CINEMATIC
TOURISM
Insights

Digital overtourism: Could AI undermine efforts to spread tourism away from crowded hotspots?

Woman holding up her mobile phone to photograph the Trevi Fountain in Rome.
A woman photographs Rome’s crowded Trevi Fountain on her mobile phone, reflecting the popularity of the city’s best-known tourist attractions.

Tourism organisations and destinations are working to encourage visitors beyond established hotspots. Yet research published during 2026 suggests that AI travel recommendations can favour familiar destinations, even when travellers ask for alternatives. With AI becoming more influential in holiday planning, could those recommendations compromise efforts to distribute tourism more widely?

Overtourism has become a significant challenge for destinations, particularly where visitor numbers are concentrated in a handful of popular locations or during peak seasons. Managing those pressures, encouraging visitors to explore beyond established hotspots and spreading tourism’s economic benefits more widely across regions and communities have become important priorities.

Destinations are responding to concerns about tourism pressures in certain areas by developing new experiences, improving regional connectivity and encouraging travel outside peak periods. Screen tourism is one part of that wider effort, with film and television providing opportunities to draw attention to lesser-known locations and encourage visitors beyond familiar attractions.

Now AI could introduce another complication to these efforts. Research published by the University of Aveiro in Portugal identified a pattern termed digital overtourism: AI travel systems repeatedly favouring familiar destinations, including when travellers ask for less crowded alternatives. The researchers also identified a related issue of horizontal substitution, where AI suggests different places without substantially broadening the geographical choice.

But there is another problem. AI can overlook destinations that would be a good fit for travellers’ interests and preferences. Skift’s research found that almost a third (29%) of these strong matches were never recommended by the AI systems tested. Travellers may never realise what other suitable destinations they are missing.

With nearly a quarter of UK adults now using AI-assisted sources for holiday inspiration, according to ABTA (the UK travel trade association), the implications extend beyond search visibility. As more travellers turn to AI for inspiration, could these narrow recommendations reinforce existing tourism concentrations and undermine efforts over the last decade to distribute visitors and create more sustainable tourism?

The challenge of overtourism

Concerns about overcrowding, pressure on infrastructure and the impact on local communities became especially prominent in international tourism discussions during the late 2010s.

In 2018, the UN World Tourism Organization identified dispersing visitors geographically and across different seasons among its strategies for managing overtourism. Yet concentration remains a substantial issue. According to the OECD’s Tourism Trends and Policies 2026, approximately 47% of tourist nights were concentrated in just 10% of the regions covered by its data.

When AI keeps recommending familiar places

In their April 22026 research paper, Digital overtourism in AI travel recommendations: evidence from a comparative analysis, Najafi and Costa from the University of Aveiro examined 420 responses from ten AI systems, using 14 structured tourism queries repeated three times.

The researchers tested general destination recommendations alongside requests for sustainable travel and alternatives to crowded locations. Although the responses collectively mentioned numerous destinations, the recommendations frequently focused on a narrower selection of iconic and semi-iconic places.

When asked for alternatives to Santorini, for example, some systems suggested other Cycladic islands, including Milos and Naxos, rather than taking travellers outside an established tourism circuit. The researchers also tested requests for hidden gems in Spain and ways to avoid crowds in Italy, finding that recommendations across the study frequently remained concentrated around familiar destinations.

The researchers describe this as horizontal substitution: offering somewhere different without substantially broadening the geographical choice.

[VISUAL INSERT 1: Horizontal substitution]

Sustainability-focused searches revealed a similar problem. AI systems sometimes described destinations as ‘eco-friendly’ or ‘responsible’, even when recommending places already associated with mass tourism. The language suggested a more sustainable holiday, but the choice of destinations had barely changed.

The AI recommendation loop

How existing destination visibility could become reinforced through repeated AI recommendations.

