Why this is hard to get right
A Real-World Scenario: When "Write a Text" Isn't Enough
Maria runs lifecycle marketing for a mid-size DTC skincare brand. She manages a team of two and owns cart recovery across email, push, and SMS. When her ESP added native SMS, her director asked her to launch an abandoned cart flow within two weeks.
Maria knew the basics. She had open rate benchmarks, a working sense of timing, and a 10% discount offer ready to go. What she didn't have was a copywriter — her contractor was booked, and her budget was locked for the quarter.
She turned to an AI assistant and typed: "Write abandoned cart text messages for skincare." The output came back in seconds. Three messages, loosely worded, no character counts, no timing logic, no opt-out language, and a tone that felt generic at best. The phrases like "Don't miss out!" and "Your cart is waiting!" could have come from any brand on the internet.
The real problem wasn't the AI — it was the prompt. Without knowing her audience (first-time buyers in the 25–40 range), her offer mechanics (10% off, code GLOW10, 24-hour window), her character constraints (150 characters per SMS), and her compliance requirements (TCPA opt-out in the final message), the model had no foundation to work from. It produced the statistical average of every abandoned cart SMS it had ever seen.
Maria's second attempt looked different. She spent eight minutes writing out her context: the role she needed the AI to play, her audience's buying stage, the exact offer details, the character limit, the tone she wanted (friendly and confident, never guilt-driven), and the need for two copy variations per message for A/B testing. She also specified send timing — 1 hour, 12 hours, and 23 hours after abandonment.
The output was night-and-day different. The messages fit within 150 characters, the first-message variation that opened with the product name outperformed the urgency-led version by 18% in click-through, and her compliance team approved the sequence in one pass because the opt-out language was already in place.
She had one more insight: the prompt she wrote was reusable. When the brand launched a new SPF line three months later, she updated the product name, the offer code, and the expiry window. Total revision time: under five minutes.
That's what a well-structured prompt actually delivers — not just a better first draft, but a repeatable system. The difference between "write a text" and a complete prompt brief is the difference between random output and a production-ready asset.
Common mistakes to avoid
Ignoring the 160-Character SMS Limit
Most AI models don't auto-enforce SMS character limits unless you state them explicitly. Without specifying 160 characters maximum (or 150 for safety), the model will write messages that split into two segments, doubling carrier costs and killing readability. Always include a hard character cap in your prompt.
Omitting the Offer Mechanics
Telling the AI to 'include a discount' without specifying the amount, the code, and the expiry window produces vague copy like 'save today' — which creates no real urgency. Include the exact percentage, the promo code, and the hours remaining so the AI can write copy that actually converts.
Skipping Compliance Requirements
TCPA and carrier guidelines require opt-out language in every new SMS thread. If you don't specify where and how to include it, the AI will skip it entirely. State which message in the sequence must carry opt-out language and what form it should take, such as 'Reply STOP to opt out.'
Leaving Tone Undefined
Without tone guidance, AI defaults to generic urgency copy. For premium or wellness brands, that tone actively damages trust and signals desperation. Define what you want (friendly, confident) and what you don't want (guilt-driven, pushy) so the model has clear creative guardrails.
Asking for One Message Instead of a Sequence
A single SMS rarely recovers a cart. The value is in the timed sequence — an immediate reminder, a mid-window nudge, and a final-hour close. Prompting for one message means you get a single tactic when you need a strategy. Always specify the number of messages and the send-time logic.
Not Requesting Copy Variations
A prompt that asks for one version per message gives you nothing to test. Specify two or three copy variations per message so you can run A/B tests from day one. Without variations, you're optimizing blind and you'll have to re-prompt every time you want to run a new test.
The transformation
Write some abandoned cart text messages that get people to buy.
You’re a lifecycle marketing copywriter for a DTC skincare brand. Create a **3-message abandoned cart SMS sequence** for shoppers who added items but didn’t check out. 1. **Audience:** US customers, 25–40, first-time buyers 2. **Offer:** 10% off, expires in **24 hours** (code: GLOW10) 3. **Tone:** friendly, confident, never guilt-driven 4. **Constraints:** **≤150 characters** per SMS, include a clear link placeholder: [CheckoutLink] 5. **Compliance:** include “Reply STOP to opt out” in message 3 Output: label each message with **send time** (1h, 12h, 23h) and include 2 copy variations per message.
