GEO Statistics: Generative Engine Optimization by the Numbers
Generative engine optimization is young enough that the data is still forming. This hub collects the most useful GEO and AI-citation numbers published so far, each with a named source and a link. It is updated monthly and will grow first-party numbers from CitationWorks monitoring as our dataset matures.
What moves visibility
The original GEO research found that optimizing content (adding quotations, statistics, and citations to sources) can improve a site's visibility in generative engine answers by up to 40%.
The paper that coined the term GEO. The gains came from making pages easier to quote as evidence, not from keywords.
62% of AI citations do not lead to a brand mention in the answer, according to a Semrush analysis of citation behavior.
Getting cited and getting named are different outcomes. Monitoring both is the point of tracking a trust gap.
Where AI citations come from
Reddit was the most-cited source across ChatGPT, Google AI Mode, Gemini, Perplexity, and AI Overviews, with YouTube, LinkedIn, Wikipedia, and Forbes rounding out the top five.
Third-party surfaces dominate. An AI answer about your category is usually assembled from places you do not control.
LinkedIn appeared in roughly 11% of AI responses on average across ChatGPT Search, Perplexity, and Google AI Mode, making it the second most-cited domain in Semrush's dataset.
Citation shares are volatile: in Semrush's tracking, ChatGPT cited Reddit in close to 60% of responses early in the study window and around 10% about six weeks later.
A single model or retrieval update can reshuffle sources in weeks. This is why one-off audits mislead and continuous monitoring matters.
An index synthesizing more than 680 million AI citations across six published studies found that a group of roughly 50 websites decides most of what brands are visible inside ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.
Citation concentration is far higher than classic search. Earning presence on a short list of trusted surfaces goes a long way.
First-party CitationWorks data
Cross-model agreement, own-domain citation rates, and accuracy rates from live daily monitoring of ChatGPT, Claude, and Gemini will publish here as our monitored dataset reaches publishable scale.自社データ
We only publish first-party numbers once the sample is large enough to be meaningful, with methodology alongside.
Where these numbers come from
Public GEO data is early and uneven, so this page holds a higher bar than most stats roundups: every third-party number links to the study that produced it, with scope and dates where the source states them. Aggregator blogs quoting other aggregators do not count as sources here.
The most valuable GEO statistics do not exist in a public dataset yet, because almost no one measures AI answers at scale. That is what CitationWorks does daily. As monitored coverage grows, this page adds first-party data: real citation rates, real accuracy rates, real mention-share shifts, sourced from live monitoring.