Best 60 Claude Marketing Prompts

Sixty marketing prompts, ten stages, one repeating shape. AI Central's Best 60 Claude Marketing Prompts is a copy, paste and fill-in pack running from audience research through to client case studies, and almost every prompt is written as a specification rather than a request. Word counts are fixed, required elements are named, common habits are banned outright. That is why the output comes back usable, and why the pattern is worth stealing for your own prompts.

Ten stages, in the order a funnel actually gets built

AI Central groups the sixty into ten named blocks of six. The order is not alphabetical and it is not by channel. It is the order a marketing engine gets assembled in, starting with who you are talking to and ending with proof that you helped them.

• Brand Builder Blueprint, audience research, language, pillars and positioning.
• Offer and Sales Architect, the spine of a sales page.
• LinkedIn Content Creation, hooks and five post shapes.
• Ad Lab, paid variants, retargeting and a pre-roll script.
• Lead Magnet Creation, from ten ideas to the closing page.
• Email Sequence Engine, a four-email welcome run plus case study and objection emails.
• Webinar and Event Promotion, registration page through to the opening script.
• Video and VSL Scripts, talking head and video sales letter sections.
• DM Conversation Kit, six kinds of LinkedIn direct message.
• Testimonial and Case Study, the proof layer.

That sequencing is the first argument the pack makes. You cannot write the ads until the offer is defined, and you cannot define the offer until you know the exact words your audience uses. The second prompt in AI Central's Brand Builder Blueprint says so outright, because it opens by pointing back at the output of the first.

Using the audience analysis, build a 3-part language library for my ICP

The library it asks for has three parts, the vocabulary people use for their problems, the language they use for success, and the way they describe their own expertise. AI Central tells the model to keep those phrases raw and true to how people actually speak. That output then feeds the pillars, the positioning statement and every headline downstream.

The prompts are specs, not requests

Ask a model for a LinkedIn post and you get the average of every LinkedIn post it has ever seen. The prompts here close that gap by taking choices away.

Almost every one carries three things at once: a form constraint, a content constraint, and a list of banned moves. The hook prompt asks for five LinkedIn hooks of two lines each, tells the model what each line has to do, then ends flat.

No questions, exclamation marks, or em dashes.

Banning a habit works better than describing a tone. An instruction like professional but friendly gives a model nothing it can act on. No exclamation marks is checkable, by the model and by you. AI Central uses that lever repeatedly, and the bans are always specific to the asset.

• Vague adjectives and questions, in the sales page headlines.
• Generic founder insights, in the audience research.
• Vague claims, in the differentiator work and the video sales letter solution section.
• The words Welcome and Congratulations, in the first welcome email.
• Quotes, unless they are exact, in the sales page social proof section.

That last one is the sharpest instruction in the set. A model asked to write a social proof section will cheerfully invent a customer quote. AI Central heads that off inside the prompt rather than catching it in the edit, which is where most people find it too late.

Make the model grade its own shortlist

A second pattern repeats across the pack, ask for ten, then force a decision. Ten sales page headlines, then pick the top three and explain why they will perform best. Ten ad hooks across varied angles, then select the top three. Ten lead magnet titles, then choose the top three and explain why they are strongest.

Volume is the cheap half. The ranking is the half that saves time, because it drags the reasoning into the open where you can disagree with it. When the explanation is thin, the headline usually is too, and you learn that in ten seconds instead of after a week of running it.

Everything is sized for the slot it runs in

The numbers give away that these were written by someone who has shipped the assets. A webinar registration page is capped under 450 words. The 24-hour reminder sits under 180. The one-hour reminder is under 100 and asks for three things only, the time, the join link, and one reason the person signed up.

Video gets timed rather than counted. AI Central's pre-roll prompt budgets a 30-second script down to the beat, five seconds of hook, eight of problem, ten of solution, seven for the call to action, with brief visual notes for each. The long talking head script is specified twice over, three to five minutes and 450 to 750 words, so the model cannot hand back something that reads long and runs short.

Most of the email prompts end the same way, include three subject lines. It is a small habit with an obvious payoff. You never sit down to write a sequence and then stall on the one line that decides whether any of it gets read.

One interview, four assets

The final block is the one most people skip, and AI Central structures it so that skipping costs you. It does not open with a case study prompt. It opens with ten open-ended testimonial questions and a structured interview framework, both aimed at getting a client to talk about their starting point, what they had already tried, why they decided to invest, what they actually did, the results with numbers, and what surprised them.

Only after that does the material get spent. The same interview becomes a 400 to 600 word case study told chronologically, a client result email under 280 words, a first-person LinkedIn post of 200 to 260 words, and a sales page proof section of three or four short studies. One conversation, four assets, nothing invented.

What to actually do with this

The pack assumes you bring the specifics. Your audience, your offer, your competitors, your numbers, your dates. Those blanks are left open on purpose, and the output is only ever as concrete as what you drop into them. Give it a real metric and a real timeframe and the copy lands. Leave them vague and you get exactly the average prose the constraints were built to prevent.

• If you are starting cold, run the first block in order, audience profile, then language library, then content pillars, then the positioning statement. Everything after it reads better because those four exist.
• If you already have an offer, open the block that matches this week's bottleneck. The blocks are self-contained enough to enter in the middle.
• If you are running a launch, the webinar and email blocks work as one sequence, registration page, confirmation, 24-hour reminder, one-hour reminder, promo post, opening script.

The bigger takeaway is the construction, not the collection. A word band, a named required element, and a short list of banned habits will do more for your output than any amount of persona setup. That is transferable to any prompt you write from here on, for any tool.

Do these prompts only work with Claude?

No. AI Central built them for Claude and states plainly that they work with ChatGPT too. Nothing in the wording depends on model-specific syntax, they are briefs, and a brief travels.

Which Claude model should I use?

The prompts in the first two blocks, the research and positioning set and the sales page set, carry a model tag of Sonnet 4.7. The rest run untagged. The reasonable reading is that the heavier thinking is wanted where the strategy gets decided, and matters less once the format is already fixed.

How much do I have to fill in myself?

More than you might expect, and that is deliberate. AI Central describes each prompt as an instruction you copy, paste and fill in with your details. The prompts reference your audience, your offer, your proof points, and in the webinar confirmation case, a real title, date and time. Claude writes the copy. You supply the facts.

Where should I start if I only have an hour?

Run the audience profile prompt, then the language library that builds on it, then the five LinkedIn hooks. That gives you the vocabulary and something you can post the same day. The audience work is reusable across all ten blocks, so the hour compounds instead of evaporating.