Why this works
Role Primes the Model
The After Prompt opens with 'You're a lifecycle marketing copywriter for a DTC skincare brand.' This single line shifts the model's frame of reference. It stops generating generic retail copy and starts writing from the perspective of someone who understands retention strategy, tone calibration, and ecommerce conversion pressure.
Numbered Structure Removes Ambiguity
The After Prompt uses a numbered list — Audience, Offer, Tone, Constraints, Compliance — to separate each decision the model must make. Without this structure, the AI interpolates between competing signals and produces averaged output. Discrete numbered items force the model to address each variable independently.
Hard Constraints Produce Usable Copy
By specifying '150 characters per SMS' and a required link placeholder '[CheckoutLink],' the prompt turns abstract requirements into technical constraints the model can verify. The result is copy that fits within carrier limits and slots directly into your sending platform without manual editing.
Offer Details Create Real Urgency
The After Prompt includes '10% off, expires in 24 hours (code: GLOW10)' rather than a vague instruction to 'add urgency.' Real discount amounts, specific promo codes, and hard deadlines give the model the raw material to write copy that converts — not copy that implies conversion without delivering it.
Requesting Variations Enables Testing
The instruction 'include 2 copy variations per message' combined with send time labels (1h, 12h, 23h) means the output is structured for an A/B testing workflow out of the box. You get a deployment-ready sequence with built-in optionality, not a single draft that requires another round of prompting to expand.
The framework behind the prompt
The Strategy Behind Abandoned Cart SMS Recovery
Cart abandonment is one of the most studied problems in ecommerce. The Baymard Institute consistently reports abandonment rates above 70% across industries, making recovery a high-leverage channel for any brand with meaningful traffic.
Why SMS outperforms email for cart recovery in many segments: SMS open rates average above 90%, compared to 20–30% for email, and most messages are read within three minutes of delivery. For time-sensitive offers — like a 24-hour discount — that read speed matters enormously. The catch is that SMS is also more intrusive, which means tone errors are amplified and opt-out rates spike quickly if copy feels pushy.
The sequencing logic comes from behavioral economics. The first message capitalizes on the peak-end rule: shoppers are most receptive immediately after the experience of browsing. The second message, sent around 12 hours later, targets decision-stage friction — the moment when a shopper is likely comparing alternatives or waiting for payday. The third message uses loss aversion, a principle from Kahneman and Tversky's prospect theory: people respond more strongly to losing something they already "have" (a discount, a cart) than to gaining an equivalent benefit.
The AIDA framework (Attention, Interest, Desire, Action) maps naturally onto a 3-message SMS sequence. Message 1 recaptures attention. Message 2 rebuilds desire through offer mechanics. Message 3 drives action through a specific, expiring call to action.
Compliance and consent aren't just legal requirements — they're trust signals. Research in direct response marketing consistently shows that transparent opt-out language reduces anxiety and can actually improve click-through rates in audiences that are privacy-conscious. Including opt-out phrasing doesn't just protect you legally; it signals that you respect the recipient's inbox.
Effective SMS copywriting also draws from direct response principles: specificity over generality, one clear action per message, and friction-matched messaging — copy that speaks directly to the reason someone didn't complete the purchase.
Prompt variations
You are a direct response copywriter for a premium home fitness brand with an average order value of $800.
Write a 3-message abandoned cart SMS sequence for customers who added a resistance training system to their cart but did not check out.
- Audience: Fitness-committed adults aged 30–50, homeowners, have browsed the site at least twice
- Offer: No discount — lead with value, social proof, and risk reversal (free 30-day returns, 5-year warranty)
- Tone: Confident, aspirational, never salesy or urgent in a scarcity-based way
- Constraints: Maximum 150 characters per message; include [CheckoutLink] placeholder
- Compliance: Include 'Reply STOP to opt out' in message 3
Label each message with send time (2h, 24h, 47h). Provide 2 copy variations per message. Do not use discount language or countdown framing.
You are a retention copywriter for a monthly wellness subscription box brand.
Create a 2-message abandoned cart SMS sequence targeting first-time subscribers who started checkout for the intro box but did not complete it.
