Writing a text message that fits in a single segment, sounds like a person wrote it, and still gets a reply is harder than it looks. Many SMS marketers now draft copy with an AI model, and the quality of the output depends almost entirely on the instructions behind it. If you don’t want to start from scratch, you can look for premium prompt templates that are already structured for messaging work, then adapt them to your own brand, audience, and carrier rules.
Why generic prompts produce weak texts
Ask a chatbot to write a promotional text and you will usually get something long, cheerful, and full of exclamation points. That copy might look fine on a screen, but it often runs past the character limit, buries the offer, and includes phrases that sound like spam. The problem is not the model. The problem is that the prompt never told it the constraints it needed to respect.
A useful SMS prompt treats the message as a small, tightly controlled object. It names the channel, the audience, the single goal, the offer, the required disclosures, and the length. Once those pieces are fixed, the model has far less room to wander.
The anatomy of a working SMS prompt
Most effective prompts for text messaging share the same building blocks. You can reuse this list as a checklist every time you write one:
- Channel: SMS, not email or push notification, so the model avoids subject lines and HTML.
- Audience: who is receiving the message and what they already know about your business.
- Single goal: one action, such as confirm an appointment, reply YES, or book a slot.
- Offer or information: the exact detail that must appear, including dates, times, and amounts.
- Length limit: a target character count, not just a word count.
- Tone: two or three adjectives, plus words to avoid.
- Required elements: sender name, opt-out instructions, and any legal wording.
Give the model two or three examples of messages you have already sent and liked. Ask it to match their rhythm rather than invent a new voice. This single step often improves results more than any clever phrasing.
Respect the character budget
Character count is not a suggestion in SMS. Standard GSM-7 messages, which cover basic Latin letters and common punctuation, fit 160 characters in one segment. Longer messages are split, and each segment in a multipart message holds 153 characters because of the header data. If your text includes emoji or certain non-Latin characters, the message switches to UCS-2 encoding, which allows only 70 characters per single segment and 67 per segment when concatenated.
Put the limit directly in your prompt, for example: “Keep the full message under 150 characters including the opt-out line.” Then count the output yourself before you schedule anything. Many messaging platforms display a segment counter, and it is worth checking it on every template, especially after a model rewrites a line.
Prompt templates for common SMS jobs
Most small senders use the same handful of message types. Here are starting prompts you can adapt. Replace the bracketed fields with your details.
- Appointment reminder: “Write an SMS reminder for [business] confirming a [service] appointment on [date] at [time]. Include a reply option to confirm or reschedule. Under 140 characters. Friendly and direct. No emoji.”
- Delivery update: “Write an SMS telling a customer their order [number] has shipped and will arrive by [date]. Include the tracking link placeholder [LINK]. Plain, calm tone. Under 160 characters.”
- Abandoned cart or booking nudge: “Write one SMS reminding a customer they left [item] in their cart. Mention the price, avoid pressure words like urgent or last chance, and end with a clear link placeholder. Under 150 characters.”
- Re-engagement: “Write an SMS to subscribers who have not purchased in six months. Ask one question, offer a single incentive of [offer], and include opt-out wording. Under 155 characters.”
- Opt-in confirmation: “Write a confirmation text for a customer who just subscribed to [program]. State what they will receive, how often, the sender name, and how to reply STOP to opt out. Under 160 characters.”
Notice that each template states the number of messages, the tone, the required parts, and the length. That structure is what makes a prompt reusable across campaigns. To go deeper, explore The marketplace for AI prompts that actually work.
Build a compliance check into every prompt
Language models do not know your consent records or the rules that apply in your region. They will sometimes produce copy that implies urgency you cannot support, omits the business name, or forgets an opt-out instruction. Add a review step to your workflow that asks the same questions every time:
- Does the message identify who is sending it?
- Is there a clear way to opt out, where required?
- Does the message match the consent the recipient actually gave?
- Are any claims about price, availability, or deadlines accurate right now?
- Is it scheduled within the recipient’s local quiet hours?
You can also ask the model to critique its own draft against this list. That is useful as a first pass, but a person should still approve anything that goes to real subscribers.
Test before you send to everyone
Once a template produces copy you like, send it to an internal list first. Check how it renders on several phones, because line breaks, links, and emoji can display differently across devices. Confirm that links work and that the opt-out reply is processed by your platform.
When you have two strong variations, split a small segment of your audience between them and compare reply rates, click-throughs, or conversions. Treat the results as a signal for your specific list, not as a universal truth. A short question that works for a dental clinic may fall flat for a retail store.
Keep a living prompt library
The real value comes from treating prompts like any other working asset. Store each one with its purpose, the date it was last tested, the character count of its best output, and any notes about what changed. When a carrier rule, a legal requirement, or your own offer changes, you update the template once instead of rewriting dozens of messages from memory.
Over time you will notice patterns. Certain openers consistently get replies. Certain calls to action get ignored. Feed those findings back into your prompts so the next draft starts from a stronger baseline.
Final thoughts
AI can speed up SMS writing considerably, but speed is only useful when the messages are clear, compliant, and built around one goal. Start with tight prompts that specify the channel, length, audience, and required elements. Check every output against your character budget and your consent rules. Test with real segments, and keep a library of what works. Do that consistently, and your texts will read like they were written with care, because the instructions behind them were.
Leave a Reply