AI Won't Do Your Marketing for You — and a 37,000-Star Repo Just Proved It

You believe AI is about to do your marketing for you.

That with a smart enough model, your company will write its own content, optimize its own ads and nurture its own leads — while you put your feet up. The whole market is running on that belief. Everyone is hunting for a better tool, a newer model, a more "magical" prompt.

I'm about to show you something that turns that belief upside down.

There is a repository on GitHub called marketingskills, built by Corey Haines — former Head of Marketing at Baremetrics and SavvyCal, and founder of Swipe Files. At the time of writing it has more than 37,000 stars and 6,000 forks. That is a level of attention 99% of software projects never get near.

And here is the part that made me write this whole post: there is not a single line of AI code in that repo.

No model. No algorithm. No "artificial intelligence" in the sense you'd expect. It is just 60-plus markdown files — plain text — holding one expert's marketing knowledge: how to optimize conversion, how to write copy, how to do SEO, how to price, how to fight churn.

The whole world is obsessed with AI. Yet the thing 37,000 people chose to star is human expertise, written down systematically. Sit with that paradox for a second.

AI is now a commodity. Expertise is not.

To understand why that repo is valuable, look at what the market actually looks like.

According to industry statistics roundups, 69.1% of marketers have already built AI into their strategy, and 79% use AI to write content; that figure is expected to hit 94% in 2026 (SEO.com). In other words: nearly every one of your competitors is using the same kind of AI, with the same handful of tools, often with the same handful of prompts.

When everyone has the same hammer, the hammer stops being an advantage. It becomes the baseline.

And the consequences are already obvious. While marketers pour money into AI content, consumer enthusiasm for AI-generated content collapsed from 60% in 2023 to just 26% in 2025 (eMarketer). People have an unflattering name for what's flooding their feeds: "AI slop" — bland, repetitive, soulless AI content.

Corey Haines's repo isn't valuable because it "runs AI". It's valuable because it packages the most expensive, scarcest part of marketing — an expert's judgment — into something an AI can read and act on. The AI is the free part. The expertise is the part worth paying for.

Rule #1 of the whole system: positioning first, tactics second

If you take only one thing from this post, take this.

Across the entire repo, one skill is set up as the foundation for all the others. It's called product-marketing. Its design principle, paraphrasing the docs closely, is:

"The product-marketing skill is the foundation — every other skill checks it first to understand your product, your customers and your positioning, before doing anything."

Read that again. Before doing anything.

Before writing a line of copy. Before spending a dollar on ads. Before optimizing a single button. The smartest AI marketer this 37,000-strong community has endorsed is required to answer three questions first: What do you sell? Who do you sell it to? Why do they choose you over anyone else?

That's not a technical detail. It's a philosophy. It contradicts, head-on, the popular belief that "a good tool produces good results". It doesn't. A brilliant tool that doesn't know who you are will just produce more polished AI slop — fluent, grammatical, and completely meaningless to your customers.

It's also the thing I repeat to every client: a beautiful website without clear positioning is just an expensive brochure. Website-building tools keep getting cheaper and faster. Whether that website actually makes money is decided in the strategy layer underneath the interface.

What is positioning worth? A concrete number

Saying "strategy matters" is easy. Let's look at a real number.

HubSpot analyzed more than 330,000 calls to action (CTAs) over six months and found that CTAs personalized to the viewer converted 202% better than generic ones (HubSpot).

202%. More than three times as well.

Same website. Same product. Same AI writing the copy. The only difference: one side knew who it was talking to and what to say at the right moment, and the other didn't. That's positioning turning into money — not a fancier AI model.

That's why I say AI isn't your leverage. Positioning is your leverage. AI just helps you execute that positioning faster — if you have it. If you don't, AI just helps you produce the wrong content faster.

So what do you do? A three-layer framework instead of tool-hunting

Stop looking for a smarter AI. Start building the part AI can't come up with for you. Here's the right order.

Layer 1 — Foundation: Positioning (first, and non-negotiable)

Before any campaign, write down in plain language:

• What you sell, described the way customers say it, not the way your team says it internally.
• Who you sell to — specific enough that you can picture a real person, not "all small businesses".
• Why you — one reason customers pick you over a competitor, or over doing nothing at all.

If those three answers are fuzzy, everything after them will be fuzzy too — whichever AI you use.

Layer 2 — Tactics: Let AI execute on the foundation

Once positioning exists, AI becomes genuinely useful. Writing the homepage, building emails, drafting ads, proposing A/B tests — but always feed it Layer 1 first. AI can't guess your positioning; it only amplifies what you give it. Garbage in, garbage out; sharp strategy in, sharp content out.

Layer 3 — Measurement: Let data decide, not gut feel

Set up real conversion tracking. Measure what makes money (leads, orders), not what pretties up a report (views, likes). That 202% figure only exists because someone measured it. You need to measure too, so you know which positioning wins.

Notice the order: AI lives in Layer 2, not Layer 1. A community of 37,000 engineers and marketers has effectively agreed on that. Reversing the order is the most common — and most expensive — mistake I see businesses make today.

What I got wrong

I'm not saying this from the stands.

There was a stretch, when AI writing tools first exploded, when I got swept up too. I once handed a client a website with AI-written copy — smooth sentences, perfect grammar, very "professional". I was rather proud of how fast it shipped.

It converted at almost zero.

The reason looks obvious in hindsight: we had jumped straight to Layer 2 and skipped Layer 1. The copy said lovely things about "us", but it didn't hit the pain of any specific customer. Grammatically right, positionally wrong. We had to rewrite it — this time starting from the three foundation questions — and only then did the conversion rate start to move.

A lesson I paid for in money and time, handed to you for free: the faster AI gets, the more expensive a positioning mistake becomes — because you replicate the mistake at machine speed.

Bottom line: You're not short on tools. You're short on positioning.

Go back to the original paradox. A repo with no AI in it was starred by 37,000 people in tech, in the middle of the biggest AI frenzy in history. Because they understand something the crowd hasn't yet:

In the AI era, the scarce thing is no longer the ability to produce content. The scarce thing is knowing what to say, to whom, and why.

AI has turned execution into a near-free commodity. That doesn't lower the value of strategy — it pushes all of the value toward strategy. The companies that win the next five years won't be the ones with the best AI. They'll be the ones that know most clearly who they are, and only then let AI amplify it.

If your current website runs on "AI slop" — copy that sounds nice but doesn't sell — the problem almost certainly isn't the tool. It's the positioning layer underneath. That's exactly where I start every project: see how I work, or talk to me directly to reset the foundation before you put another dollar into tools.

Stop looking for a smarter AI. Become the business an AI is worth working for.

---

Sources:

• Corey Haines's marketingskills repo — GitHub: https://github.com/coreyhaines31/marketingskills
• Corey Haines / Swipe Files: https://www.swipefiles.com/
• AI adoption in marketing (69.1% / 79% / 94%) — SEO.com: https://www.seo.com/ai/marketing-statistics/
• Consumer enthusiasm for AI content fell from 60% (2023) to 26% (2025) — eMarketer: https://www.emarketer.com/content/exclusive--ai-slop-threat-creator-economy
• Personalized CTAs convert 202% better (analysis of 330,000+ CTAs) — HubSpot: https://blog.hubspot.com/marketing/personalized-calls-to-action-convert-better-data