Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{200145,
        author = {Srushti Pawar and Samruddhi Shedekar and Payal Sharma and Pratiksha Kute and Prof. N. R. Shetty},
        title = {Group Sync: An AI-Powered Group Trip Planner},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {7697-7712},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=200145},
        abstract = {Group travel planning constitutes a complex multi-stakeholder decision problem requiring simultaneous balancing of individual preferences, budget constraints, and fairness considerations. Existing travel recommendation systems either target individual users or require explicit preference elicitation, leaving a significant gap for systems capable of passively inferring preferences from natural group conversation. This paper presents Group Sync, a full-stack AI-powered group travel recommendation system addressing three core research gaps: (1) passive preference extraction from natural language group chat via a Large Language Model (LLM), (2) fairness-aware destination scoring using a variance-based fairness index, and (3) city-level vibe-union clustering that recommends geographically co-located places collectively satisfying diverse group interests. Group Sync employs a multi-criteria ranking mechanism evaluating destinations across four dimensions: vibe compatibility, budget fitness, group suitability, and destination quality rating. Evaluation was conducted through a structured user study involving 50 participants across 15 independent group planning sessions. Group Sync achieved a mean overall satisfaction score of 3.92/5 (78.4%), with destination relevance and vibe match accuracy both exceeding 83%. Budget accuracy, the lowest-performing dimension at 76.0%, reflects the system's reliance on static dataset cost averages rather than real-time pricing. A re-usage intent rate of 76% among first-time users confirms practical adoption potential. Across 15 independent planning sessions, the system produced 15 geographically distinct destination recommendations without repeating the same output across different preference profiles, confirming the absence of popularity bias. The system is implemented using Python, FastAPI, OpenAI GPT-4o-mini, Pandas, React, and Vite, and operates as a zero-cold-start, session-based recommendation engine requiring no prior user history.},
        keywords = {group recommender system; travel recommendation; LLM preference extraction; fairness-aware recommendation; multi-criteria decision making; vibe-union clustering; Indian tourism.},
        month = {May},
        }