Context
Chen and colleagues compared Google top-ten results with citations from four generative search configurations using 1,000 ranking-style prompts and found low average domain overlap, alongside intent- and vertical-specific source differences.
What changed
An original January 2026 study quantified mean domain Jaccard overlap with Google top-ten results at 4.0% for GPT-4o, 11.1% for Gemini 2.5 Flash, 12.6% for Claude 4.5 Sonnet and 15.2% for Perplexity Sonar Pro in its controlled ranking-query sample.
Why it matters
The observed differences show why Google search position cannot be treated as a direct proxy for AI citation visibility, especially when source categories and freshness differ by intent and system.
Visibility OS interpretation
Visibility OS interpretation: audit generative citations independently across engines, query intents and dates, and measure source overlap and answer contribution separately from conventional SEO rankings.
What remains uncertain
How do these historical configurations generalize to later model versions and live products? How sensitive are the findings to query templates, country, date extraction and API settings? Citations do not prove referral traffic.