AI Marketing
AI Prompts for Marketing: 20 Templates That Work
AI prompts for marketing that produce usable work: the structure behind a good prompt and 20 templates for research, content, ads, email and SEO.

AI prompts for marketing fail for one reason more than any other: the model knows nothing about your product, your audience or what "good" looks like, and the prompt doesn't tell it. The templates below fix that by building context into every request. Copy them, fill the brackets and attach your own material.
First, the structure that makes any prompt work. Then 20 templates grouped by job.
Why most AI prompts for marketing fail
"Write a LinkedIn post about our new feature" produces the same post every other company gets. The model fills the gaps with averages.
Anthropic's prompt engineering overview recommends defining success criteria and a way to test against them before you even write the prompt. For marketing, that means deciding up front: who is this for, what should they do next, and how will you judge the draft?
The structure of a prompt that works
Anthropic's prompting best practices boil down to a few habits: be clear and direct, add context and the reason behind instructions, use examples, structure long prompts with tags and give the model a role. For marketing, I use five blocks:
CONTEXT: product, audience, where this will appear
TASK: one clear job
CONSTRAINTS: length, tone, claims to avoid, words we never use
EXAMPLE: a past piece that performed well
FORMAT: exactly what to returnEvery template below follows this pattern. Paste your brand guide and a few past winners into a Claude Project once, and you can skip the context block after that.
Research and strategy prompts
1. Audience pain map
Here are 30 customer reviews and 5 sales call notes: [paste].
List the 7 most repeated problems, in the customers' own words,
with 2 quotes each. Then rank them by how often they appear.2. Competitor positioning
Here are the homepages of 3 competitors: [paste text].
For each, write the promise, the proof they offer and who they target.
Then list 3 positions none of them claim that we could own.3. Offer angles. Ask for ten angles for [offer], each with the audience segment, the pain it addresses and one headline, ranked by how different they are from competitors.
4. Objection list. Give the model your sales notes and ask for the 5 objections prospects raise before buying, with a one-paragraph answer to each backed only by facts you provided.
Content and copywriting prompts
5. Brief from a keyword. Provide the keyword, audience and the top-ranking page headings, and ask for a brief with the search intent, headings, questions to answer and what the existing pages miss.
6. Hook variations
Topic: [topic]. Audience: [audience].
Write 10 opening lines for a LinkedIn post. Each must use a different
approach: a number, a mistake, a question, a contrarian claim, a story.
No emojis. No "In today's world". Max 20 words each.7. Repurpose a long piece. Paste an article or transcript and ask for one newsletter section, three LinkedIn posts and five short social captions, each standing alone and pointing back to the original.
8. Edit for voice. Paste your voice guide and a draft, and ask for a revision that keeps every fact, removes filler and flags any sentence that makes a claim you didn't supply.
9. Case study draft. Provide the customer, problem, what you did and the measured result, and ask for a 300-word case study in problem, approach, result order with no invented numbers.
Ads and email prompts
10. Ad variations. Ask for 12 ad headlines and 6 primary texts for [offer], grouped into three angles, within the platform's character limits, with a note on which audience each angle suits.
11. Landing page critique. Paste the page copy and ask for the five biggest reasons a skeptical visitor would leave, ranked by likely impact, with a rewrite for the top one.
12. Welcome sequence. Ask for a five-email welcome sequence with the goal of each email, subject line options and body copy under 150 words, using only the product facts you paste.
13. Subject line test. Paste an email and ask for ten subject lines across curiosity, benefit and urgency, each under 45 characters, plus a preview line.
14. Re-engagement email. Describe the inactive segment and ask for a short, plain email that asks one question and offers one reason to come back.
SEO prompts
15. Search intent check. Paste the top ten result titles for a keyword and ask what the searcher wants, what format wins and whether a blog post can realistically rank.
16. FAQ from real questions. Paste People Also Ask questions and forum threads, and ask for concise answers of 40 to 60 words each, flagging any that need a source.
17. Title and meta description. Ask for five title tags under 60 characters and five meta descriptions between 120 and 160 characters, each including the keyword naturally.
18. Internal link suggestions. Paste a draft and a list of your existing URLs with titles, and ask for the five most relevant links with the exact anchor text and the sentence where each belongs.
19. Content refresh. Paste an old article and current top-ranking headings, and ask what to add, cut and update, in priority order.
20. On-page review. Paste a draft and ask for a checklist review: keyword in the opening, question-style headings, unsupported claims, missing sources and paragraphs over four sentences.
Turn prompts into a team library
The prompts that work should stop living in one person's chat history:
- Store each prompt with the task, required inputs and a sample output.
- Put the best ones into project instructions or Claude skills, so the process runs the same way for everyone.
- Review the library monthly and delete what stopped working.
Before a prompt goes into the library, test it. Run it on three different real inputs, not the one you wrote it for. If two of the three outputs need heavy editing, the prompt is missing context or constraints. Fix it, run the three again, and only then share it with the team.
For marketing-specific skills, see best Claude skills for marketing, and to connect prompts into workflows, read AI marketing automation. If you want your team building this library on real campaigns, it's part of my AI marketing training. HubSpot Academy's lesson on AI prompting is a good free companion.
If you want a prompt library built around your real briefs and your team trained to extend it, that is part of my AI marketing agency work.
Frequently asked questions
What makes a good AI prompt for marketing?
A good marketing prompt includes context about the product and audience, one clear task, constraints such as length, tone and claims to avoid, an example of good output and the exact format you want back. Missing context is the most common reason for generic results.
Do these prompts work in both ChatGPT and Claude?
Yes. The structure is model-agnostic. Claude tends to follow long constraint lists closely, and both models improve when you add examples and your own reference material.
How do I stop AI marketing copy sounding generic?
Give the model your real inputs: customer quotes, reviews, past high-performing copy and a list of words and claims you never use. Then ask for several distinct angles instead of one draft, and edit the best one.
Should my team share a prompt library?
Yes. Store tested prompts with the task they solve, the inputs they need and an example output. Review them monthly, retire the ones that stopped working and turn the best into reusable skills or project instructions.
Sources
Christian Monge
Builds AI second brains with Obsidian and Claude Code, co-founded the competitive-intelligence SaaS Pengu Insights, and trains teams in English and Spanish from Madrid.
More about meTeam Training
Bring this to your team.
Hands-on Claude training in English or Spanish, built around the tasks your people repeat every week.
Plan a team workshopKeep reading

AI Marketing
AI CRM: When to Add AI and When to Switch
What an AI CRM actually does, which features earn their price, and how to decide between adding AI to the CRM you have and migrating to a new one.

AI Marketing
AI Email Marketing Without Burning Your List
Where AI email marketing helps: segmentation, send timing, subject testing and first drafts, plus the volume trap that quietly kills deliverability.

AI Marketing
AI for Ecommerce: The Five Jobs That Pay Back
Where AI for ecommerce earns money: product data, search, support triage, merchandising and creative volume, plus the failures that cost trust and refunds.