5 AI Prompts Your Marketing Team Should Be Using Today

TL;DR

• Five prompts I run on real client work: (1) SEO + paid-media landing pages, (2) lead-magnet drafts, (3) CRM/pipeline pattern detection, (4) SEO topic clusters, (5) paid-media account audits.
• The three rules that decide output quality: ask the AI to ask you questions first; design the prompt in a separate session from where you run it; feed it the same context the running model needs.
• Biggest mistake: a one-line prompt with no context, then concluding "AI is useless".
• This article was itself built with the workflow at the end, it's not hypothetical.
• AI doesn't replace strategy. It removes the repetitive analysis around it.

I'll be honest: most "AI for marketing" posts you've read were written by someone who has never exported a CRM at 9pm trying to figure out why a campaign that looks fine is quietly dead. These five prompts are the opposite of that, they're the ones I actually run, most weeks, on real client data.

And to prove the point: this article was itself built with the workflow at the end. Here are the five, copy-paste ready, plus the three rules that decide whether a prompt gives you something useful or generic mush.

Personally, the one I use the most is data-analysis of the CRM. We know how important understanding the qualification of leads is for measuring actual success of marketing actions. Working with prompts like the ones I’m showing you below help me not spend 4h / week / client in normalizing ad platform and CRM data and it makes it way easier for me to see what works and what doesn’t in just a quick look.

What you need to get started

These prompts assume you can export data from tools like your CRM, Google Ads / Meta Ads, Google Search Console and Google Analytics. No advanced technical knowledge required, just the ability to export, structure and paste data.

How do you build a prompt that actually works?

Before the prompts, the three principles that separate mediocre output from genuinely useful results. Apply them every time you design or adapt a prompt.

1. Ask the AI to ask you questions first

When you're about to create a complex prompt, start by asking the AI to interview you: "Before writing the prompt, ask me everything you need to know to do this well." Your answers feed a much richer context, and the final output is more precise and actionable.

2. Build and refine the prompt in a separate session

Don't build the prompt in the same chat where you'll run it. Use a separate conversation, or even a different model, to design, improve and clean it. You iterate on the prompt without polluting the working session with noise. A well-crafted prompt prepared in advance produces consistent results from the first attempt.

3. Give the AI the same context that will run the prompt

The model running the prompt needs to know who you are, what you do, who your audience is and the specific goal. The more relevant context up front, company, vertical, tone, constraints, the less you correct afterwards. Context isn't noise: it's the difference between a generic answer and a useful one.

4. For big projects, split into one prompt per phase

The most common mistake with AI is asking for too many things at once. AI Agents exist for this, they work on one very specific task. No agents? No problem: you get the same result just by conversing. Break the tasks into smaller, manageable pieces.

General rule: a well-built prompt with good context always beats a quick prompt with none. Invest the time up front, it saves you iterations and errors.

Prompt 1, How do you generate SEO + paid-media landing pages fast?

The problem

Creating landing pages for long-tail keywords or paid-media experiments takes copywriting time and structure. Most teams leave these opportunities untapped because production cost is high. With a good prompt you generate structured drafts in minutes.

Step-by-step workflow

• Identify keyword clusters to target (Search Console, Ahrefs, Semrush), or ask the AI to build them for you ;).
• Create a landing strategy spreadsheet: primary keywords, search intent, target audience, product/service description.
• Add context on your differential value proposition.
• Provide all of it to the prompt.

The prompt

You are an SEO strategist and conversion copywriter.

I will provide you with a keyword cluster we want to target with a landing page.

Your task is:
1. Design the complete landing page structure optimised for SEO and conversion.
2. Write a main H1 and the key headings (H2, H3).
3. Include sections that address the search intent and the main objections.
4. Suggest internal linking opportunities with existing pages on the site.
5. Ensure the content also works for paid media traffic
   (clarity, strong CTA, no friction).

Tone: professional, clear and conversion-oriented, maintaining SEO relevance.

What you'll get

A complete landing page structure with H1, headings, objection sections, CTA and internal linking suggestions, ready to send to design or publish directly.

Prompt 2, How do you draft a lead magnet without burning a week?