DIGITAL OVERTOURISM? Existing digital prominence Uneven representation AI selection Filtering and ranking Familiar destinations appear repeatedly Concentrated recommendations Reinforced visibility Potential feedback
420
AI-generated recommendations
10
AI systems tested
14
Structured queries
Research found that destination recommendations repeatedly converged on a relatively narrow group of prominent places, despite appearing varied. The animated cycle illustrates a proposed mechanism of visibility reinforcement, not proven changes in physical visitor flows.

Source: Najafi and Costa (2026), Digital overtourism in AI travel recommendations .

Why does this happen?

The researchers identify popularity bias and uneven digital representation as possible explanations. Well-known destinations already have extensive coverage across travel publications, websites, reviews and other online sources. AI systems trained on or retrieving information from these sources can reproduce that imbalance.

Separate research from digital marketing agency SALT provides another indication of the problem, this time within Google AI search. Its May 2026 analysis of 7,900 Google AI Overviews relating to UK staycations found that established cities, including London, Edinburgh and Manchester, repeatedly appeared across different travel searches. Regions such as the Cotswolds and Lake District featured but were less prominent as primary recommendations.

SALT’s research suggested that familiar cities benefit from extensive, consistent online information, while geographically broader regions can be harder for AI systems to identify and present as distinct destinations. Its findings suggest that the way places are represented online may be as important as the range of experiences they offer.

There is also the prospect of a reinforcing loop, where places already having a dominant presence online receive repeated recommendations, potentially attracting further attention and making it harder for smaller destinations and options to compete for visibility.

Of course, established destinations have always enjoyed advantages in travel media, guidebooks and conventional search. But AI introduces a new layer of complication. Rather than browsing several sources and comparing suggestions, travellers may receive a ready-made shortlist that appears tailored to their interests. They see the recommendations, but not necessarily the destinations that were left out. For smaller destinations, the challenge may therefore extend beyond being visible online to fighting for a place in that initial selection.

Personalised recommendations, familiar choices

Skift Research’s October 2026 report, The Mystery of the Disappearing Destination, identified a related problem.

To investigate how much choice AI actually offers, Skift tested ChatGPT, Gemini and Claude against 21 destinations and six traveller profiles, covering different interests, budgets and travel needs. Its research found that almost a third (29%) of destinations assessed as strong matches for particular travellers were never recommended by the AI systems tested. The implication is significant: a destination can be well suited to someone’s interests and be filtered out when AI narrows the choices.

The research also found that AI often recommended the same destinations to travellers with different interests and budgets, and only changing the descriptions to suit their preferences rather than suggesting different places. Families might receive suggestions about child-friendly attractions while budget-conscious travellers were offered affordability advice, yet the destinations themselves suggested often remained much the same.

For travellers, the result can look like a broad range of personalised choices when in reality the AI is not necessarily offering much geographical variety. And unless they look elsewhere, they may have little reason to suspect that suitable destinations have been excluded.

The research also raises questions about how widely AI spreads attention between destinations. The five most frequently recommended places accounted for 38% of all recommendations, rising to 48% for the top ten. That means nearly half of the recommendations went to just ten destinations. For tourism organisations trying to promote lesser-known places, this imbalanced concentration matters: even destinations that are well matched to a traveller’s interests may struggle to compete for a place on the shortlist.

There was, however, a more encouraging finding. Skift found that AI recommendations were less concentrated on popular destinations than existing travel patterns might suggest. The ten destinations AI recommended most frequently accounted for 48% of its suggestions. By comparison, those destinations received 56% of all international tourist visits in 2024.

This is a potentially encouraging finding. AI may be giving travellers a wider choice of destinations than existing travel patterns suggest. But that doesn’t mean people are actually visiting different places. And it doesn’t resolve the bigger concern: AI is still overlooking destinations that would suit travellers’ interests and preferences.

There is also a practical reason why AI offers limited choices. In a US survey included in Skift’s research, 63% of respondents preferred a curated shortlist of ten destinations over a much larger catalogue or a single AI-selected itinerary.

