---
title: "AI Prompts for Marketing: 6 Templates You Can Paste Today"
description: "Six copy-ready AI prompts for marketing: strategy, positioning, email, ads, content briefs and campaign reviews. Works in ChatGPT, Claude and Gemini."
url: https://www.promptpal.app/blog/ai-prompts-for-marketing-templates
author: "PromptPal Team"
published: 2026-11-09T08:02:44.630Z
updated: 2026-11-18T22:22:30.309Z
category: "General"
tags: ["marketing", "prompt templates", "chatgpt", "claude", "prompt management"]
language: en
reading_time_minutes: 8
image: https://images.ctfassets.net/xzkuxcljsk12/5CTn19H8aSxrUqQq4Vt0r6/f1e90098b92ec8fe582fda3b182e06d3/image.png
publisher: PromptPal
---

# AI Prompts for Marketing: 6 Templates You Can Paste Today

Six copy-ready AI prompts for marketing: strategy, positioning, email, ads, content briefs and campaign reviews. Works in ChatGPT, Claude and Gemini.

The best AI prompts for marketing give the model four things: who you're selling to, what you're selling, what a good result looks like, and the facts it must not invent. Generic prompts like "write a marketing plan for my product" return generic plans. Below are six complete prompts you can paste into ChatGPT, Claude or Gemini, plus a way to keep them from rotting in a doc nobody opens.

## What every marketing prompt needs

Most weak marketing output traces back to missing inputs, not a weak model. If you don't say who the buyer is, the model writes for "businesses" and "professionals". If you don't give it your real numbers and claims, it makes some up.

Every prompt below follows the same skeleton:

- **Role and task.** One line. "You are a B2B email copywriter. Write a three-email sequence."
- **Audience.** A job title, a pain point, and what they've already tried.
- **Facts.** Product details, pricing you've approved, proof points. Anything not on this list is off limits.
- **Constraints.** Length, tone, channel, words to avoid.
- **Output format.** Numbered list, subject line plus body, a short rationale. Whatever you can use without reformatting.

The prompts work in all three major models. Claude tends to follow tight formatting and "don't invent" instructions closely; ChatGPT and Gemini usually do too, but check the first output for claims you didn't supply. If you move between models a lot, [this post on getting consistent results across Claude, ChatGPT and Gemini](https://promptpal.app/blog/consistent-ai-results-claude-chatgpt-gemini) covers what to adjust.

## 1. A marketing strategy prompt that asks questions first

Strategy is where a one-shot prompt does the most damage. The model will happily write a confident plan from three sentences of context. Make it interview you instead.

> `You are a senior marketing strategist helping a [company size] [business type] company.`
>
>
>
> `Goal: build a 90-day marketing plan for [product or service] aimed at [target customer, e.g. operations managers at logistics firms with 50-200 staff].`
>
>
>
> `What I already know:`
>
> `- Current channels: [channels in use]`
>
> `- Monthly budget: [amount you've approved]`
>
> `- Team: [who does the work]`
>
> `- Main constraint: [e.g. no designer, long sales cycle]`
>
>
>
> `Before writing the plan, ask me up to 5 questions you need answered to avoid guessing. Wait for my answers.`
>
>
>
> `After I answer, produce:`
>
> `1. The one audience and one channel to focus on first, with a two-sentence reason.`
>
> `2. A week-by-week plan for the first 30 days.`
>
> `3. Three metrics to check weekly and what result would make us change course.`
>
> `4. What we should deliberately not do this quarter.`
>
>
>
> `Do not invent benchmarks or industry statistics. If you're unsure, say so.`

The "what we should not do" line is the most useful one. Models default to recommending everything: blog, podcast, LinkedIn, webinars. Forcing a cut list gives you a plan a small team can actually run.

## 2. A positioning prompt built from customer words

Positioning copy improves fastest when you feed it what customers actually said. Paste in sales call notes, reviews or support tickets.

> `You are a positioning consultant.`
>
>
>
> `Below are raw comments from [number] customers about [product]. Read them and extract:`
>
>
>
> `1. The problem they describe, in their own words (quote up to 5 short phrases).`
>
> `2. The alternatives they mention using before us.`
>
> `3. The outcome they care about most.`
>
> `4. Three possible one-sentence positioning statements in this format: "For [audience] who [problem], [product] is the [category] that [key benefit], unlike [alternative]."`
>
>
>
> `Only use information in the comments. If the comments don't support a claim, don't make it.`
>
>
>
> `Customer comments:`
>
> `[paste comments]`

This is a good job for AI because the work is reading a lot of messy text and finding patterns. Still, you pick the statement. The model can't tell which one your sales team can say with a straight face.

## 3. Cold email sequence with a hard fact boundary

Cold emails go wrong when the model fills gaps with flattery and fake specifics ("I loved your recent post on..."). Close that door.

> `You are a B2B email copywriter. Write a 3-email outreach sequence.`
>
>
>
> `Recipient: [job title] at [type of company]`
>
> `Their likely problem: [specific problem]`
>
> `What we offer: [one sentence]`
>
> `Proof I can use: [one real result or customer, or write "none"]`
>
> `Call to action: [e.g. 15-minute call, reply with a yes or no]`
>
>
>
> `Rules:`
>
> `- Email 1 under 90 words, email 2 under 70, email 3 under 50.`
>
> `- Plain text. No exclamation marks, no "I hope this finds you well".`
>
> `- Do not reference anything about the recipient that I haven't given you. Use [first name] and [company] as placeholders.`
>
> `- Each email gets two subject line options under 6 words.`
>
> `- Email 3 should make it easy to say no.`

If you write this kind of email weekly, add one real example of an email that worked as a few-shot sample. It does more for tone than any adjective you can write. We cover that approach in [few-shot prompting](https://promptpal.app/blog/few-shot-prompting-examples-how-to).

