple Mail Privacy Protection (MPP) auto-fires opens, inflating reported numbers by 15–20+ points. Optimize for click rate, which MPP cannot fake.
- Use the prompts in order: subject (1, 9) → first sentence (2) → body outline (3–6, 8) → CTA (10) → re-engagement (7) → post-send diagnosis (11) → next-quarter planning (12).
- Any current model handles these well. GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro all draft strong subject variations; Claude tends to hold voice across a long body paste better.
Best for
- Weekly creator and personal newsletters
- Product newsletters (release notes, updates)
- Curation-style newsletters with 3–5 picks
- Founder-led startup updates
- Re-engaging a dormant subscriber list
What “good” looks like in 2026
Benchmarks vary wildly by list quality and sender, but these are useful reference points (verify against your own ESP dashboard):
| Metric | Reference point (as of June 2026) | Source note |
|---|---|---|
| Subject line length | 30–50 chars; 2–4 words test best | First ~30 chars visible on mobile |
| Reported open rate | ~20% (excl. MPP) / ~34% (incl. MPP) | MPP inflates by 15–20+ pts |
| Strong send windows | Tue–Thu 9–11am; Friday rising | Always test against your list |
| The metric that matters | Click rate, not open rate | MPP cannot fake a click |
Treat open rate as a rough signal and let click rate be your real scoreboard.
Which model to use
Each prompt below is model-agnostic. For the big jobs:
- Subject-line batches and A/B pairs (#1, #9): any of GPT-5.5, Claude Sonnet 4.6, or Gemini 3.1 Pro produce 12 strong variants in one shot.
- Long-body work — rewriting a blog into a newsletter, or first sentences from a pasted draft (#2, #6): Claude Sonnet 4.6 / Opus 4.7 (1M-token context) and Gemini 3.1 Pro (1M) ingest a full long post without trimming. On ChatGPT Plus the in-app window is roughly 320 pages, which is plenty for a single issue.
- Send-day forensics and topic planning (#11, #12): a reasoning mode helps — GPT-5.5 Thinking or Claude Opus 4.7.
For more on getting a long draft into shape, see writing with Claude.
1. Subject lines that get opened
I am writing a newsletter on [topic]. Write 12 subject lines: 3 curiosity, 3 contrarian, 3 specific-number, 3 plain-promise. Keep each under 50 characters and front-load the first 30. Mark your top 3 picks and explain why.2. First sentence that earns the second
My newsletter body is below. Write 5 opening sentences that pull the reader past the preview line. Avoid "Hello again, friends." Lean concrete and specific.
[paste body]3. Curation newsletter outline
Outline a curation newsletter with 5 picks. For each: title (8 words max), a 2-line description that earns the click, and 1 contrarian "but watch out for X" line. Open with a 60-word personal note tying the picks together.4. Deep-dive newsletter outline
Outline a 1,000-word deep-dive on "[topic]". Structure: hook scene, the question I started with, what I learned (3 sections), what I am still unsure about, what to read next.5. Personal-essay newsletter
Outline a personal-essay newsletter about [experience]. Voice: warm, slightly self-deprecating, ends on insight not advice. 600 words.6. Newsletter from a long blog
Below is a long blog post. Strip it to a 350-word newsletter: 1-line hook, 3 takeaways, 1 personal note, 1 CTA. Keep the voice casual.
[paste]7. Re-engagement newsletter
Write a re-engagement newsletter for subscribers who have not opened in 60 days. Tone: honest, no guilt-tripping. Offer 2 options: stay (here is what is new), or leave (one-click unsubscribe). End with one clear ask.8. Launch announcement newsletter
I am launching [product]. Outline a launch newsletter: hook with the reader's problem, what I built, who it is for and not for, social proof or process note, 2 CTAs (try it, share it).9. Subject-line A/B test pairs
For this newsletter, write 5 subject-line pairs for A/B testing. Each pair: one curiosity line, one specific line. Briefly note what hypothesis each pair tests.
Topic: [topic]10. CTA that earns the click
My newsletter is about [topic] and ends with a CTA to [action]. Write 5 CTA variants that vary by style: imperative ("Try X"), question ("Want X?"), benefit-led ("Get X without Y"), curiosity ("See what X looks like"), and social ("Join 12K others who...").11. Send-day forensics
My last newsletter click rate was [X]% (baseline [Y]%). Below: subject line, preview text, first paragraph, send time, segment. Diagnose whether opens dropped, clicks dropped, or both, and whether the cause was subject, timing, segment, or content. Recommend 3 specific changes for the next send, ordered by likely lift.
[paste]Track click rate as your primary signal here. Reported opens are unreliable now that MPP auto-loads tracking pixels.
12. Topic-bank generator for the next quarter
I send a [weekly/biweekly] newsletter on [niche]. Based on my 10 best- and worst-performing past issues below, generate 12 topic ideas for the next quarter. For each: a tentative subject line, the angle, why it is likely to beat the average, and which past issue it would cannibalize. Cluster into 3 themes.
[paste]Common mistakes
- Spending 90% of effort on the body and 10% on the subject line — the subject decides whether anyone reads the body.
- A first sentence that explains what the newsletter is, instead of pulling the reader in.
- No clear CTA, or three competing CTAs that split attention.
- Same length and structure every week, so engagement decays from boredom.
- Never cleaning the list — dead subscribers tank your sender reputation and your deliverability.
- Judging sends by open rate alone in the MPP era; let click rate be the scoreboard.
FAQ
Is open rate still worth tracking in 2026? As a rough trend line, yes; as your main metric, no. Apple Mail Privacy Protection pre-fetches the tracking pixel, so a chunk of “opens” are bots, not readers — inflating reported opens by 15–20+ points. Use click rate, reply rate, or click-to-open as your real scoreboard.
How long should a subject line be? Aim for 30–50 characters, and front-load the first ~30 because that is roughly all most mobile clients display. In 2026 data, two-to-four-word lines tend to win, and performance drops past about seven words. Avoid ALL CAPS — it lowers opens and trips spam filters.
When should I send? Tuesday through Thursday, 9–11am local time, is the classic strong window, though 2026 data shows Friday performing surprisingly well on some lists. There is no universal best time — A/B test send time on your own list and segment by time zone.
Which AI model is best for newsletters? For subject batches, any of GPT-5.5, Claude Sonnet 4.6, or Gemini 3.1 Pro is fine. For pasting a full long draft and keeping your voice, Claude Sonnet 4.6/Opus 4.7 and Gemini 3.1 Pro (both 1M-token context) handle the length without trimming.
Will AI-written subject lines sound generic? They will if you accept the first draft. Generate 12, keep the 2–3 that match your voice, then rewrite by hand. The prompt does the divergent thinking; you do the taste.
Related
- Email subject line prompts
- Meta description prompts
- Sales copy prompts
- Brand story prompts
- How to Use AI to Write a Newsletter Issue: Theme, Three Links, Your Take, One CTA
- AI Newsletter Lead Magnet: PDF, Checklist, or Template That Converts
External: Brevo email benchmarks and Apple Mail Privacy Protection overview for the open-rate caveat.