Can AI and Shared Travel Stories Help Us Fly Less? Lessons From a Swiss Holiday Study

The researchers built Treeps around “sufficient holidays,” meaning trips that use fewer resources and produce less carbon pollution, through choices such as travelling overland or replacing a short flying break with leisure activities closer to home.

Planning a holiday often starts with searching for flights, even when a train journey or somewhere closer to home could offer the break we need. Cheap fares, familiar booking habits and the appeal of distant destinations can push the environmental cost of getting there out of the picture.

Researchers tested whether an app combining artificial intelligence, practical travel advice and stories from other travellers could encourage people to reconsider flying. Their study, "Toward less air travel: evidence from a multi-component field experiment combining GenAI and peer-based digital nudges," published in Frontiers in Sustainable Cities, examined a three-month experiment in Lugano, Switzerland.

The results offer a cautious picture: giving people access to the tool produced no clear average change in their attitudes or flying intentions. People who engaged more with it reported lower intentions to fly, a finding that raises useful questions about what keeps travellers interested and whether that interest eventually changes their bookings.

Making holidays without flights easier to imagine

The researchers built Treeps around "sufficient holidays," meaning trips that use fewer resources and produce less carbon pollution, through choices such as travelling overland or replacing a short flying break with leisure activities closer to home.

Treeps combined a smartphone app with an Instagram community. Its GPT-3.5 chatbot suggested destinations from a collection of 200 places selected by the research team, supported by photographs, descriptions, public transport information, booking links and estimated carbon savings compared with flying.

Travellers could share experiences involving trains, overnight trains, buses, cycling or walking, giving other users examples of holidays they could picture taking themselves. Posts appeared in the app and on its connected Instagram account, where participants could exchange advice and encouragement. Points rewarded sharing and engagement, with 50 Swiss franc vouchers for the three highest-scoring users. Notifications prompted people to explore destinations and contribute their experiences.

The design followed the Model of Action Phases, which describes change as a progression from questioning a habit to preparing, acting and maintaining a new routine. Treeps tried to support that process by making alternatives attractive, reducing planning difficulties and showing that other people were choosing them.

What the Swiss trial actually found

The experiment recruited 204 volunteers who lived, worked or studied around Lugano, randomly assigning 102 to receive Treeps and 102 to a control group without access. The intervention ran from May to August 2025. Only 146 participants completed both the initial and follow-up surveys: 62 in the Treeps group and 84 in the control group. Most were working-age adults with high levels of education, and the sample travelled frequently, averaging 0.7 holiday flights per person in the three months before the initial survey.

Researchers compared changes within individuals and between the groups, examining personal beliefs, perceptions of other people's behaviour, attitudes toward planes and trains, the perceived ease of using ground transport, and intended holiday flights over the next three months. The main analysis found no detectable average effect on any of these outcomes. Assignment to Treeps was counted regardless of whether someone used it, reflecting the practical challenge that offering an app does not guarantee participation.

A separate analysis examined engagement through app visits, chatbot messages and shared experiences. Higher engagement was associated with lower flying intentions; each additional point on the combined engagement index corresponded to a reduction of about 0.107 intended flights in the model's comparison of changes.

People chose how much to participate, so this relationship cannot show that Treeps caused the reduction. More active users might already have been more motivated to change their travel habits, and estimates became less precise at high engagement levels because few participants used the tool so intensively.

Repeated suggestions and an unexpected social effect

Of the 102 people offered access, 67 opened the app at least once. Twenty-one participants shared 89 holiday experiences, with just three people contributing 50 of those posts. The chatbot also produced 196 destination suggestions covering only 67 distinct places, sometimes recommending the same destination to the same person more than once. The researchers suggest this limited variety may have discouraged participation, especially because destinations became available in the app's browsing section only after the chatbot recommended them.

The social findings were more complicated than expected. Greater engagement was associated with a smaller decline in participants' estimates of how many Swiss people fly for holidays, even as their own flying intentions moved downward. An additional exploratory analysis found that participants without Instagram accounts, representing 18% of the Treeps follow-up group, engaged more with the app. Their estimates of how widespread holiday flying was also increased more than those of Instagram users. Instagram ownership did not clearly alter the result for flying intentions.

Researchers suggest that regular contributors seeing only a handful of people sharing lower-carbon holidays might have felt their choices were unusual. Instagram users could have encountered more travel examples through their feeds, making those choices seem more common. Researchers did not measure how often Treeps posts appeared in participants' Instagram feeds, and this small subgroup analysis was outside the original analysis plan.

What a better travel tool would need

Dropout limits how confidently the results can be read: 39% of the Treeps group missed the follow-up survey, compared with 18% of the control group. Researchers adjusted for measured differences linked to dropout, but could not account fully for unmeasured factors such as motivation or available time.

The remaining sample was also too small to reliably detect some modest changes, so the lack of an average effect does not settle whether the approach could work. Its self-selected, largely highly educated volunteers cannot represent every traveller.

Because Treeps delivered AI advice, carbon feedback, social content and rewards together, the study cannot identify which component contributed to the observed patterns. It examined intentions and psychological factors, with no evidence presented that participants actually took fewer flights or reduced emissions.

Researchers recommend improving chatbot reliability, making all destinations accessible from the start and encouraging participation beyond a small group of enthusiastic contributors. Clearly labelled examples supplied by researchers could help keep the community useful when user contributions are sparse.

Larger studies and longer follow-up would help establish whether changes in intention survive contact with fares, schedules and real booking decisions. The practical lesson is that useful suggestions, easy planning and a community people want to return to deserve as much attention as the technology behind the app.