5 cart-abandonment emails written by AI — full sequence with timing and copy
Contents
The first time I built a cart-abandonment (弃购挽回) sequence, I spent three days writing it. Five emails, two rounds of revisions, a copywriter to polish the subject lines, and a designer to lay them out. The sequence recovered about 8% of abandoned carts over a quarter. Not bad — but it took me a quarter to learn whether it worked.
Last year I rebuilt the same flow with AI. I drafted all five emails in an afternoon, the open rate climbed 22%, and the recovery rate hit 11.4%. The lift came from being able to AB test more variants per email, not from AI writing better copy than I would have. Let me show you the exact sequence I use, the timing, the actual copy, and the prompts that generated it.
The benchmark most stores leak against
Across ecommerce (电商), cart abandonment hovers between 70% and 77% depending on the industry. The Baymard Institute's long-running studies put the average at roughly 70%. In fashion and luxury it climbs to 80%+, in electronics it's closer to 65%. The top three reasons shoppers leave are predictable: extra costs (shipping, tax) at 48%, required account creation at 24%, and a checkout that felt too slow or complicated at 18%.
You cannot fix those reasons by emailing harder. But the 70% who leave aren't a lost cause — roughly 10% to 15% of them will come back if you send a smart sequence. Multiply that by your abandoned-cart volume and the number stops being academic. On a Shopify store doing $200k/month, a 5% lift in recovery is a six-figure annual swing.
The sequence below assumes a non-food, direct-to-consumer product: apparel, beauty, home goods, supplements, gadgets. Food and consumables shift toward subscription and reorder flows; high-AOV (average order value, 平均客单价) items like furniture and B2B need longer windows. Adjust the windows; the structure holds.
The sequence, email by email
| # | Send time | Goal | Tone |
|---|---|---|---|
| 1 | 1 hour after abandon | Re-open the tab | Helpful, low-pressure |
| 2 | 24 hours after abandon | Remove the objection that stopped them | Reassuring, specific |
| 3 | 72 hours after abandon | Rekindle desire with social proof | Friendly, slightly playful |
| 4 | 5 days after abandon | Discount with a hard deadline | Direct, urgent |
| 5 | 10 days after abandon | Final breakup / last chance | Honest, brief |
Five emails is the sweet spot. Three leaves money on the table; seven burns the list. The Honest Company, Chubbies, and Everlane all run sequences in this range.
Email 1 — the 1-hour reminder
This is the highest-open, highest-click email in the entire sequence. Send it fast. The shopper still has working memory of your product, and a one-hour gap feels like a coincidence, not a chase. Do not include a discount — you'll train your list to wait for one.
Subject line: You left something in your cart
Preview text: It's still here if you want it.
Body copy:
Hi {{first_name}},
You added the {{product_name}} to your cart an hour ago and stepped away. No pressure — your cart is still saved and waiting.
[Product image with "Complete your order" button]
If you had a question that didn't get answered, just reply to this email. A real person on our team reads every response.
— The {{brand_name}} team
The AI prompt I used (Claude / GPT-4, pick either):
You are an email copywriter for a DTC (direct-to-consumer) brand selling
{{product_category}}. Write a 1-hour cart-abandonment reminder email.
Constraints:
- 80 words or fewer in the body
- No discount or coupon mention
- One CTA (call to action, 行动号召) button, copy: "Complete your order"
- Subject line under 50 characters
- Preview text under 90 characters
- Tone: helpful, low-pressure, like a friend pointing out something you forgot
- Mention that a real human reads replies
- Personalize with {{first_name}} and {{product_name}}
Output: subject line, preview text, body. Plain text only.Run this prompt. Then trim 20% of the words. AI writes too long by default — your real copy is the one with the second sentence cut.
Email 2 — the 24-hour objection handler
Most carts die for a reason. Your job is to guess the right one and answer it. If your analytics show shipping costs are killing you, lead with free-shipping or a shipping calculator. If returns are a common hesitation, lead with the policy. Don't know the reason? Use your highest-confidence bet and AB test the alternative next month.
