No bullsh*t, no spam. Only exclusive SEO & AI/ML tips
Live Last read · last published · next check
This week I published something I've wanted to write for a long time: a deep dive into rerankers, probably the least discussed layer of the AI search stack in the GEO/AEO/AI search world.
We put about 1 billion pages online. Then we watched who paid for what.
You write in your language, about your country. Google cites Reddit's machine-translated version instead. Here's the data.
I took Meta AI's TRIBE v2 model and turned it into a free SEO analysis tool. TRIBE v2 was trained on 1,115 hours of real fMRI brain recordings from over 700 humans.
The things are getting interesting. I'm getting natural inbound links and growing my link profile for free. After indexed 37k+ pages. Part 2 of the $10 AI website experiment.
Nowadays, I run OpenClaw as my primary SEO automation agent. It schedules tasks, generates content, monitors rankings, and handles the repetitive work that used to eat my mornings.
Recently I have been experimenting with extracting query fanouts from Google AI Mode. There is a simple prompt structure that has been working surprisingly well since the release of Gemini Pro 3.
A wild experiment in AI Slop programmatic SEO and an unexpected discovery about AI crawler behavior.
Most Discover advice sounds like this: publish great content, use big images, post consistently, and let the algorithm find you. That’s not wrong. It’s just incomplete.
A comprehensive weight-level analysis revealing how the model processes, routes, and generates content and how to architect your pages to align with its internal mechanisms.
