AI Chatbots Could Give Every Voice in Community Tourism a Seat at the Table

Community-based tourism has long been celebrated as a model of travel that puts local people first, but the reality on the ground is often messier than the brochures suggest. Villages that welcome visitors must balance the demands of government administrators, outside investors, returning entrepreneurs, small business owners, marginalized residents, and the tourists themselves. Each of these groups arrives at the table with different priorities, different vocabularies, and very different levels of power. A new experimental study published in Humanities and Social Sciences Communications asks a strikingly modern question: could a generative artificial intelligence platform help simulate, and perhaps even improve, the way these stakeholders talk to one another?

The research, conducted by Jinsheng Jason Zhu of Guilin Tourism University and City University of Macau and Aranya Siriphon of Chiang Mai University, set out to test whether a generative AI system could be configured to represent the perspectives of six distinct stakeholder groups within a community-based tourism ecosystem. Rather than treating artificial intelligence as a single, neutral assistant, the researchers explored what they call identity-based chatbot interactions, in which the AI is prompted to adopt and respond from specific stakeholder identities. The premise is deceptively simple: if a chatbot can convincingly embody the concerns of a marginalized resident or the risk calculus of an investor, it might serve as a rehearsal space for the negotiations that determine whether tourism enriches or erodes a community.

The platform at the center of the experiment was Kimi.ai, a generative AI assistant developed by the Chinese company Moonshot AI. The choice is notable in itself. Much of the public conversation about generative AI in tourism and hospitality has centered on Western models, yet community-based tourism in Asia often unfolds in linguistic and cultural contexts that such models handle unevenly. By working with a platform trained with strong Chinese-language capabilities, the researchers positioned their experiment closer to the environments where many of the world’s community tourism projects actually operate, from the karst landscapes of Guangxi to the highland villages of northern Thailand.

The experimental design was carefully bounded. Thirty-six undergraduate students, aged 18 to 22 and drawn from six tourism-related majors, each took on one of six simulated stakeholder roles: government officials, tourism investors, returning entrepreneurs, local business owners, marginalized residents, and inbound tourists. The authors are explicit that these simulated roles did not correspond to the participants’ actual occupations or social identities; the students were role-playing within a controlled research setting. Recruitment, written informed consent, and data collection all took place in December 2024, after the study protocol received prospective ethical approval from the University Research Ethics Review Board at Guilin Tourism University in May of that year, and the research was conducted in accordance with the Declaration of Helsinki.

Once the simulations were underway, participants interacted with the generative AI chatbot while assuming their assigned stakeholder identities, and the resulting interaction logs became the raw material for analysis. The researchers applied two complementary techniques: thematic analysis, which identifies recurring patterns and concerns across the conversations, and sentiment analysis, which tracks the emotional tone of the exchanges. This combination allowed the team to ask not only what the AI-mediated stakeholders talked about, but also how the conversations felt from different positional vantage points. Did the marginalized residents’ concerns register with the same weight as the investors’? Did the tone of negotiation shift when the AI was addressing a government official rather than a tourist?

The findings, as reported in the paper, suggest that identity-based chatbot interactions can effectively address stakeholder-specific needs. The AI was able to tailor its responses to the identity it was engaging, supporting conversations about cultural preservation and encouraging economic inclusivity across the simulated ecosystem. In practical terms, this means the system could discuss heritage protection with a resident concerned about commodification while simultaneously helping an entrepreneur think through a business plan, without flattening those perspectives into a single generic voice. The authors argue that this capacity positions generative AI as a tool for enhancing stakeholder engagement, resource coordination, and problem-solving within community-based tourism ecosystems.

Why does this matter for the broader field? Community-based tourism has always struggled with a participation gap. The stakeholders with the most time, literacy, and confidence, typically administrators and investors, dominate planning meetings, while marginalized residents, whose lives are most directly affected by tourism’s arrival, often lack the channels or the standing to make their concerns count. Traditional participatory planning methods, from town halls to focus groups, try to correct this imbalance but are expensive, slow, and dependent on skilled facilitators. A generative AI system that can simulate stakeholder perspectives offers a low-cost laboratory in which planning scenarios can be tested before they meet reality, potentially surfacing conflicts and inequities while they are still cheap to fix.

The study’s authors are careful to frame their work as an experimental approach rather than a finished solution. The participants were students simulating roles, not actual community members navigating real tourism decisions, and the gap between a rehearsed negotiation and a contested village meeting is substantial. Sentiment and thematic analyses of chatbot logs can reveal patterns, but they cannot fully capture the embodied, historical, and sometimes fraught relationships that define real communities. Still, the methodological promise is considerable: the same simulation framework could, in principle, be deployed with real stakeholders, allowing a village to converse with an AI that has been primed to understand its specific cultural and economic context, or allowing planners to stress-test policy proposals against multiple simulated perspectives at once.

The implications extend to policymakers and practitioners as well. The authors present their results as offering practical insights for those designing tourism governance, highlighting generative AI’s potential role in advancing equitable and sustainable tourism. For a local government weighing whether to license a new wave of homestays, for instance, an AI-simulated stakeholder consultation could quickly map how investors, business owners, and residents might each respond, flagging points of friction before public money is committed. For NGOs and development agencies working in destinations with limited technical capacity, such tools could democratize access to the kind of scenario analysis that has traditionally required expensive consultants.

The research also arrives at a moment when the tourism sector is scrambling to understand what generative AI means for its future. Most attention so far has gone to consumer-facing applications, such as itinerary generation, translation, and personalized recommendations. This study points in a different direction, toward AI as infrastructure for governance and deliberation rather than merely for marketing. That reframing carries its own risks, including the possibility that simulated voices could be mistaken for genuine ones, or that AI-mediated consultation could become a substitute for, rather than a supplement to, real community participation. The authors do not claim to have resolved these tensions, but their experiment makes them concrete and testable.

Looking forward, the researchers encourage future work on emerging technologies to further strengthen decision-making and operational effectiveness in community-based tourism development. The collaboration itself, spanning Guilin Tourism University in China and Chiang Mai University in Thailand and partially funded by Chiang Mai University’s International Research Fellowship program, reflects the cross-border character of the challenge, since community tourism in Southeast Asia and southern China increasingly serves the same mobile pool of visitors while navigating distinct political and cultural systems. Whether identity-based AI simulation matures into a standard planning tool or remains a provocative experiment, the study marks an early attempt to bring the deliberative core of community-based tourism, the messy, essential conversation among unequal stakeholders, into conversation with the most consequential new technology of the decade. If it works, the beneficiaries would be the people community tourism was always supposed to serve: the residents whose homes, heritage, and livelihoods are on the line.

Subject of Research: Generative AI simulation of stakeholder interactions for inclusivity and sustainability in community-based tourism

Article Title: Advancing generative AI for stakeholder empowerment in community-based tourism: an experimental approach to inclusivity and sustainability

Article References: Advancing generative AI for stakeholder empowerment in community-based tourism: an experimental approach to inclusivity and sustainability. (n.d.). https://doi.org/10.1038/s41599-026-09374-3

Image Credits: AI Generated

DOI: 10.1038/s41599-026-09374-3

Keywords: generative AI, community-based tourism, stakeholder engagement, Kimi.ai, Moonshot AI, sentiment analysis, thematic analysis, cultural preservation, economic inclusivity, sustainable tourism, chatbot simulation, tourism governance