With almost two-thirds of US travellers surveyed preferring a shortlist of destinations, AI’s ability to narrow the choices is clearly appealing to many. For destination marketers heading into 2027, the bigger challenge is what to do when even places that are a strong match for travellers may never make the list.

Why growing AI adoption matters

What might once have been a relatively limited problem could become much more significant for destinations as AI influences more people’s choices about where to go.

According to ABTA’s Holiday Habits 2026–27 research, the proportion of UK adults using AI tools for holiday inspiration has more than tripled in two years, rising from 4% in 2024 to 14% in 2026. Among 25–34-year-olds, almost a third (29%) now use these tools. When those using AI-generated answers in ordinary Google searches are also included, the proportion of UK adults using AI for holiday inspiration rises to nearly a quarter (24%).

The trend is not confined to the UK. The International Air Transport Association’s October 2026 Global Passenger Survey found that nearly half (49%) of surveyed passengers worldwide had used AI for travel planning, although only 17% reported using it specifically for destination inspiration. Searching for flights or comparing fares is not the same as asking AI where to go, but the figures demonstrate how widely the technology is entering travel planning.

AI also appears to play a greater role when travellers are deciding where to go than when they make a booking. Skift’s research found that AI accounted for around one in five (21%) recorded travel-inspiration interactions, compared with just 6% at the booking stage.

Travellers may discover destinations through AI but book elsewhere, making its influence difficult to measure. As more people use AI for inspiration, recommendations favouring already crowded destinations could place additional strain on local infrastructure, services and communities, undermining efforts to manage tourism more sustainably.

Another study (Propellic, September 2026), involving 38 US travellers planning real trips, found that most stuck with one AI platform rather than comparing recommendations across several. If that behaviour is widespread, any bias towards already crowded destinations could go unchallenged, potentially adding to existing tourism pressures.

The concern is what happens if AI adoption continues to grow while recommendations remain concentrated on familiar destinations.

Could growing AI use intensify digital overtourism?

For destinations already struggling with overcrowding, the implications with growing AI use could extend well beyond just online visibility. If AI influences more travel decisions without offering greater variety, it could add to pressures on local infrastructure, public services and communities.

Taken together, the findings are a warning the tourism industry should take seriously. AI risks adding to the pressures already facing overcrowded destinations, from stretched infrastructure and public services to housing shortages and the impact on local communities. Those tensions were evident during the summer of 2026, with major protests in Mallorca, demonstrations over access to public spaces in Ibiza, opposition to coastal tourism development in Portugal and campaigns over beach access in Italy. At a time when destinations are trying to ease overcrowding and encourage tourism in less-visited regions, the last thing they need is technology working against those efforts.

What can destinations do now?

Tourism organisations aware of the challenges faced, are tackling this problem and reviewing how their destinations appear in AI-generated travel advice. 

Tourism Ireland: Promoting regional experiences

Tourism Ireland chief executive Alice Mansergh at Skift’s Transatlantic Summit in Dublin in October 2026, discussed efforts to bring lesser-known experiences into AI-generated travel advice, including sheepdog demonstrations in Sligo and traditional boatbuilding on the Aran Islands.

Tourism Ireland is improving information about these experiences across its own channels and platforms such as Reddit, YouTube and Tripadvisor. The organisation reports that its content contributes to nearly one in five AI-generated answers about Ireland.

Singapore: Measuring AI visibility

Singapore Tourism Board has been tracking how its tourism information appears in AI-generated responses. An agency-submitted case study published by The Drum reported that 81 of 108 priority VisitSingapore.com pages were cited in AI-generated answers. It also recorded a 1.34-fold increase in measured AI visibility between September and December 2025.

This marks a positive shift in how tourism boards measure their effectiveness in AI search. Rather than focusing solely on where their websites rank in search results, they can now assess whether their information is being used to answer travellers’ questions.