- Audience: Women aged 28–45, new to the brand, likely comparing options
- Offer: First box ships free; cancel anytime messaging is key to reducing commitment anxiety
- Tone: Warm, encouraging, low-pressure — like a trusted friend recommending something they love
- Constraints: 160 characters maximum per message; include [CheckoutLink] and brand name in message 1
- Compliance: Include opt-out instruction in message 2
Label messages with send times (45 minutes, 20 hours). Write 2 copy variations for each message. Focus on reducing friction around commitment, not on urgency.
You are an ecommerce SMS copywriter for a fashion accessories brand running a 24-hour flash sale.
Write a 2-message abandoned cart SMS sequence for customers who added sale items but did not complete checkout during the flash sale window.
- Audience: Existing customers aged 22–35 who have purchased before; they know the brand
- Offer: 30% off storewide, ends in 6 hours from send time; no code needed, discount applies at checkout
- Tone: Energetic, direct, time-aware — these customers respond to urgency because the deal is real
- Constraints: 150 characters per message maximum; include [CheckoutLink]; skip brand name in message 2 to save characters
- Compliance: Include 'Reply STOP to opt out' in message 2
Label messages with send times (30 minutes after abandonment, 5 hours after abandonment). Provide 2 copy variations per message. Make the countdown feel specific, not manufactured.
You are a product marketing writer for a B2B SaaS company offering a 14-day free trial of a project management platform.
Write a 3-message SMS sequence for leads who started the trial signup form but did not complete it.
- Audience: Operations managers and team leads at companies with 20–200 employees; they evaluate tools for their team, not just themselves
- Offer: No discount — emphasize zero credit card required, 5-minute setup, and a free onboarding call
- Tone: Professional, helpful, peer-to-peer — not a sales pitch, more like a product expert checking in
- Constraints: 150 characters per message; include [SignupLink]; avoid exclamation marks
- Compliance: Include opt-out language in message 3
Label each message with send time (1h, 24h, 72h after form abandonment). Provide 2 copy variations per message. Lead with value and ease, not urgency.
When to use this prompt
Ecommerce Marketing Teams
Build cart recovery SMS flows for key categories and seasonal promos without rewriting from scratch.
Product Managers Owning Growth Experiments
Generate test-ready SMS variants with clear timing and constraints for fast A/B testing.
Customer Success Teams for Subscription Brands
Recover stalled starter-kit purchases while keeping the tone helpful and low pressure.
Sales Teams at High-Ticket DTC Brands
Create compliant, personal-feeling follow-ups for carts with large average order values.
Pro tips
- 1
Specify your top cart items and price range so you can match urgency to purchase intent.
- 2
Add your brand’s “never say” list so the copy avoids words that hurt trust.
- 3
Set a goal per message, like “remind,” “reduce friction,” and “close with incentive,” to keep the sequence focused.
- 4
Define the checkout friction, like shipping cost or shade selection, so the SMS can address it directly.
Most abandoned cart SMS prompts treat every shopper the same. If your data is richer than that, your prompt should be too.
Segment by cart value. A customer with $240 in their cart needs a different message than one with $35. In your prompt, specify the cart value tier you're writing for and tell the AI whether to mention the total, the savings, or the product name explicitly.
Reference the specific product. Generic recovery messages say 'your cart.' High-performing ones say 'your Vitamin C serum.' Add a line to your prompt like: 'Reference the hero product by name in message 1.' Even if you use a dynamic variable in your ESP, the AI can write the placeholder into the copy so your team knows where it goes.
Acknowledge browse behavior. If your platform tracks that a customer viewed the same item three times before adding it, that's a signal worth using. Prompt the AI to write copy for high-intent shoppers separately from impulse browsers. The tone and urgency level should differ significantly.
Timing variation by segment. First-time buyers may need more time to decide. Loyal customers may respond faster to a shorter window. Define the segment in your prompt and adjust the send-time logic accordingly — 1h/12h/23h for urgency-driven buyers, 2h/24h/48h for deliberate ones.
Compliance failures in SMS marketing can result in TCPA fines starting at $500 per message, per recipient. A strong prompt builds compliance in from the start rather than patching it at review.