Subject line: Quick question about your {{product_name}}
Preview text: Want to make checkout easier?
Body copy:
Hi {{first_name}},
A few customers told us checkout wasn't clear enough, so we want to flag the things that might have stopped you yesterday:
- Free shipping on orders over $50 (yours qualifies)
- 30-day returns, no restocking fee
- Checkout as guest — no account required
Your cart is one click away: [Resume checkout]
— {{brand_name}}
The AI prompt:
Write the 24-hour cart-abandonment follow-up email for a DTC brand selling
{{product_category}}.
The goal is to address the most common checkout objections.
The top three objections in our category are:
1. {{objection_1}} (e.g., "shipping cost was higher than expected")
2. {{objection_2}} (e.g., "didn't want to create an account")
3. {{objection_3}} (e.g., "wasn't sure about the return policy")
Constraints:
- Each objection becomes a bullet that answers it
- Subject line asks a question, under 50 characters
- 90 words or fewer
- One CTA, copy: "Resume checkout"
- No discount
- Personalize with {{first_name}} and {{product_name}}
Output: subject line, preview text, body bullets, closing.Bullet three works because real objections are specific. The AI will write a generic "we want you to feel confident" if you don't list the actual objections. Feed it the three your analytics show.
Email 3 — the 72-hour social proof nudge
By day three, desire has cooled. The shopper doesn't need another reminder; they need a reason to come back. Social proof is the cleanest lever: a UGC (user-generated content, 用户生成内容) photo, a press mention, a "this is our most-returned item" line. Don't fabricate reviews. Pull them from your real review platform.
Subject line: {{first_name}}, see what {{reviewer_name}} just said
Preview text: 4.8 stars from 1,200+ reviews
Body copy:
Hi {{first_name}},
Wanted to share what a customer said about the {{product_name}} this week:
"I've been wearing this three times a week since it arrived. The fit is exactly what the photos suggested, and the fabric has held up to a dozen washes already." — {{reviewer_name}}, verified buyer
Your cart is still waiting: [Pick up where you left off]
— {{brand_name}}
The AI prompt:
Write a 72-hour cart-abandonment email for a DTC brand.
The product is {{product_name}}, average rating {{rating}} from
{{review_count}} reviews.
Pick a review from this list and quote it verbatim — do not paraphrase:
{{paste 3-5 real reviews}}
Constraints:
- Subject line includes the reviewer's first name and the rating
- Body quotes ONE review in full, attributed
- Under 100 words total
- One CTA, copy: "Pick up where you left off"
- No discount
- Do not invent quotes or reviewers; only use what's in the list
Output: subject line, preview text, body, attribution line.The most important constraint is the last one. AI will hallucinate reviews — confident, polished, fictional quotes. By forcing the prompt to pull from a real list and forbid invention, you sidestep the single biggest trust failure in AI-generated email.
Email 4 — the 5-day discount email
Now you bring out the discount. The 5-day gap is enough that you've earned the right to give one. The discount needs a deadline — a real one, not "limited time" — and the deadline needs to be short: 24 to 48 hours. A deadline measured in days is invisible; a deadline measured in hours changes behavior.
Subject line: 10% off, only for the next 36 hours
Preview text: Use code COMEBACK10 at checkout.
Body copy:
Hi {{first_name}},
We held the {{product_name}} in your cart for five days. We can't hold it forever.
Use code COMEBACK10 in the next 36 hours for 10% off your order.
[Resume checkout]
After 36 hours the code expires and the cart is released.
— {{brand_name}}
The AI prompt:
Write a 5-day cart-abandonment discount email for a DTC brand.