New Zealand: Improving regional information

In July 2026, Tourism New Zealand completed AI discoverability workshops to help tourism businesses improve their online information.

The government also committed NZ$800,000 to improving tourism information systems and adding regional content to Tourism New Zealand’s AI travel assistant, with regional visitor dispersal among its objectives.

What These Case Studies Mean for Destination Marketers

For destinations, these examples provide a practical starting point: test AI-generated itineraries against local tourism priorities, identify overlooked experiences and improve the information available about them.

A small set of realistic travel questions, tested regularly across different platforms, could reveal gaps that conventional website analytics would miss.

There is also a responsibility on AI providers. More comprehensive destination information will achieve little if recommendation systems continue prioritising popularity without considering the pressures destinations face or their capacity to accommodate visitors.

What this means for screen tourism

VisitBritain’s 2026 research identifies screen tourism as an opportunity to encourage visitors to explore lesser-known UK locations and travel outside peak seasons.

A production might feature locations across several regions, yet an AI-generated itinerary could send visitors to only its most recognisable filming sites. That would limit the value of screen tourism initiatives designed to encourage exploration beyond the headline attractions.

For film commissions and destination marketers, the useful test is whether AI-generated travel plans include the wider filming locations and experiences they have been promoting, rather than just the obvious stops.

When AI redirects the journey

The destination is trying to relieve pressure on an overcrowded location. AI recommendations may direct travellers straight back towards it.

TRAVEL DISCOVERY Traveller considers where to go DESTINATION STRATEGY Promote alternative experiences AI RECOMMENDATIONS Favour an existing hotspot AI SHORTLIST ALTERNATIVE LOCATION LOCATION ALREADY UNDER STRAIN
Destination’s intended diversionAI-recommended route

Digital overtourism: A growing concern for 2027

There is an uncomfortable prospect for the tourism industry heading into 2027. AI could add considerably more to an already crowded in-tray, bringing new strategic headaches on top of the daily pressures of tackling overtourism and developing more sustainable tourism practices. Destinations can spend years developing visitor-management strategies, yet an increasingly influential part of travel planning now appears to operate outside their control.

The concern grows as more people turn to AI for holiday ideas. Travellers can receive advice apparently tailored to their interests while being offered the same destinations, repackaged with different descriptions. A tourism organisation might be trying to relieve pressure on a particular neighbourhood or attraction, only to find it appearing in AI-generated itineraries despite those efforts.

Yet tourism authorities are not powerless or without hope. The challenge is navigating a rapidly changing field of AI search optimisation, often referred to as Answer Engine Optimisation (AEO) or Generative Engine Optimisation (GEO). As AI systems themselves continue to evolve, the methods and advice for improving visibility must also adapt, creating further uncertainty for destinations that are feeling pressure to take action or risk losing visibility or facing additional visitor pressures. Whilst there are established ways to improve a destination’s online information, there is no universally agreed method for ensuring that particular places or experiences make it into travellers’ recommendations. Even when that information is available to AI systems, the destination itself may still be left off the traveller’s shortlist.

With AI-generated answers also reducing clicks to websites from traditional Google searches, destinations face another difficulty: reaching travellers through channels they can measure and influence.

The initiatives in Ireland, Singapore and New Zealand demonstrate positive ways to improve destination information, monitor AI responses and work with local tourism businesses. The University of Aveiro researchers go further, calling for changes to how AI systems rank destinations, better representation of overlooked places and the inclusion of visitor-capacity information.

That raises a larger challenge for 2027: destination management’s digital strategy cannot stop at the boundary of a tourism board’s website and other digital touchpoints. As digital overtourism becomes a growing concern, the wider industry needs closer dialogue with AI providers about how recommendation systems operate, how they account for local visitor pressures and how destination-management priorities can be reflected in their decisions. Otherwise, tourism organisations risk being left to manage the consequences of recommendations they had little say in producing.