Email marketing dies at the unsubscribe rate, not the click rate. Every send that earns the click but loses the next open is borrowing from a list balance you can’t refill. Worse, in 2026 a sloppy list is a deliverability problem: Gmail classifies any domain sending 5,000+ messages a day as a bulk sender, permanently, and enforces a spam-complaint rate at or above 0.30% by withholding delivery help. Google’s own guidance is to stay under 0.10%. So the job of every email is two-sided: earn the next open, and never earn a spam click.
These 12 prompts focus on exactly that — welcome flows that pay value before they ask, cart recovery that handles the specific objection (not “you forgot something!”), post-purchase that drives a review without sounding desperate, and re-engagement that respects the inactive subscriber. The last prompt is a deliverability audit tuned to the current Gmail and Yahoo bulk-sender rules.
For Shopify lists, pair the lifecycle emails below with strong on-site discovery — our AI Shopify collection description workflow turns category pages into the “how to choose” landing pad your email CTAs deserve.
TL;DR
- Copy a template, swap the
[bracketed]placeholders, paste into any model. All 12 prompts work in ChatGPT (GPT-5.5), Claude (Sonnet 4.6 / Opus 4.7), or Gemini 3.1 Pro — no special setup. - Best model for email copy as of June 2026: Claude Sonnet 4.6 for voice consistency across a sequence; GPT-5.5 Instant for fast subject-line volume; Gemini 3.1 Pro when you paste a long source doc (1M-token context).
- Two non-negotiables baked into the prompts: one clear CTA per email, and a real reason behind every discount.
- Prompt 11 is the deliverability gatekeeper — run it before any campaign that touches more than a few hundred Gmail/Yahoo inboxes.
What “good” looks like (2026 benchmarks)
Use these as guardrails, not goals. Numbers vary by industry and list age, and Apple Mail Privacy Protection inflates open rates by auto-loading images, so treat opens as directional and judge real lift on clicks and revenue.
| Metric | Ecommerce average (2026) | Notes |
|---|---|---|
| Open rate (campaign) | ~30–33% MPP-inclusive; ~20% excluding bot opens | Apple MPP auto-opens skew this high |
| Click-through rate | ~2.1% | Broadcast promos |
| Cart-recovery open rate | ~45–50% | Triggered flows beat broadcasts |
| Cart-recovery conversion | 3–5% average; 10%+ is excellent | Top stores hit ~$28 revenue/recipient |
| Spam-complaint rate | Keep under 0.10% | 0.30%+ triggers Gmail/Yahoo enforcement |
Sources: Brevo 2026 benchmarks, Google sender guidelines.
Best for
- DTC brands running lifecycle flows
- SaaS onboarding and retention sequences
- Creator newsletters
- E-commerce promotions and segment splits
How to use these prompts
Each block is a copy-ready template. Replace every [bracketed placeholder] with your specifics, then paste into your model of choice. For a multi-email sequence, keep one chat thread open so the model carries voice and offer context between emails. If you want the model to match an existing brand voice, paste 2–3 of your best past emails first and add “match this voice” to the prompt.
1. Welcome series (3 emails)
For brand [brand name], design a 3-email welcome series:
D0 (delivery + onboarding), D2 (value-first content, no sale),
D5 (introduce offer). For each: subject line + 150-word body.2. Cart abandonment email
For [product type], write a 3-email cart abandonment sequence:
1h (gentle reminder), 24h (objection handling), 72h (last-chance +
small incentive). Each email 120 words or fewer. One clear CTA per email.3. Post-purchase delight email
For [product], write a post-purchase email sent 7 days after delivery.
Goal: drive a review + repeat purchase. Voice: warm, no salesy push.
150 words or fewer.4. Re-engagement (60-day inactive)
Subscribers who haven't opened in 60 days. Write a 3-email sequence:
nudge with new content, ask for a preference update, soft farewell
with one-click unsubscribe. Honest tone, no guilt.5. Product launch email
Launching [product]. Write a launch email to the existing list:
hook with the problem we solve, what's new, who it's for, one CTA.
200 words or fewer.6. Lifecycle “you missed this” email
For a SaaS user who signed up but didn't finish onboarding: write an
80-word email that nudges them to step [N] with one specific benefit.
No guilt, one CTA.7. Promotion without sounding cheap
For [brand], write a 20%-off promotion email that doesn't devalue the
brand. Frame the discount with a real reason (end-of-season, anniversary,
restock). The reason must come before the discount in the copy.8. Newsletter from a long blog
Below is a 1,500-word blog post. Convert it to a 350-word newsletter
version: hook, 3 takeaways, 1 personal note, 1 CTA.