## 4. Ad copy variations that differ on purpose

Ask for ten ad variations and you usually get one idea rewritten ten ways. Ask for different angles instead.

> `You are a paid social copywriter.`
>
>
>
> `Product: [product and one-line description]`
>
> `Audience: [audience]`
>
> `Platform: [e.g. LinkedIn, Meta, Google Search]`
>
> `Character limits: [headline limit], [body limit]`
>
> `Approved claims: [list]`
>
>
>
> `Write 6 ads, two each for these angles:`
>
> `- Pain: the cost of the current way of doing it`
>
> `- Outcome: what life looks like after`
>
> `- Proof: a specific result from the approved claims`
>
>
>
> `Label each ad with its angle. Stay inside the character limits. Use only the approved claims. After the ads, tell me which one you'd test first and why, in two sentences.`

Check the character counts yourself. Models miscount, and a headline that's three characters over gets truncated in the live ad.

## 5. A content brief for a writer or for the model

A good brief saves more time than a good draft prompt. This one works whether you hand the output to a freelancer or feed it back into the model.

> `You are an SEO content strategist.`
>
>
>
> `Target keyword: [keyword]`
>
> `Audience: [who searches this and what they want to do next]`
>
> `Product angle: [how our product legitimately relates]`
>
>
>
> `Produce a content brief with:`
>
> `1. Search intent in one sentence.`
>
> `2. A working title and a URL slug.`
>
> `3. The first two sentences of the article, which must answer the query directly.`
>
> `4. An outline with H2s that each say what the reader gets.`
>
> `5. Three things competing articles probably skip that we should include.`
>
> `6. Facts we need from a subject expert before writing (list as questions).`
>
>
>
> `Do not cite statistics or studies. If a claim needs a source, flag it.`

Item six matters. It turns "AI wrote something plausible" into "here's what a human needs to supply", which is where the quality comes from.

## 6. A campaign review prompt for after the numbers come in

Most people use AI to make marketing and never to review it. Give it your results and ask for a plain reading.

> `You are a marketing analyst.`
>
>
>
> `Campaign: [name and goal]`
>
> `Dates: [range]`
>
> `Results (paste raw numbers): [impressions, clicks, signups, spend, etc.]`
>
> `What we expected: [targets]`
>
> `What changed during the campaign: [new creative, budget shift, seasonality]`
>
>
>
> `Tell me:`
>
> `1. What the numbers say, in three bullet points, without speculation.`
>
> `2. Three possible explanations for the biggest gap versus target, ranked by how well the data supports each.`
>
> `3. One cheap test to separate the top two explanations.`
>
>
>
> `Do not calculate anything you can't derive from the numbers above. Show any arithmetic you do.`

"Show your arithmetic" is worth keeping. It lets you catch errors quickly instead of trusting a percentage that appeared from nowhere.

## Adapting these for digital marketing and different models

For channel-specific work, such as SEO, paid search or lifecycle email, change the audience and constraints lines and leave the structure alone. The skeleton transfers; the details are what you swap.

A few notes on models:

- **ChatGPT** is quick for variations and brainstorming. Ask it to flag assumptions explicitly, or they tend to slide by.
- **Claude** is a good fit for the long-input prompts, like positioning from a pile of customer comments. There's a set of general templates in [Prompts for Claude](https://promptpal.app/blog/prompts-for-claude-templates).
- **Gemini** works well if your source material lives in Google Docs or Sheets. More templates in [Gemini prompts for work](https://promptpal.app/blog/gemini-prompts-templates-work-images).

Model behaviour changes between versions, so treat any "model X is better at Y" claim, including these, as a starting point and test it on your own task.

## Keeping your marketing prompts from going stale

The real cost shows up in month two. Someone tweaks the cold email prompt, it gets better, and the improvement lives in one person's chat history. The rest of the team keeps using the old version.

A workable routine in PromptPal:

1. Paste each prompt above into your own library and replace the bracketed fields you never change (company name, approved claims, tone rules) with real values.
2. Leave the fields that change per use as `[placeholders]`.
3. When you improve a prompt, save it as a new version instead of overwriting, so you can go back if the new one performs worse. See [prompt version control](https://promptpal.app/prompt-version-control).
4. Put the approved ones in a shared workspace so the whole marketing team starts from the same text.

If you'd rather start from existing templates than from a blank page, the [public prompt library](https://promptpal.app/prompts/public) has prompts organised by role and task.

## Mistakes that cost the most time

- **Skipping the approved-claims list.** This is how invented stats end up in a draft. Always supply the facts, and always say "nothing else".
- **Asking for the final asset in one go.** For anything over a paragraph, ask for an outline or angles first, approve, then draft.
- **Rewriting adjectives instead of adding inputs.** "Make it punchier" rarely helps. A real example of copy you like helps a lot.
- **Never testing the prompt twice.** Run it three times on different inputs before you save it. A prompt that works once might be luck.

## Next step

Pick the one prompt above you'd use this week, fill in your real facts, and save it. [Browse the PromptPal prompt library](https://promptpal.app/features/prompt-library) to store it, version it and share it with your team.

---

Originally published at https://www.promptpal.app/blog/ai-prompts-for-marketing-templates
