About
How the guides on this site are made
AI search is a young field with more claims than evidence. This page explains exactly how guides here are researched, drafted, reviewed, and kept current — so you can judge how much to trust them.
What's the short version?
Guides are researched against primary sources, drafted with AI assistance, then fact-checked, tested, and edited by a human — me — before publishing. Claims carry dates and sources, uncertainty is stated rather than smoothed over, and pages are revisited as the field changes.
How does a guide get made?
Every published guide goes through the same five stages.
-
1
Research
Each topic starts with primary sources: published studies, official platform documentation, and industry data from named research organizations. Blog folklore and "everyone knows" claims don't qualify as sources on their own.
-
2
Draft
Guides are drafted with the help of AI tools, working from that research. This is disclosed rather than hidden — it's how one person can maintain a library this size. The draft is a starting point, never the finished product.
-
3
Review and test
I fact-check the draft against its sources, cut anything that can't be supported, and where possible test the advice directly — on this site and in the AI tools themselves. Advice that survives is advice I'd follow with my own website, because I do.
-
4
Publish with dates and sources
Statistics and platform-specific claims are attributed and dated in the text, so you can see how current something is and check it yourself. Every guide shows when it was last updated.
-
5
Revisit
AI search changes fast enough that an accurate page can quietly become a wrong one. Guides covering volatile details are reviewed quarterly; durable-principle guides are reviewed when the field gives a reason to. Updates change the visible date.
How do you handle what's unknown?
Honestly: a lot about AI search is unknown, including to the people who build the engines. How models weigh sources shifts between versions, most published "GEO tactics" haven't been rigorously tested, and vendors have an incentive to sound more certain than the evidence allows.
The rule here is to label claims by their strength. Some things are well-established (engines can't cite pages they can't crawl). Some are supported by early research (structure and clarity appear to matter for citation). Some are plausible but unproven — and when a guide includes one of those, it says so. You should never have to guess which kind of claim you're reading.
I'm also developing this site's first original research project — systematically tracking how AI engines describe real organizations — so that over time more of what's published here rests on data gathered firsthand rather than secondhand.