# ChatGPT prompts for email marketing: 34 prompts for a small online shop

> Good ChatGPT prompts for email marketing state five things: context about your shop, the audience, the goal of the email, the constraints (length, offer, words to avoid) and the format you want back. The 34 prompts below follow that pattern for planning, subject lines, body copy, segments, rewriting, translation and results, and work the same way in Claude.

Canonical: https://tellnod.com/guides/chatgpt-prompts-for-email-marketing  
Last updated: 2026-10-22  
Author: Tellnod team

## A prompt structure that works

Most weak output comes from a prompt that leaves the model guessing. Anthropic’s advice is to be specific about the output format and constraints, and to explain why an instruction matters [S1]; OpenAI’s guide likewise recommends supplying relevant context, examples and the format you want [S2]. For a marketing email, that comes down to five parts.

| Part | What to include | Example |
| --- | --- | --- |
| Context | Who you are, what you sell, your voice | Small apparel brand: knitwear, outerwear, basics. Plain, warm voice |
| Audience | Who receives it and what they already know | Customers who bought a coat last winter |
| Goal | The one action you want | Click through to the new knitwear page |
| Constraints | Length, offer terms, words to avoid, facts to use | Under 120 words, no discount, no exclamation marks |
| Format | What the answer should look like | Subject, preheader, body, call to action, as separate lines |

Put together, a complete prompt reads like this: “You write email for Ninefold, a small apparel brand selling knitwear, outerwear and basics. Our voice is plain and warm; we never use hype. Write an email to customers who bought a coat from us last winter, introducing the new merino knitwear. Goal: a click to the knitwear page. Under 120 words, no discount, no exclamation marks. Return the subject line, preheader, body and call to action as separate lines.”

Save the context part once and paste it at the top of each chat. The prompts below assume you have done that, so they stay short. Replace the details in each one with your own.

## Prompts for planning campaigns

1. “Plan a three-month email calendar for a small apparel shop from November to January. One campaign a week, no more than two sales. For each week give the theme, the audience and the product to feature.”
2. “We launch a new outerwear line in four weeks. Suggest a sequence of three emails before launch day and one after, with the purpose of each.”
3. “Here is our product list with stock levels. Which three products should we feature in email this month, and why?”
4. “List ten email ideas for a knitwear brand that do not involve a discount, such as care guides, how a sweater is made, or styling one piece three ways.”
5. “We send one email a week to 1,800 subscribers. Suggest a simple structure for a weekly newsletter we can repeat without writing from scratch.”

For more angles, see [email campaign ideas for online stores](https://tellnod.com/guides/email-campaign-ideas-for-online-stores) and the [holiday email marketing calendar](https://tellnod.com/guides/holiday-email-marketing-calendar).

## Prompts for subject lines and preheaders

1. “Write ten subject lines for an email about our new merino crewnecks. Under 45 characters, no emoji, no capital letters for emphasis. Group them by approach: product detail, question, plain news.”
2. “For each of these five subject lines, write a [preheader](https://tellnod.com/glossary#preheader) of 40–80 characters that adds information instead of repeating the subject.”
3. “Rewrite this subject line so it says exactly what is inside the email, without implying a deadline we don’t have: [paste subject].”
4. “Give me two subject lines for an A/B test that differ in one thing only, and tell me what each version tests.”
5. “Which of these subject lines could read as misleading or spammy, and why? [paste list]”

> Keep subject lines honest. CAN-SPAM prohibits deceptive subject lines [S3], and Google asks senders not to start a subject with “Re:” or “Fwd:” unless the message is a real reply or forward [S4]. You can test a line with the [subject line checker](https://tellnod.com/tools/subject-line-checker).

## Prompts for body copy

1. “Write a new-arrival email for our Harbor wool coat. Use these facts only: [paste fabric, fit, price, sizes]. One short paragraph, one call to action.”
2. “Write a back-in-stock email for the oatmeal crewneck, sizes S–L. Do not say it is limited or sold out fast unless I tell you so.”
3. “Write a thank-you email for customers after their first order. No discount; include how to wash merino and a link to our returns page.”
4. “Draft a win-back email for customers who haven’t ordered in nine months. Acknowledge the gap without guilt, show three new pieces, and offer free shipping until Sunday.”
5. “Write a short email announcing our winter sale: 20% off outerwear from Friday to Monday, code WINTER20. Lead with the dates, then the products.”
6. “Turn these three product descriptions into an email that shows one outfit built from a coat, a knit and a basic tee. Under 150 words.”

> Models can produce details that sound right but are not true [S5]. Give the facts in the prompt, ask the model to use only those, and check prices, dates and codes before sending. Templates for these emails: [new arrival](https://tellnod.com/templates/new-arrival), [win-back](https://tellnod.com/templates/win-back).

## Prompts for segments

1. “Our customer export has these columns: [paste headers]. Suggest five [segments](https://tellnod.com/glossary#segment) worth emailing separately, and the rule that defines each.”
2. “How should the email to people who bought outerwear twice differ from the one to people who have only bought basics?”
3. “Define a VIP segment for a shop with 2,000 customers using order count and total spend. Suggest thresholds and explain the trade-off.”
4. “Which customers should we leave out of a sale email, and why? Consider recent buyers, people who returned items, and anyone who has not opened in a year.”

A chat assistant can only suggest segments unless it can see your list. Every segment should contain only people who agreed to receive marketing email; see [email segmentation for e-commerce](https://tellnod.com/guides/email-segmentation-for-ecommerce).

