TL;DR: AI’s SEO output is underwhelming because AI doesn’t have your expertise earned from years of real SEO experience. Better prompts won’t fix this: the solution is to formalise your knowledge, standards, workflows and skills in a way that AI can work with. And it doesn’t have to be a big effort. This article presents an actionable solution.

What’s wrong with AI’s SEO output?

If you work in SEO, you’ve probably been using AI for a while now. For content drafts, technical analysis, research, brainstorming and more.

Let’s be honest: the results are often… meh. Especially for more advanced stuff. Usable after editing, but rarely at the quality level you’d produce yourself. That’s because the AI doesn’t know what you know: it doesn’t have your years of experience, your knowledge that goes beyond public documentation, or your instinct for what actually works in your specific area of specialisation.

That’s not a criticism of AI in general. It’s just the reality of working with a tool that has no access to your real-world expertise.

What happens when you give AI access to your expertise?

Think about what you actually know after years in this field. Not just the things you could write down in a checklist, but the judgement calls, the patterns you recognise instinctively, the lessons you learned the hard way, or the things you’d correct in a junior SEO’s work without even thinking about it. Most of that knowledge has never been written down (mainly because ain’t nobody got time for that). It lives in your head, and you apply it every time you work.

What if you could formalise that? Not as a one-off exercise, but as something that builds up gradually while you just do your normal work?

No, this doesn’t mean even more work

This is where most of you probably expect the catch. Surely formalising years of experience takes a massive upfront effort? Writing everything down, structuring it, keeping it up to date.

The approach I’ve been using works differently. I just do my work: audits, analyses, content reviews, whatever the project requires. And a process runs in the background that watches how I work. When I correct something AI produces, it notices. When I make a judgement call, it captures the pattern. When I apply a standard that isn’t written down anywhere, it logs it.

Over time, this builds into a set of documented processes and preferences that AI can actually work with. Not because I sat down and wrote a manual, but because the manual wrote itself while I was doing the work. And the next time I work on something similar, the AI already knows how I’d approach it.

A concrete example: technical documentation

Let me give you a concrete example: Over the past 15 years working as an SEO consultant, I’ve developed a specific way of writing technical documentation: audit reports, implementation recommendations, analysis write-ups.

I wasn’t even 100% aware of how exactly I structure findings, how much context I include, how I frame recommendations for different audiences. I never defined or documented my process, I just slowly got better at writing this stuff over the years.

When I started using AI to help draft technical documentation, the output was mostly technically correct (with some help from me) but never sounded like “me”: too generic, wrong level of detail, missing the style I’d developed over years of doing this kind of work. I’d end up having to rewrite most of it.

Now, I have a documented skill that captures how I actually write. My AI knows how I address my audience, that I lead with what matters rather than burying it in context, that I write recommendations as specific next steps rather than vague suggestions. These aren’t rules I sat down and defined: they were captured from how I actually work, refined with every AI draft that I fixed.

The same approach works for any other repeatable process and is not limited to writing. I’ve formalised how I create implementation tickets for dev teams, how I approach specific types of analysis, how I evaluate my own consulting business decisions. And each documented process was created from real work, not from a blank page.

Your work produces more than just the deliverable

Here’s what surprised me most: this didn’t add more work. I was already doing the audits, writing the tickets, and fixing AI drafts and workflows that weren’t right. The only difference is that now, those corrections go somewhere. They compound into something powerful instead of disappearing the moment the task is done.

You do your normal work, and you end up with two things: the deliverable you were going to produce anyway, and a documented, repeatable process that gets better over time.

The second one is where value accumulates: and it costs you almost nothing extra.

One caveat: you do need the real expertise

There’s an obvious caveat and it’s bad news for the grifters: this only works if there’s real expertise underneath. If you don’t have a clear standard for what SEO excellence looks like, there’s nothing to capture. The formalisation amplifies what’s already there: it doesn’t create it from scratch.

And it’s not magic. Over the first few sessions, the documented processes are rough. They improve with use, because every correction refines them. It’s a compounding effect: the more you work, the better the system gets at reflecting how you work.

This is where AI is heading for SEOs

This is where I think AI is actually heading for SEOs and knowledge workers in general:

Not replacing expertise, but making it repeatable and scalable.

Every time you fix an AI output, that’s a correction that could be captured. Every standard you apply instinctively is a pattern that could be documented. The question is whether those corrections and patterns grow into something lasting, or whether they disappear when you close the tab.

My solution is a Claude Cowork meta-skill

I’ve open-sourced the Claude meta-skill that I built to document and formalise my knowledge, experience and expertise:

It’s called “One Skill to Rule Them All” and it works best with Claude Cowork, but you can give it a first try in the Claude web interface (with limitations).

To be clear: By using this meta-skill you’re not getting my expertise – You’re getting a tool that will help you make yours scalable and repeatable.

And to turn up the meta: The most beautiful thing about this meta-skill is that it also observes itself, so it gets better at its own job over time. After using it for a while, you will have your own personal version of this meta-skill, exactly adjusted to your needs. And, of course, you’ll have all the other skills it creates and improves just for you.

Curious? You can start right now

If you’ve been looking for a way to make AI actually reflect the high quality standards of your work, my open-sourced Claude Cowork meta-skill might be worth a look.

If you’re completely new to Claude, just get familiar with Claude Skills first, or skip that step and head straight over to GitHub to install my meta-skill:

One Skill to Rule Them All on GitHub

I’m genuinely curious to see how other SEOs experience this. If you give it a try, I’d love to hear what it captures from your work, or if you feel that it’s not really working out for you.

And if you have any questions at all, please don’t hesitate to reach out to me. I’m always happy to help.