All roles
for marketers

AI for marketers

Marketing is where AI shows up most visibly — and where it most easily produces forgettable output. The marketers winning aren't generating MORE content. They're using AI to do the boring connective work (variants, repurposing, first drafts) so they can spend more time on the few pieces that actually matter.

8 use cases·6 tools·30-min starter
Get the Marketer skill pack (13)Hand-written workflows you install in ~/.claude/skills/

What AI handles well

First-draft blog post from outline

The problem. You have an outline and a deadline. The blank page is where the time goes — not the editing.

What AI does. Feed AI the outline + your brand voice samples. Get a first draft in 5 minutes. Then edit ruthlessly. The 80% that AI writes is the part you don't mind. The 20% you write is what makes it sound human.

Use the promptTools:Claude is best for long-form prose. Customize with your brand voice docs.

Social repurposing (one piece → many)

The problem. You wrote a 1500-word blog post. You want a LinkedIn post, a Twitter thread, a newsletter snippet, and a video script. Doing them all manually = 4x the work.

What AI does. Use AI to extract distinct angles from the same source. The thread isn't a summary — it's the post reframed for that platform's context.

Brand voice consistency at scale

The problem. Your team is growing. New writers don't fully match your brand voice. The output drifts toward generic.

What AI does. Distill your brand voice into a guidelines doc using AI on your best existing samples. Use it as a system prompt for any AI-assisted writing. Audit drafts against it.

A/B variants for emails / ads / headlines

The problem. You need 5 subject line variants. You write 2, get stuck, and ship one without testing.

What AI does. Generate 8-12 variants across different angles (curiosity, urgency, identity, outcome). Test 3 of them. Most variants will be bad — the variance is the point.

Hot-take generation from data

The problem. You ran a survey or analyzed product usage data. You have insights but they're trapped in a spreadsheet. Your audience won't read a 2000-word data dump.

What AI does. Have AI extract the 3 most counterintuitive insights, then write a hot-take piece (LinkedIn, blog, talk) anchored on those. Lead with the surprising stat, end with the implication.

Competitive content audit

The problem. Your competitor just published 5 case studies and an industry report. You need to understand what they're saying, what gaps exist, and how to position yourselves differently.

What AI does. Paste competitor content into AI. Ask it to identify the unspoken claims, the customer pain points referenced, and what they're NOT saying that you could own.

SEO keyword expansion

The problem. You have 3 target keywords. You need 30. Tools spit out lists but they're full of irrelevant variants.

What AI does. Use AI to expand each keyword into 10 variations grouped by user intent (research vs comparison vs purchase) so you can plan content per stage of the funnel.

Image generation prompts that work

The problem. You need hero images, social graphics, blog headers. Generic prompts give you stock-photo-feeling AI slop.

What AI does. Build prompts with style anchors (specific photographers, art movements, lighting). Iterate on a few before committing. Keep a swipe file of prompts that worked.

Your AI stack

Start with the foundation. Add specialized tools as the work calls for them.

Foundation LLM

Claude
Best for long-form writing and matching brand voice from samples. Notably less robotic than ChatGPT for narrative content.
ChatGPT
Strong at structured tasks, lists, variants. Built-in image generation. Custom GPTs for repeated workflows.

Specialized add-ons

Notion AI
For teams already in Notion. Drafts in your existing docs without context-switching.
Midjourney
Best image quality for hero images and visual content. Steeper learning curve than DALL-E but worth it.
Beehiiv
Newsletter platform with built-in AI for subject lines, content recommendations, audience growth tools.
Cursor or v0
For marketers who write code occasionally — landing pages, microsites, custom forms — without bothering eng.

Prompts ready to use

Get started in 30 minutes

1

Pick a foundation LLM and create a brand voice project

10 min

Set up Claude Project or ChatGPT Custom GPT with: 5 of your best brand-aligned samples, your style guide, and a "what we DON'T do" list. Now every draft starts pre-loaded.

2

Run the brand voice extraction prompt

10 min

Use the "Extract brand voice from samples" prompt with your best content. Save the resulting voice doc — use it as system context going forward.

3

Repurpose your last published piece into 3 social variants

10 min

Take a recent blog post or newsletter. Use the Twitter thread prompt to get a thread, then ask for a LinkedIn variant, then a newsletter teaser. Notice the difference vs generic summaries.

Common mistakes

  • Using AI for the parts you should be doing yourself. Headlines, lead paragraphs, and the actual angle of a piece — these are where YOU should add value, not the AI.

  • Generating 50 variants and shipping 50. The point of variants is to TEST 2-3 and pick a winner. Not to flood the channel.

  • Skipping the brand voice setup. Without your voice loaded, AI defaults to generic marketing-speak. "Empower," "leverage," "transform" — that's the smell of a missing brand voice doc.

  • Trusting AI to fact-check itself. AI hallucinates statistics, misattributes quotes, and invents citations. If you're publishing it, verify it.

  • Optimizing for content volume not signal. If you're shipping 3x more content but engagement is flat, you're cluttering your audience's feed without building anything.

Related roles