AI SEO Experiments
Real-world experiments testing modern SEO and AI search strategies on learnaiseocourse.com. Method first, result second, whichever way it goes.
What counts as progress
Ranking for the target query is the last of four things that have to happen, not the first. Each stage is reported as it is reached, and reaching none of them is also a result.
- IndexedThe target page is crawled and appears in Google’s index.
- ImpressionsSearch Console records the page being shown for any query.
- Long-tail rankingsThe page holds positions for longer, less contested queries.
- Target-query positionA measurable position for "AI SEO course" itself.
The log
Technical foundation
Canonical URLs, sitemap, robots.txt, hand-rolled metadata component and the first structured data.
SERP and competitor research
Eight competitor profiles, an intent and content-gap analysis, an entity map, and a draft blueprint for the course page.
Course architecture drafted
The earlier plan was to build the course page straight from the blueprint. Instead, a nine-module architecture and an evidence model were drafted first, for review.
Architecture reviewed and locked
Claims checked against primary sources and several corrected. The curriculum on the course page was replaced, and the change was announced on the page.
Lesson specifications
The first 27 lessons specified before any were written, each claim tied to a source and an evidence label.
First lesson published
Lesson 6.1, Crawlability, published on the course page and checked in production.
Search Console and first indexing request
Verify the site in Search Console, submit the sitemap, and record when the course page is first indexed.
What this cannot prove
- One website, one query, one attempt. Nothing here is a controlled trial, and no result generalises on its own.
- Page-one competitors are established domains. A new domain is not the same starting line, and the comparison is not fair to either side.
- Search Console and analytics both sample and both lag. Numbers are read as directions, not as facts to the decimal.
- AI answers vary between sessions and accounts. Any answer sampling is a snapshot of one prompt set on one date.
- Correlation is all a single site can show. Where a ranking moves, the cause stays a hypothesis.
The decisions behind each change are written up as they are made, and the course explains the reasoning. See the curriculum.