Context
A March 2026 original preprint compared 11,000 real-world information queries across GPT-4o-mini API configurations, Google AI Overviews and Google organic results. It reported low overlap between top cited domains of the OpenAI API web-search condition and Google, and uneven use of cited source content in generated answers.
What changed
Researchers analyzed source-domain overlap, response language and source-to-summary coverage in 11,000 Natural Questions queries. The top-100 cited-domain overlap was 25% between the GPT-4o-mini web_search API condition and Google organic results, and 24% between that API condition and Google AI Overviews. A source appearing in citations did not imply proportional contribution to the answer.
Why it matters
AI visibility measurement must distinguish source discovery, citation appearance and source contribution to synthesized answers. Observed source pools and synthesis patterns differed by tested system; a single citation count or traditional Google ranking cannot represent every AI discovery surface.
Visibility OS interpretation
Visibility OS interpretation: treat retrieval/source inclusion, citation exposure and answer-level content absorption as separate metrics, and compare systems and query topics separately. These observations do not identify ranking factors or justify content-length manipulation. The authors' 'Search GPT' refers to a GPT-4o-mini API web_search tool setup, not the consumer ChatGPT Search product.
What remains uncertain
How stable are these findings after model updates, across locations and languages, and on current commercial queries? How much do automatic content-unit decomposition and entailment scoring affect estimated source contribution? Can these patterns be reproduced using consumer search products?