BrightonSEO

I ran 30+ GEO experiments so you don't have to: What 200M AI citations say actually works

San Diego, September 2026

Your boss wants an AI search strategy. LinkedIn gurus want to sell you one. Nobody has data - so I spent a year running 30+ controlled GEO experiments across 200M AI citations to test every tactic the industry swears by.

Some results confirmed what we suspected. Most didn't. Why do some Reddit communities earn 9x more AI citations than others with identical content?

Which LinkedIn format gets cited - and which one is a waste of time?

Where does Claude actually pull its citations from? (Hint: not where you think.)

Can AI even see your images and design? And what happened when I built a 1,000-blog AI slop site is stranger than you'd guess - Google noticed fast, but the AI engines are still citing it today.

You'll leave knowing which GEO tactics measurably move visibility on ChatGPT, Claude, Gemini, and AI Overviews, which are folklore, and how to run these experiments on your own site.

Share this talk: