<p>Generative artificial intelligence (AI) is spreading quickly in tourism, but destination managers still lack rigorous methods for deciding which applications to prioritise. This paper addresses that gap with a stakeholder-anchored multi-criteria decision-making (MCDM) framework that evaluates generative AI strategies at the destination scale. Drawing on stakeholder theory and sustainability transition theory, we conduct 35 semi-structured interviews with tourism stakeholders in the French Riviera and derive five evaluation criteria: visitor experience (XP), environmental performance (ENV), governance and trust (GOV), economic value (ECO), and feasibility (FEAS). Interview evidence is translated into criteria weights through AHP, and four strategies (itinerary planning, content generation, automated marketing, and service chatbots) are ranked using an ensemble of complementary methods: AHP scoring, TOPSIS (distance to ideal), PROMETHEE II (threshold-based outranking), and ELECTRE III (veto-based outranking with governance and environmental non-negotiables). Cross-method convergence identifies itinerary planning as the top-ranked strategy because it embeds dispersion, lower-carbon mobility, and capacity-aware timing at the point of choice. This advantage depends on strong governance safeguards and reliable local data. Content generation ranks second, while automated marketing and service chatbots perform weakest unless their policies explicitly encode sustainability objectives. Robustness checks (weight perturbations, threshold sweeps, and score noise) confirm the ordering across all four methods. The framework is transferable to other destinations by recalibrating weights and thresholds with a local stakeholder panel while retaining the same multi-method decision logic.</p>