Field note
What the research says actually moves AI visibility
A Princeton study proved content changes lift AI visibility (up to 40%). The catch is the lesson.
It's easy to treat "getting mentioned by AI" as guesswork. The founding research says otherwise, and its most useful finding probably isn't the one you'd guess.
In 2023, researchers at Princeton introduced Generative Engine Optimization and built GEO-bench, a controlled test set of thousands of queries, to test whether changing your content actually changes how AI engines represent you. It does: their methods raised a source's visibility in AI answers by up to 40% (Aggarwal et al., 2023).
But the headline number isn't the lesson. The asterisk is. What works varies by domain. A content change that lifts you in one field can do little in another; the paper found the effectiveness of each strategy shifted by topic. The strategies it tested were content-quality moves (for example, adding relevant quotations) rather than keyword manipulation, and even those paid off unevenly. (The per-strategy detail is in the paper.)
So the honest takeaway is two-sided: yes, AI visibility is measurable and improvable, not luck. And no, you can't copy another industry's "AEO checklist" and expect it to transfer. Test what moves your answers in your category, and keep what works.
More terms in the glossary; the method is in the overview.
New here? Start with the overview → · By Denes Csaszar · written in the open