Discount: {{discount_value}} (e.g., 10% off or $15 off)
Code: {{code}}
Expiry: {{expiry_window}} from when the email is opened
Constraints:
- Subject line includes the discount and the time window
- Code is bolded in the body
- Mention that the code expires and the cart is released
- 70 words or fewer
- One CTA, copy: "Resume checkout"
- Do not include a "no purchase necessary" or legal disclaimer; legal will add it
- Match the brand's tone: {{tone}} (e.g., warm, witty, plainspoken)
Output: subject line, preview text, body.Two important things AI gets wrong here. First, it loves to add "exclusions apply" and "while supplies last" boilerplate — that eats words and adds no conversion lift. Cut it. Second, it writes deadlines like "ends Sunday" instead of "in 36 hours." Concrete hour-based urgency outperforms named-day urgency in my tests, roughly 14% higher click-through on this specific email.
Email 5 — the 10-day breakup email
This is the email most brands skip. It's the email that does two things at once: gives the most stubborn shoppers one last shot, and gives the rest of the list permission to stop hearing from you for a while. A breakup email has a higher unsubscribe rate than the others — and that's the point. People who stay past email 5 are your best customers. People who leave are saving you deliverability problems.
Subject line: Last call for the {{product_name}}
Preview text: After this, we'll stop emailing you about it.
Body copy:
Hi {{first_name}},
This is the last cart reminder we'll send about the {{product_name}}. After today, the cart is released and we'll only email you about things you've actually asked for.
If now's the right time: [Complete your order]
If not, no hard feelings. You can update your preferences anytime at the link below.
— {{brand_name}}
The AI prompt:
Write the final 5th email in a 10-day cart-abandonment sequence.
This is the breakup / last-chance email.
Tone: honest, brief, slightly self-aware. No begging, no fake urgency.
The email should:
- Acknowledge this is the last cart email about this product
- Give the shopper a clear out (unsubscribe or update preferences)
- Offer one last chance to convert
- Reinforce that the list is respected, not spammed
Constraints:
- Under 80 words
- Subject line: "Last call for the {{product_name}}"
- One CTA, copy: "Complete your order"
- One secondary link, copy: "Update email preferences"
- No discount
- No exclamation marks
Output: subject line, preview text, body.A 10-day window is a judgement call. For apparel with seasonal urgency, 7 days works better. For supplements and consumables, push to 14 days because the next purchase cycle is closer. The structure is the same.
What AI does well, and what it ruins
A few patterns I trust AI to handle: subject line variations at scale, body copy under tight word limits, and tonal rephrasing ("rewrite this in a more playful voice"). On a sequence like this, AI can produce 10 to 20 subject line candidates in a few minutes — including the weird, off-pattern ones that turn out to be the winners. Human writers self-edit away from "ugly" subject lines; AI doesn't care.
A few patterns I won't trust AI to handle: invented reviews and testimonials, anything that names a specific price or shipping threshold you haven't confirmed, and legal-adjacent copy like discount exclusions. AI is also weak at humor and at long-form storytelling. The discount email above has zero jokes on purpose.
Two practical rules when you ship these into your ESP (Email Service Provider, 邮件营销平台 — Klaviyo, Mailchimp, Omnisend, etc.):
1. Always personalize with live merge tags, not AI-generated names. AI will write a nice email, then put "Hi friend" in the salutation because no merge tag was passed. The merge tag system is the personalization. Don't let the model fake it.
2. Always add a real reply-to address. The phrase "a real person on our team reads every reply" in email 1 is a conversion driver — but only if it's true. A no-reply@ address kills the trust. If you're going to promise a human, route the inbox to a person.
A quick way to test this works for you
Don't ship the full five-email sequence on day one. Ship email 1 and email 4 only. The reminder and the discount. Run them for two weeks, then layer in email 2 (objection handler) once you have analytics on which objection your audience actually has. Add email 3 if your review count is healthy. Add email 5 if your list churn from emails 1-4 is hurting deliverability.
A five-email sequence is the destination, not the starting line. AI gets you to "all five emails drafted" in an afternoon. The hard part — picking the right objections, finding the real reviews, deciding when to discount — still belongs to you.
If you want the prompts above in a copy-pasteable format, drop me a line and I'll send the full prompt sheet. Or open any of these in your own model of choice; the structure transfers.