Generative AI and Search Engine Orchestration (GASEO): First Experimental Report — Initial Evaluation of Dual-Dimension Suitability Scoring for Generative AI and Search Engine Routing

The increasing functional overlap between generative artificial intelligence (AI) systems and traditional search engines creates a need for a unified mechanism through which users can express information needs without having to determine in advance which technology should process their requests. Generative AI systems are particularly suitable for tasks involving content generation, reasoning, transformation, synthesis, and interactive assistance, whereas search engines are particularly suitable for retrieving existing information, locating websites, discovering products and services, and accessing current or externally verifiable information. Nevertheless, many real-world requests involve both capabilities, making a rigid separation between generative AI and search engines insufficient. This report presents the first experiment of Generative AI and Search Engine Orchestration (GASEO), a framework designed to coordinate generative AI and search engine capabilities through a unified task-oriented interface. The original GASEO approach is based on intent understanding, capability suitability assessment, automatic execution-mode selection, hybrid execution when necessary, and the possibility of user-directed mode selection. The framework is intended to move beyond requiring users to choose between a search engine and a generative AI system before submitting a request. The first experiment investigates an initial component of this approach: whether the relative suitability of an input for search engines and generative AI can be assessed using two independent scores. Ten generative AI prompts and ten search engine queries were first generated using Gemini 3.5 Flash-Lite. The resulting twenty items were then randomly reordered and submitted to GPT-6 (ChatGPT Free). GPT-6 was instructed to assign each item a Search Engine Score and a Generative AI Score, each ranging from 0 to 100. The item-level classification was determined by comparing the two scores and selecting the capability with the higher score. The reported Search Engine Score total was 1,035, corresponding to a mean of 51.75, while the Generative AI Score total was 1,165, corresponding to a mean of 58.25. More importantly, all twenty items were classified consistently with their original intended categories: ten generative AI prompts were assigned higher generative AI scores, and ten search engine queries were assigned higher search engine scores. This produced an observed classification agreement of 20/20, or 100%, on the experimental sample. These results provide an initial demonstration that comparative suitability scoring can distinguish clearly constructed generative AI-oriented prompts from search-oriented queries under the experimental conditions. However, the experiment does not establish generalizable routing accuracy, actual search retrieval performance, or end-to-end GASEO effectiveness. The findings should therefore be interpreted as an initial proof-of-concept result that motivates larger, more diverse, and independently evaluated experiments.