AI Environments
Models with and without web search, and consumer surfaces. Vorentus observes each class separately because they are not methodologically equivalent.
Why the classification exists
A memory-only model answer and an answer assembled from live web search are produced under different conditions and answer different questions. Combining them into one number implies an equivalence that does not exist.
Vorentus records the environment, the model actually served, the retrieval mode (with or without web search), the access mode and the repetition count with every observation, and reports web-search observations separately from memory-only observations.
ChatGPT (consumer) is a separately identified consumer-facing surface captured via a licensed data provider, not an OpenAI API response. OpenAI API, retrieval-enabled APIs and Google consumer search experiences are documented as distinct observation methods, never merged as though equivalent.
Gemini is observed through the Gemini API, a programmatic measurement environment without web search. It is not the Gemini consumer app, Google AI Mode or Google AI Overviews, and its results are never presented as what those products show.
Gemini with Google Search grounding is a possible methodological modality, but it is not currently available in Vorentus measurements. Where web-search observation is in scope, it is performed through the retrieval-enabled APIs named in each report.
Location
Measurements requested from 187 UN-recognised countries, 27 territories and 1,175 cities, using exact location codes. Every result states whether the AI engine confirmed the location it used.
Detailed coverage in Portugal, Brazil and Spain, including district, state and provincial capitals, and Europe's main cities.