Writing SMS Copy with AI Prompts That Actually Work

If you have tried using AI to draft text message copy, you have probably seen the gap between a clever-sounding draft and a message you would actually want to send to a customer’s phone. Many small teams now look at chatgpt prompt packs as a faster starting point, but a prompt is only useful if it respects the realities of SMS: short messages, personal delivery, strict opt-in rules, and very little room for fluff.

Why most AI-written texts fail on the phone

Language models are trained on long-form writing. Left to their own defaults, they produce greetings, multiple exclamation points, a sign-off, and a link buried in the third sentence. On a website or in an email, that structure might be fine. In a text message, it reads like spam and gets ignored or reported.

The fix is not a better general prompt. It is a prompt that encodes the constraints of messaging from the start. A working SMS prompt tells the model who the recipient is, what the message must accomplish, how long it can be, and what it must never do.

The constraints every SMS prompt should include

Before writing any prompt, list the rules your messages must follow. For most businesses these include:

• Length: A single standard SMS segment holds 160 characters of GSM-7 text. Anything with emoji or certain special characters switches to UCS-2 encoding, which drops the limit to 70 characters per segment and can split your message into several billable parts.
• Identification: The sender name or business name should appear early, so the recipient knows who is writing before they open the message.
• Opt-out language: Promotional messages need a clear way to stop receiving them, such as a reply keyword like STOP. Your prompt should require this wording every time, not as an afterthought.
• Call to action: One action per message. A reply, a short link, or a booking request. Not three.
• Timing: Messages should respect local quiet hours and the recipient’s time zone. The model cannot know this unless you tell it, and your sending platform should enforce it regardless.
• Tone: Plain, direct, and written as though a person is sending it, not a brand announcement.

A prompt structure that holds up

Reusable prompts work best when they follow a consistent layout. Here is a structure that has proven useful for text campaign drafting:

1. Role and context: Tell the model it is writing a transactional or promotional SMS for a named business type, and describe the audience in one sentence.
2. Goal: State the single outcome you want, such as confirming an appointment, announcing a restock, or recovering an abandoned order.
3. Hard limits: Set the maximum character count, require opt-out wording for promotional messages, and forbid emoji unless you have chosen to accept the shorter segment limit.
4. Inputs: Provide variables in brackets, such as first name, appointment time, or offer code, so the same prompt can be reused across campaigns.
5. Output format: Ask for three variants, each labeled with its character count. This makes comparison fast and helps you spot drafts that run long.

A prompt built this way produces something you can check against a list of rules in seconds. A vague prompt produces something you have to rewrite from scratch.

Example of a constrained prompt

Consider a clinic that needs appointment reminders. A workable prompt might read: write three SMS reminders for a dental clinic’s patients, each under 160 characters, plain text without emoji, including the clinic name in the first five words, containing the appointment date and time in brackets as variables, and ending with the reply instruction to confirm or reschedule. Label each variant with its character count. The output will be far more usable than a request for a friendly reminder.

Testing AI-generated SMS before it goes live

No prompt should be trusted without review. A simple testing routine catches most problems:

• Paste each draft into a character counter and confirm it stays within a single segment if that matters for your budget.
• Read it aloud. If it sounds like a press release, rewrite the prompt rather than editing the output by hand, so the fix carries forward.
• Send test messages to internal phones at different carriers and check how links and special characters render.
• Confirm that the opt-out instruction is present and that your platform processes the keyword correctly.
• Check that the message makes sense to someone who has no context about the campaign.

Keep a log of which prompt versions produced copy that performed well and which produced complaints or confusion. Over time, this log becomes more valuable than any single prompt.

Building a personal prompt library

Teams that use AI for messaging tend to accumulate dozens of prompts with no naming system. A better approach is to organize them by use case: onboarding, appointment reminders, payment nudges, re-engagement, and seasonal promotions. Store the constraints block once and reuse it across prompts so a rule change only needs to be made in one place.

Versioning matters too. When you revise a prompt, note the date and the reason. If a new version produces shorter, clearer copy, you want to know exactly what changed.

Where to find starting points

You do not have to build everything from zero. If you want to see how other people structure their drafting prompts before writing your own, browsing a set of tested AI prompt collections can show you useful patterns for role-setting, constraints, and output formatting. Treat any borrowed prompt as a draft. Adapt the limits, opt-out language, and tone to your own sending rules and audience, and test it the same way you would test anything you wrote yourself.

A quick checklist before you send

• Message fits within your chosen character limit
• Business name appears early
• Opt-out wording is included for promotional sends
• Exactly one call to action
• No emoji unless you have accepted the shorter segment size
• Timing respects recipient time zone and quiet hours
• Copy reads naturally when spoken aloud
• Test message delivered and checked on real devices

Final thoughts

AI can save real time on text message drafting, but only when the prompt carries the discipline that SMS demands. Short, specific, constraint-heavy prompts produce copy that respects the recipient’s phone and their attention. Start with the rules, build a reusable structure, test every output, and keep a record of what works. The result is messaging that feels like it came from a real person who knows what they are saying, which is exactly what text message marketing needs to succeed.