Before you prompt, confirm you have:
- Written consent to send marketing SMS from every recipient on your list
- A clear record of when consent was obtained and for what purpose
- A functioning opt-out mechanism tied to the 'Reply STOP' instruction
In your prompt, always specify:
- Which message in the sequence carries the opt-out instruction
- The exact opt-out phrasing your legal team has approved
- Whether your brand name must appear in every message (some carriers require this for shortcodes)
- Any state-specific requirements — California, Florida, and New York have stricter rules than the federal baseline
After generating output, verify:
- Opt-out language is present and unambiguous
- No false urgency claims (e.g., 'only 2 left' without real inventory data)
- No prohibited language for your industry (financial services, healthcare, and alcohol have additional restrictions)
- Character counts are within limits to prevent unintended message splitting
If you send internationally, add the destination country to your prompt and ask the AI to flag any region-specific compliance considerations.
Generating a strong SMS sequence is only half the job. Getting it into your sending platform without manual rework is where most teams lose time.
Use standardized placeholders. The After Prompt on this page uses [CheckoutLink] as a placeholder. When you prompt the AI, use the exact variable syntax your ESP or SMS platform expects — for example, {{checkout_url}} in Klaviyo or %CHECKOUT_URL% in Attentive. This eliminates a copy-paste translation step.
Request labeled output. Ask the AI to label each message with its send time, its segment (if you're writing for multiple audiences), and its character count. Clean labeling makes it faster to map each message to the right automation step in your flow builder.
Structure for easy handoff. If another team member or contractor will load the messages into the platform, ask the AI to output a simple table: message number, send time, variation A, variation B, character count. A table is faster to QA than a block of prose.
Save the prompt as a template. Once your sequence performs well, save the full prompt in a shared doc. Next season's update takes five minutes — change the offer details and expiry, keep the structure, re-run. This is how a single well-built prompt pays off over multiple campaigns.
When not to use this prompt
When This Prompt Pattern Is Not Appropriate
This prompt structure works well for brands with strong SMS consent lists and defined offer mechanics. It's not the right tool in every situation.
Don't use this pattern if:
- You don't have confirmed opt-in consent for SMS marketing. Using this prompt to generate messages for unconsented recipients creates TCPA liability regardless of how good the copy is.
- Your product has a long consideration cycle (B2B enterprise software, luxury real estate, custom manufacturing). SMS cart recovery assumes short-window decision-making. For high-stakes purchases with multi-week sales cycles, use email nurture sequences or sales outreach instead.
- Your brand operates in a heavily regulated category without legal review first — healthcare, financial services, and alcohol all have channel-specific restrictions that require compliance sign-off before any template goes live.
- You have fewer than 500 subscribers on your SMS list. The investment in building and testing a multi-message sequence rarely pays off at small list sizes. A single well-crafted message may be more appropriate.
- You're recovering carts from an offline or in-store experience where customers didn't initiate a digital checkout. The prompt assumes a tracked digital abandonment event with a recoverable checkout URL.
In these cases, consider email-based recovery prompts, direct outreach scripts, or retargeting ad copy prompts as better-matched alternatives.
Troubleshooting
The AI writes messages that exceed 150 characters even after I specified the limit
Add a verification instruction at the end of your prompt: 'After writing each message, display the character count in parentheses and flag any message that exceeds 150 characters.' This forces the model to self-audit. Alternatively, ask for messages at 130 characters maximum to give yourself a buffer for dynamic variable expansion when your ESP inserts the actual checkout URL.
All three messages in the sequence feel identical in tone and structure
Assign a distinct goal to each message in your prompt. For example: 'Message 1 goal: remind with no pressure. Message 2 goal: introduce the offer with mild urgency. Message 3 goal: final push with opt-out included.' When each message has a different strategic job, the AI writes them differently — without that instruction, it defaults to repeating the same hook three times.
The copy sounds too salesy and pushes against the brand tone we defined
Add a 'never use' list to the Tone section of your prompt. For example: 'Never use the phrases: don't miss out, limited time only, hurry, or act now.' Negative constraints often work faster than positive tone descriptions because they eliminate specific patterns the model defaults to. Also check that your tone and urgency instructions don't contradict each other — asking for 'high urgency but no pressure' creates ambiguity the model resolves unpredictably.
