The AI-Driven Digital Wormhole: A Double Machine Learning Audit of Algorithmic Distance Suppression and the Decay of Tobler's Law

As Agentic AI begins to curate global travel, traditional spatial constraints are being superseded by structural metadata. This study introduces the "Digital Wormhole" hypothesis, proposing that high-density digital metadata—specifically JSON-LD structured data—compresses physical distance in AI-driven destination selection. Utilizing a quasi-experimental natural laboratory design across the Gujarat Heritage Circuit, we deploy Double Machine Learning (DML) to isolate the causal impact of structural legibility from geographical and institutional confounders. Our results reveal a significant causal lift (ATE = 0.7616), where structurally optimized remote sites, such as Dholavira, achieve inclusion probabilities equivalent to proximal urban anchors. This empirical decay of Tobler’s First Law suggests that in agentic ecosystems, metadata, rather than territory, defines tourism geography. We conclude that algorithmic gatekeeping creates a new digital divide, where digital luster serves as the primary signifier of a site's existence. The study provides a robust methodological framework for DMOs to transition from conventional content marketing to structural optimization.