## Prompts for rewriting and translation

### Rewriting

1. “Cut this email to half its length without losing the product details or the call to action: [paste].”
2. “Rewrite this so it sounds like our last two emails, pasted below. Keep the facts, change the voice.”
3. “Remove every phrase from this draft that could appear in any other shop’s email, and replace it with something specific to the product.”
4. “Rewrite the opening so the news comes in the opening sentence.”
5. “Give me three calls to action for this email that name the product, instead of ‘Shop now’.”

### Translation

1. “Translate this email into Japanese for customers in Japan. Keep it polite but not stiff, convert sizes to Japanese sizing notes, and keep prices in yen as I give them: [paste].”
2. “Translate into Traditional Chinese for customers in Taiwan. Flag any phrase that does not carry over naturally rather than translating it word for word.”
3. “Adapt this email for UK customers: British spelling, prices in pounds as listed below, and change any US-only references.”
4. “Translate this email into Thai, then translate your version back into English so I can check the meaning matches the original.”

Ask a native speaker to read a translated email before it goes to a large list, and keep the legal footer accurate for each market.

## Prompts for analyzing results

1. “Here are the results of our last six campaigns: [paste sends, clicks, orders, revenue, unsubscribes]. Which performed best by [revenue per recipient](https://tellnod.com/glossary#revenue-per-recipient), and what did those emails have in common?”
2. “Compare these two campaigns. They went to lists of different sizes on different days; tell me which differences in [click-through rate](https://tellnod.com/glossary#click-through-rate) are likely to be meaningful and which are not.”
3. “Our unsubscribe rate doubled on the last two sends. List possible causes and how to check each one.”
4. “Based on these results, suggest one A/B test for next week, what to change, and how many recipients each version needs before I read anything into it.”
5. “Summarize this month’s email results in five lines I can share with my business partner.”

Treat open rates with caution: Apple’s Mail Privacy Protection downloads remote content in the background when a message arrives, which hides when a person actually opens it [S6]. Clicks, orders and [conversion rate](https://tellnod.com/glossary#conversion-rate) are more reliable. More on this in [email marketing metrics](https://tellnod.com/guides/email-marketing-metrics).

## Which prompts become actions with Tellnod connected

On its own, ChatGPT or Claude returns text. With Tellnod added as an [MCP connector](https://tellnod.com/glossary#mcp-connector), some of the same requests act on your actual list and email, and the result appears as a card in the chat. The send itself always needs your click.

| You type | Without a connector | With Tellnod connected |
| --- | --- | --- |
| “Customers who bought outerwear twice” | A suggested rule | A saved segment with a recipient count, unsubscribed contacts removed |
| “Draft a new-arrival email for the Harbor coat” | Text to copy elsewhere | A draft in your brand’s logo, colors and fonts |
| “Show me how it looks on mobile” | Not possible | A preview card with the rendered email on desktop and mobile |
| “Send me a test” | Not possible | A test email to your own address |
| “Schedule it for Thursday at 9 am” | Not possible | A prepared send and a confirmation card with recipients, exclusions and time; you press Confirm |
| “How did last week’s email do?” | Needs numbers pasted in | A results card with delivered, opens, clicks, unsubscribes and bounces |

The assistant can prepare a send but cannot confirm it: the Confirm button lives on the [approval card](https://tellnod.com/glossary#approval-card), and a scheduled campaign can be cancelled until it goes out. In ChatGPT, mention @Tellnod in the prompt; setup takes about a minute on the [connect page](https://tellnod.com/connect), with full steps for [ChatGPT](https://tellnod.com/guides/send-marketing-email-from-chatgpt) and [Claude](https://tellnod.com/guides/send-email-campaign-from-claude). Tellnod is free up to 500 contacts and 2,000 emails a month; see [pricing](https://tellnod.com/pricing).

## FAQ

### How do I write a good ChatGPT prompt for a marketing email?

Include context about your shop, the audience, the goal, the constraints and the format you want back. Paste real facts such as prices and dates, and add one or two past emails as examples of your voice.

### Do these prompts work in Claude?

Yes. The structure is the same for Claude and ChatGPT: both respond better to specific context, constraints and examples [S1] [S2].

### Can ChatGPT send marketing emails?

Not by itself. ChatGPT writes the text; sending needs an email tool. With a connector such as Tellnod, ChatGPT can draft, preview and prepare a campaign, and you approve the send on a confirmation card.

### Why does ChatGPT write generic marketing emails?

Usually because the prompt is generic. Without details about your products, customers and voice, the model falls back on common phrasing. Add one concrete product fact, a list of words to avoid and a past email as an example [S1].

## Sources

- [S1] Claude Docs: Prompting best practices: <https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices> (checked 2026-10-22)
- [S2] OpenAI API docs: Prompt engineering: <https://developers.openai.com/api/docs/guides/prompt-engineering> (checked 2026-10-22)
- [S3] FTC: CAN-SPAM Act: A compliance guide for business: <https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business> (checked 2026-10-22)
- [S4] Google Workspace Admin Help: Email sender guidelines: <https://support.google.com/a/answer/81126> (checked 2026-10-22)
- [S5] Claude Docs: Reduce hallucinations: <https://platform.claude.com/docs/en/test-and-evaluate/strengthen-guardrails/reduce-hallucinations> (checked 2026-10-22)
- [S6] Apple Support: Protect email privacy in Mail on Mac: <https://support.apple.com/guide/mail/protect-email-privacy-mlhlp1205/mac> (checked 2026-10-22)