[paste blog post]Tip: for very long source docs, run this in Gemini 3.1 Pro or Claude — both hold a 1M-token context, so you can paste the full article instead of trimming it first.
9. Subject + preheader A/B set
For an email about "[topic]", write 8 subject + preheader pairs.
Vary: question hook, curiosity gap, specific benefit, gentle urgency,
social proof, plain-statement, personal-confession, contrarian.
Subject 50 chars or fewer, preheader 90 chars or fewer. Mark which
avoid spam triggers and which fit a transactional vs promo voice.10. Segment-aware promo split
For a 15%-off promo on [product], write 3 segment-tailored variants:
first-time browser (no purchase history), past-customer (1+ purchase),
high-LTV (3+ purchases). Same discount, three different angles. The
high-LTV version must NOT use urgency tactics — it earns the open via
signal, not pressure.11. Deliverability + spam-trigger audit
Below is my email draft + subject + preheader: [paste].
Flag: (a) words that risk spam filters (FREE!!!, ACT NOW),
(b) imbalance of image vs text, (c) all-caps lines,
(d) overly long preview text, (e) missing plain-text version cues,
(f) any missing or unclear one-click unsubscribe cue.
Output a rewritten version that preserves the message but lifts
deliverability, targeting a spam-complaint rate under 0.10%.12. Lifecycle map for a new SaaS
For SaaS [product] across signup → trial → paid → retention, design a
6-email lifecycle map. For each email: trigger event, day offset,
single goal, single CTA, success metric. End with the 1 email I should
send if a user goes silent for 21 days.Deliverability in 2026: the rules these prompts assume
Prompt 11 only matters because the inbox rules tightened. As of June 2026, if you send bulk mail to Gmail or Yahoo:
- Authenticate fully. SPF, DKIM, and a published DMARC record are required for bulk senders.
p=noneis an acceptable starting point, but inbox providers expect you to progress towardp=quarantineorp=reject. - One-click unsubscribe is mandatory. Bulk senders must support RFC 8058 — a
List-Unsubscribeheader plus aList-Unsubscribe-Postheader carryingList-Unsubscribe=One-Click. Your email platform sets this; verify it’s on. You also have 48 hours to honor any unsubscribe, so suppress, don’t just queue. - Keep complaints under 0.10%. Gmail enforces at 0.30%, but that’s the danger line, not a target. The cleanest way to stay low: prompt 4’s honest re-engagement, then suppress non-openers instead of mailing them forever.
No prompt fixes a bad domain reputation. These help you write copy that doesn’t trip filters or trigger complaints, which is the part you control message by message.
Common mistakes
- Same subject template across every send. Opens decay fast when the inbox stops being surprised. Rotate the angles from prompt 9.
- Two paragraphs explaining the email instead of doing the thing. Lead with the value, not the setup.
- No single clear CTA. Multiple competing asks mean none get clicked.
- Discounts with no reason. A random “15% off!” trains subscribers to wait for the next one. Prompt 7 forces a reason.
- Guilt-trip re-engagement (“we miss you”) instead of giving a reason to stay or an easy exit.
- One-size-fits-all promos across high-LTV and first-time segments. Different temperature, different ask — that’s what prompt 10 fixes.
- Mailing dead addresses to “keep volume up.” It raises your complaint rate and drags every other email’s placement down.
FAQ
Which AI model writes the best email copy in June 2026? For a multi-email sequence where voice must stay consistent, Claude Sonnet 4.6 tends to hold tone best. For fast subject-line volume, GPT-5.5 in Instant mode is quickest. For converting a long source document into a newsletter, Gemini 3.1 Pro or Claude both have a 1M-token context, so you can paste the full draft. Any of the three handle these prompts well — pick by workflow, not hype.
Will AI-written emails hurt my deliverability? Not by being AI-written. Filters score sender reputation, authentication, and complaint rate — not authorship. What hurts you is spammy copy (all-caps, “FREE!!!”, image-only emails) and mailing people who don’t want it. Run prompt 11 on anything going to a real list.
Do I still need A/B testing if AI writes 8 subject lines? Yes. The model gives you varied, on-brief candidates fast (prompt 9), but only your audience tells you what works. Test 2 of the 8, send the winner to the rest, and feed the result back into the next prompt as context.
What’s a safe spam-complaint rate? Stay under 0.10%. Gmail and Yahoo begin withholding delivery mitigation at 0.30%, so anything approaching that is already a problem. The biggest lever is list hygiene: re-engage with prompt 4, then suppress subscribers who still don’t open.
How short should marketing emails be? Shorter than instinct says. The prompts cap most bodies at 80–200 words because a single clear message with one CTA outperforms a wall of copy. Save the depth for the landing page the email links to.