The problem

Lead magnets are one of the most effective assets for capturing demand in paid media and organic. The catch: creating them well, with logical structure and real value, takes time. AI can generate a complete draft of several in one session.

Step-by-step workflow

• Identify your audience's most recurring problem, or ask the AI to build it for you ;).
• Define the most useful format: checklist, guide, template or framework.
• Prepare: audience description, main problem, chosen format.
• Provide those inputs to the prompt.

The prompt

You are a B2B marketing strategist with experience in demand generation.

We want to create a downloadable lead magnet for our audience.

Target audience: [describe audience]
Main problem: [describe problem]
Format: [checklist / framework / guide / template]

Your job is:
1. Propose a clear and compelling title for the lead magnet.
2. Design the complete structure of the document.
3. Write a brief draft for each main section.
4. Suggest how the content can naturally lead towards our product or service
   without feeling like an advertisement.

The output should be usable as a downloadable marketing resource.

What you'll get

A complete lead magnet draft, title, structure and content per section, with a natural transition toward your service or product.

Prompt 3, How do you find what your CRM is hiding?

The problem

Marketing performance is usually measured only at lead volume. The most valuable insights appear when you analyse how leads evolve inside the pipeline: which channels generate real deals, which sources stall early, where commercial velocity is lost.

Step-by-step workflow

• Export CRM data: lead source, attribution campaign, pipeline stages, timestamps, deal outcome.
• Remove personally identifiable data before pasting.
• Organise it in a clean spreadsheet.
• Paste a representative sample (50–200 rows) into the prompt.

The prompt

You are a revenue operations analyst with experience in B2B SaaS.

I will provide you with CRM pipeline data including: lead source,
campaign attribution, timestamps per stage and deal outcome.

Your job is:
1. Identify patterns that positively or negatively affect pipeline performance.
2. Detect delays or inefficiencies in the commercial process by stage.
3. Compare the performance of different acquisition channels (quality vs. volume).
4. Suggest concrete improvements in lead management or campaign segmentation.
5. Propose 3–5 key metrics that should be monitored regularly.

Focus on practical insights that improve alignment between marketing and sales.

What you'll get

A pipeline analysis by channel and stage, bottleneck identification and concrete optimisation recommendations to improve lead quality and closing velocity.

Prompt 4, How do you compress days of SEO research into minutes?

The problem

SEO topic research means reviewing multiple sources, identifying search intents, organising clusters and prioritising, days of manual work. AI can structure a complete topic cluster, with priorities, in minutes.

Step-by-step workflow

• Choose a central topic tied to your business or a service line.
• Gather context on audience, competition and current positioning.
• Provide that with the main topic to the prompt.
• Use the output as the basis for your content roadmap.

The prompt

You are an SEO strategist with experience in B2B content marketing.

Our website wants to build authority around the following topic: [insert topic].

Your task is:
1. Generate a complete topic cluster around this topic (pillar page + cluster posts).
2. Identify informational, transactional and comparative searches by subtopic.
3. Propose concrete article ideas for each cluster with an editorial angle.
4. Detect internal linking opportunities between the proposed content pieces.
5. Recommend which topics should be published first for the fastest SEO impact.

Structure the output so it can be used directly as an editorial content roadmap.

What you'll get

A structured topic cluster: pillar page, cluster articles, mapped search intents, internal linking opportunities and a publication order prioritised by SEO impact.

Prompt 5, How do you audit a paid-media account objectively?

The problem

Ad accounts grow complex over time: duplicated campaigns, badly distributed budgets, stale segmentations. Auditing them objectively and systematically takes time and experience. With the right data, AI detects structural problems and optimisation opportunities in minutes.

Step-by-step workflow

• Export performance data from Google Ads or Meta Ads.
• Include key metrics: impressions, clicks, CTR, CPC, CPA, conversions, total spend.
• Organise by campaign and ad group in a spreadsheet.
• Paste into the prompt with context on the business objective.

The prompt

You are a senior paid media consultant with experience in B2B and e-commerce accounts.

I will provide you with performance data from Google Ads or Meta Ads campaigns.

