GEO: what makes an AI engine cite you — according to the research

If the decision moves to answer engines, the practical question is: what makes an engine cite one source and not another? Enter a new discipline — Generative Engine Optimization (GEO). Unlike intuitive advice, it already has empirical research: a paper presented at KDD 2024 (a leading data-mining conference) systematically measured which content characteristics raise the likelihood that a generative engine cites a source.

The researchers examined content across a range of queries and tested the effect of targeted changes on visibility in the answer. The findings are sharp, and all point the same way — epistemic authority:

In other words: content built as a backed answer — with a source, a statistic and an authoritative quote — is cited significantly more than generic marketing copy. The study also stressed that efficacy varies across domains, meaning domain-specific optimization is required, not a generic formula.

"Content characteristics associated with epistemic authority — the presence of statistics, source citations and quotations from credible authorities — improved visibility by up to 40%."— GEO: Generative Engine Optimization, KDD 2024

This is precisely the engineering behind a PULSE program page: every claim is written as question → answer → data → source and date, marked up in structured schema (FAQ Schema) that answer engines read. Not cosmetics — a direct application of what the research found. And as GEO itself demands: this very sentence rests on a source, linked below.