The two copy variations per message are nearly identical
Specify what dimension each variation should test. For example: 'Variation A: lead with the product name. Variation B: lead with the discount amount.' Without a defined test axis, the model writes superficial rewrites rather than genuinely different approaches. Name the strategic difference you want to test, and the AI will write toward that variable.
The AI omits the opt-out language or places it incorrectly
Move your compliance instruction out of a numbered list and into a standalone line with bold emphasis at the bottom of your prompt: 'Compliance requirement: Message 3 must end with the exact phrase: Reply STOP to opt out. This is non-negotiable — do not omit it.' Embedding compliance in a list item often causes the model to deprioritize it when balancing multiple constraints.
How to measure success
How to Evaluate Your AI-Generated SMS Output
Strong output from this prompt type should clear a specific quality bar before it reaches your sending platform. Evaluate each draft against these signals:
Structural checks:
- Every message is at or under your stated character limit (count manually or with a character counter tool)
- [CheckoutLink] or your platform's variable syntax appears exactly once per message
- Opt-out language appears in the final message with no variations from your approved phrasing
- Send time labels are present and match your automation logic
Copy quality checks:
- Message 1 reads as a reminder, not a sales pitch — it should feel like a helpful nudge
- Message 2 introduces the offer with enough specifics (amount, code, expiry) to create genuine urgency
- Message 3 closes with clarity: one action, one deadline, one opt-out
- The two variations per message differ on a testable dimension, not just word order
Brand fit checks:
- No phrases from your "never use" list appear
- The tone matches your defined parameters across all three messages
- The copy would not feel out of place next to your existing brand communications
Before sending: run output past your legal or compliance team if you haven't done so for this sequence structure before.
Now try it on something of your own
Reading about the framework is one thing. Watching it sharpen your own prompt is another — takes 90 seconds, no signup.
Build a compliant, conversion-ready abandoned cart SMS sequence tailored to your brand, audience, and offer in minutes.
Try one of these
Frequently asked questions
Most high-performing sequences use 3 messages: an immediate reminder (within 1 hour), a mid-window nudge (around 12 hours), and a final-hour close tied to offer expiry. Two messages can work for flash sales with short windows. Going beyond 3 messages typically hurts opt-out rates without a meaningful lift in recovery. Match sequence length to your offer window, not a default template.
No. Without explicit instructions, most models write at comfortable reading length — which is often 200 to 300 characters. You must specify the character limit in your prompt (150 characters is a safe cap that avoids multi-part message splitting). If you want the AI to count characters, ask it to display the character count next to each draft.
Replace the audience age range, product context, and offer mechanics with your own data. More importantly, add the specific friction point your customers face — for example, 'customers abandon because of shipping cost' or 'shade selection is confusing.' Naming the friction lets the AI address it directly, which is far more effective than generic recovery language.
At minimum, TCPA requires opt-out language in any new SMS thread. Specify which message in the sequence must carry the opt-out instruction and the exact phrasing, such as 'Reply STOP to opt out.' If you operate in California, note CCPA context. If you send to EU numbers, flag GDPR consent requirements. The AI will not add these automatically unless you instruct it to.
Generic output usually means one of three things: the role definition is too broad, the tone constraints contradict each other, or the model is averaging your instructions with prior training patterns. Try adding a 'never use' list to your prompt — for example, 'Do not use the phrases: don't miss out, your cart is waiting, or limited time.' Negative constraints often sharpen output faster than positive instructions.
The structure transfers well, but two things change: remove the character limit and replace the compliance field with email-specific CAN-SPAM requirements (physical address, unsubscribe link). Email also supports subject lines and preheader text, so add those to your output instructions. The role, audience, offer, and tone sections carry over directly.
Ask for variations that differ on a single variable, not random rewrites. For example: 'Write variation A that leads with the discount and variation B that leads with the product name.' This gives you a clean test hypothesis. If you ask for 'two different versions,' the model may vary multiple elements at once, making it impossible to know what drove the performance difference.
No. For premium brands, discounting too early trains customers to abandon carts intentionally to wait for an offer. Consider leading your sequence with value and social proof, then introducing a discount only in the final message — or not at all. Your prompt should reflect this strategy by assigning a different goal to each message rather than defaulting to discount-first.