Context
An original research preprint introduces GEOFlagBench and an intervention-paired detector for GEO-optimized webpages. Applied to 10,095 usable pages from released Google Search and Gemini-grounded retrieval results, the model flagged 8.90% as GEO-like; the authors explicitly characterize this as an estimate without ground-truth labels.
What changed
The study adds a 3,200-document GEO detection benchmark, an intervention-paired training method, a citation-URL verifiability audit, and a scoped estimate of GEO-like content among sampled retrieved pages.
Why it matters
AI visibility measurement now has a proposed way to audit whether retrieved source content resembles optimization-driven rewrites and whether its outbound citation URLs are verifiable, rather than equating citation frequency with source quality.
Visibility OS interpretation
Visibility OS interpretation: Detection of GEO-like text is a risk and measurement signal, not proof that the publisher manipulated an engine, that a page is inaccurate, or that a deployed platform favors it. Source credibility and retrieval exposure should be measured separately.
What remains uncertain
How well does the detector generalize to unseen optimization methods and languages? How do false-positive rates affect prevalence estimates? Do independent audits of current live search results replicate the sample-specific rates?