Your task is:
1. Identify campaigns or ad groups with low performance and possible causes.
2. Detect structural problems in segmentation or campaign architecture.
3. Highlight clear budget redistribution opportunities.
4. Propose 3–5 concrete experiments that could improve performance.
5. Present a prioritised action list, ordered from highest to lowest expected impact.

The goal is to produce a practical audit that guides immediate account optimisation.

What you'll get

A structured audit with campaign diagnostics, budget redistribution opportunities and a prioritised action list ready to execute.

What do most people get wrong about AI prompts for marketing?

Three things, and they're why people give up too early:

• They paste a one-line prompt with zero context, get mush, and conclude "AI doesn't work". It worked exactly as instructed, the instruction was empty.
• They design and run the prompt in the same chat. By the time the prompt is good, the session is full of noise that contaminates the output. Separate the two.
• They expect it to replace judgment. It replaces the repetitive analysis around the judgment. If you don't know what good looks like, the tool won't decide it for you.

The people getting real value aren't using better models. They're giving better context.

How I actually wrote this article (with these exact prompts)

Not hypothetical, here's the real process, including the rough edges.

1. I asked ChatGPT to generate the initial prompt. I told it the working title ("5 prompts to automate tasks: use AI in your business") and the five tasks I wanted covered, landing pages pushed to the site's GitHub, lead-magnet drafts, CRM pattern-finding, SEO research compression, paid-media audits, and ended with "ask any questions you consider important, feel free to improve the ideas."

2. ChatGPT asked back the right questions: target audience? which AI tools? include copy-paste prompts? rough word count?

3. ChatGPT produced the prompt. I copied it into Claude.

4. Claude generated a first draft, usually as a docx.

5. Then the actual work: making sure the content delivers what I want, and personalising it with sections exactly like this one.

6. To publish, I hand it to Claude Code connected to the site's GitHub. I upload the article as a PDF so Claude reads it correctly. The Claude Code prompt itself was generated by ChatGPT from a set of code-generation rules I keep predefined, so the article ships without breaking the site:

Important operating rules: make the smallest set of changes needed; don't refactor unrelated files; don't change layouts, templates, routing, sitemap generation or blog architecture; don't rename folders or move existing articles; don't touch the other-language blog; only create/update the one article from the provided content; preserve the static Astro build.

(That's the same prompt I used for my Freelancer vs Agency article, worth reading next if you want the long-form version of this workflow.)

Conclusion: AI doesn't replace strategy, it accelerates it

Claude and ChatGPT don't do your strategic thinking. They don't know what positioning makes sense for your business, which audience is most profitable, or how to structure a differential proposition. That stays human.

What they remove is the repetitive analytical work that eats time without creating value: reviewing data, structuring documents, synthesising scattered information, generating first drafts. That frees capacity for what matters, decisions, experiments, growth.

These five are designed for marketers and consultants who already have judgment and want to multiply execution. Start with the prompt that matches your current pain. Don't implement all five at once, one well-executed workflow beats five half-finished ones.

So which one are you actually going to run this week? Be honest, the one you pick says a lot about where your time is leaking.

FAQ

Claude or ChatGPT, which should I use?

Either works; the structure matters more than the model. In practice I design prompts in one (often ChatGPT) and run the heavy analytical ones in Claude. Use a separate session for designing vs running regardless of model.

Do I need to know how to code?

No. You need to export, structure and paste data. The only step that touches code (publishing via Claude Code) is optional, the prompts themselves are plain text.

Is it safe to paste CRM data into an AI?

Remove personally identifiable data first, that's a step in Prompt 3 for a reason. Paste a representative sample, not the raw export with names and emails. Check your own data-handling policy and the model's data settings.

Won't the output be generic?

It will if your input is. Output quality tracks context quality: company, vertical, audience, constraints, goal. The three rules at the top exist precisely to stop generic output.

Can I use these with client accounts?

Yes, that's what they're for. Anonymise client data, and treat the AI output as a first draft you apply judgment to, not a deliverable you forward untouched.

How do I adapt a prompt to my business?

Don't edit the prompt blindly, paste it, then add your context and ask the AI what it still needs to know before answering. That one move adapts it better than rewriting it yourself.