I don't think content SEO is dead

I don't think content SEO is dead. My small sites reached hundreds of thousands of impressions with $0 ad spend, using a CLI and AI. This is the setup.

It’s been a while on my SEO journey. SEO is something I never liked and in fact avoided. Now I’m somewhat of a pro.

Man in a suit on a campus captioned “You know, I'm something of a SEO expert myself”

Here’s Days to Expiry, which I launched in November.

My setup has pushed my small sites to hundreds of thousands of impressions purely from content SEO, with $0 ad spend.

Google Search Console performance chart with 267 clicks, 230k impressions and average position 7.3

Since I last wrote about the topic, much has changed.

I’ve gone from a bunch of prompts and MCPs to a command line interface and tools.

I’ve integrated a feedback loop and consolidated the SEO part with my Reddit marketing tooling.

For the last week I’ve been running this for three sites and the first results are promising.

My coffee site, now brewedlate.com

Google Search Console performance chart with 6 clicks, 1.79k impressions and average position 14

My oldest site, Expensesorted.com

Google Search Console performance chart with 3.53k clicks, 225k impressions and average position 11.2

The biggest benefit of this system is clearly for developers running multiple sites, but It will work just as well for a single site.

System Overview

All these sites have one thing in common, they have GitHub repos and the repos contain the markdown files with the content.

That makes it easier to manage from outside. All I need is to hook it all up to a tool to orchestrate it all.

I like to use AI via CLI, so why not also a tool that uses CLI and folder structures to manage all this.

Terminal task dashboard showing the Days to Expiry project menu

At first this might look shabby, old and outdated, but it’s really the most basic form of user interface. AI is very good at generating it quickly and you can focus on what matters.

Process and tools.

The core of it all is nothing fancy and as with the CLI UI that’s what I want. I want a workhorse that does the job.

The foundation is a simple folder structure.

Actually two. One for the “control centre” and one for the project that I’m marketing, here Days to Expiry.

Diagram of the coding folder with days-to-expiry, brewedlate and expensesorted repos orchestrated by automation-cli

The control centre contains all the prompts, tools and orchestration logic so that from here I can kick off a workflow such as the research of keywords for some new landing pages.

Terminal menu of actions such as Work on a task and Batch Mode with an available tasks list

Running something like this is pretty trivial if you think about it.

You need something like a task, the task will need to have a set of tools, such as being able to research keywords and creating a specification for a landing page.

In all of this I distinguish between two type of activities

  1. deterministic: what can be coded and will stay the same, such as calling the Google Search Console API

  2. agentic: what needs intelligence, such as the interpretation of these results

These activities can then be chained together.

Flow diagram from Search Console and Reddit data through deterministic and agentic steps to updating content and posting replies

Simple as that. No need for n8n, Openclaw or any of that stuff.

Just Python, a couple of files and AI.

In terms of AI I decided to go with CLI too.

To me, this makes the most sense. You pay for a subscription to Kimi code, Copilot, Codex etc. and that’s all you need to pay.

The main benefit for me is that you don’t need to hassle with the extra API and don’t need to pay for it either. It works just the same as the API, if not even better, because it interacts directly with your tools and files without bridging between local and remote running.

An example for the SEO workflow here would be to fetch data from Google Search Console for a site and place it in a folder. That then can be sent into the AI CLI to analyse it for issues and create specifications to resolve these issues.

Automation

You will have notices that a lot of the above are simple building blocks, that once in place, can be stringed together.

Here is an example, where every couple of weeks I will run a workflow to review and optimise articles that are positioning well or not so well in Google Search Console.

Search Console pages report of daystoexpiry.com blog URLs ranked by impressions

The workflow roughly goes like this:

  1. For a project, fetch the Google Search Console rankings

  2. Get the ranked articles and investigate how to improve them (better structure, linking, depth)

  3. Work through all articles detected and improve them

  4. Publish

Terminal output of an article optimisation task for the theta-decay-dte-guide post

The automation is that it does this for however many articles you might want.

I keep using and iterating this workflow on a daily basis and it’s quite a lot of fun this way.

Pretty sure there are lots of things I can still improve here.

I have packaged up the folder and am happy to share the foundation if someone is interested.

If you'd like a system like this set up for your own site, I offer SEO automation systems for developers and founders. Learn more here.

Originally published at https://blog.learnedlate.com/p/i-dont-think-content-seo-is-dead.