Key Takeaway
The next competition for tourist attention may not take place only on Google, Tripadvisor, or TikTok. More travellers are asking AI systems where to go, where to eat, where to stay, and what to experience. For Georgia, this creates two opposite possibilities: AI could reinforce Tbilisi, Adjara, and other established hotspots—or use personalization to connect visitors with a much wider range of regions and small businesses.
Research currently supports both scenarios. Studies of ChatGPT travel recommendations show a tendency toward established tourist hubs and a much narrower spatial map than tourist activity visible on Instagram. At the same time, chatbot answers vary by user intent, model, and past context, creating the potential for a long tail of personalized recommendations.
According to BTU researchers, the key question for Georgia is not whether travellers will use AI. They already do. The strategic question is what AI knows about Georgia’s regions, how reliable that information is, and whether a system can show different versions of Georgia to different travellers.
From the TikTok Herd to a Personal AI Guide
Travel advice has always shaped tourist flows. Guidebooks concentrated attention through editorial authority. Tripadvisor shifted influence toward crowd ratings. Social platforms added viral algorithms that can suddenly send thousands of people to one photogenic restaurant, village, or viewpoint.
AI adds a fourth layer. A traveller can ask for a quiet destination, a small family hotel, a difficult hiking route, a local winery, vegan food, Soviet architecture, a fixed budget, or a car-free itinerary. In theory, this gives recommendations much more granularity.
But personalization does not automatically create diversity. If the model mostly knows the famous places, it will keep recommending the famous places. The same technology can therefore concentrate tourism or disperse it.
Risk One: AI Often Repeats the Most Established Destinations
A 2025 Current Issues in Tourism study found a marked preference in ChatGPT travel advice for established tourist hubs. Peripheral and sustainability-oriented alternatives appeared, but more often when users explicitly asked for them.
Daniel Paül i Agustí’s Barcelona study offers a more spatially precise warning. His comparison identified 1,321 locations in tourist Instagram data but only 215 in ChatGPT recommendations.
BTUAI calculates that the Instagram location set was about 6.1 times larger, while the ChatGPT set contained about 83.7% fewer locations. The study also identified 38 recommendations that appeared to be invented.
The study does not prove that every current model will behave the same way. It does show the structural risk of a generic prompt: ask what to see in a destination, and the model may default to the most frequently represented landmarks.
Opportunity Two: The Same AI Does Not Give Everyone the Same List
Chatbots are probabilistic and increasingly personalized. Different systems draw on different sources, and the same system can give somewhat different answers to similar prompts.
In an experiment described in the user-provided Economist article, Evertune queried chatbots about pizza in Rome. Different systems favored different restaurants, while Gemini produced a longer tail of 24 distinct pizzerias across its answers. The experiment is illustrative rather than a universal benchmark, but it shows that AI recommendations do not have to collapse into a single ranking.
User intent matters even more. “What should I see in Georgia?” is likely to reproduce iconic destinations. “Where can I spend four quiet August days focused on family wineries, mountain villages, moderate hikes, and low crowds without renting a car?” creates a much broader recommendation space.
Georgia’s Tourism Is Already Geographically Concentrated
The issue matters because Georgia’s inbound tourism is already concentrated. Geostat’s full-year 2025 data show 36.2% of inbound visits in Tbilisi, 25.7% in Adjara, and 11.1% in Mtskheta-Mtianeti.
BTUAI’s verified calculation puts Tbilisi and Adjara together at 61.9% of visits, with the top three regions at 73.0%. The indicator does not mean a traveller visits only one region, but it clearly shows the concentration of recorded visits.
In Q1 2026, Tbilisi, Adjara, and Mtskheta-Mtianeti together accounted for 75.5% of the official regional distribution. Seasonality limits direct comparison with the full year, but both periods show the same structural pattern.
Geostat’s latest Q2 2026 release again identifies Tbilisi and Adjara as the most visited regions, with roughly 875,500 and 621,300 visits respectively.
AI Could Become Infrastructure for Regional Tourism Distribution
This creates an unusual opportunity for Georgia. In traditional marketing, a small municipality, guesthouse, hiking route, or workshop has difficulty competing for attention with Tbilisi, Batumi, or Kazbegi. In a personalized AI environment, relevance can matter more than mass reach.
A traveller interested in ceramics, birdwatching, village food, architecture, remote work, or a specific level of hiking difficulty can be matched with a place that would never win a broad “top ten Georgia” ranking.
But the place has to exist in the machine-readable knowledge environment. “Beautiful nature” is not enough. AI needs concrete information on access, seasonality, transport, price range, booking, physical difficulty, languages, opening hours, and who the experience is actually suitable for.
Destination Data Becomes a New Layer of Tourism Infrastructure
Tourism infrastructure used to mean roads, airports, accommodation, signage, and visitor centers. In AI-mediated travel, destinations also need a reliable digital knowledge layer.
Information about Georgian regions is often fragmented across municipal websites, social media pages, hotel listings, travel blogs, and different languages. Fragmentation makes it harder for both people and AI systems to construct a reliable itinerary.
A modern destination-data layer would include attractions, routes, access, seasonality, transport, price bands, accessibility, safety conditions, official booking channels, opening hours, and updated multilingual descriptions.
Illustrative Georgian Scenario: A Different Kakheti
Imagine a foreign visitor planning three days in Kakheti. Social media may lead them to a handful of photogenic wineries and Sighnaghi. An AI system could instead receive a detailed brief: avoid crowds, prioritize small family wineries, include one local meal, add a two-hour walk, and keep daily spending under EUR 100.
If reliable digital information exists on smaller providers, the system can build a more distributed itinerary and spread spending across businesses that mass marketing rarely surfaces.
If that information is missing, personalization fails. The model falls back to what is most visible online.
What Georgian Tourism Leaders Should Monitor
- Test how major AI assistants recommend Georgia to different traveller profiles, not only through generic prompts.
- Identify regions and experiences that rarely appear in AI recommendations and diagnose whether the problem is weak information, translation, reputation, or discoverability.
- Create thematic micro-itineraries for gastronomy, wellness, architecture, hiking, crafts, family travel, remote work, and other specific intents.
- Give small tourism businesses a simple standard for describing location, price, seasonality, booking, language, capacity, and limitations.
- Monitor invented or outdated AI travel information as a destination-reputation issue.
- Measure regional distribution, nights stayed, and local-business spending rather than only national visitor totals.
- Use personalization to relieve pressure on hotspots without creating new fragile “hidden gem” overcrowding.
Personalization Has Its Own Overtourism Risk
Dispersing tourists is not automatically sustainable. TikTok has already shown how quickly a small location can be overwhelmed. AI could do the same if a village, trail, or guesthouse suddenly becomes the “perfect hidden gem” for thousands of users.
Destination managers therefore need to connect recommendation systems with carrying-capacity thinking: parking, water, waste, trail pressure, resident acceptance, seasonality, and booking limits.
Accuracy is another risk. The 38 fabricated attractions in the Barcelona study are a reminder that incorrect travel information can move from amusing to reputational or even safety-sensitive very quickly.
BTU Researchers’ Assessment
According to BTU researchers, AI will not diversify Georgian tourism by itself. It amplifies the knowledge environment that already exists. If Georgia is digitally represented mainly through Tbilisi, Batumi, Kazbegi, and a few iconic attractions, AI will tend to reproduce that concentration.
If regions build high-quality, multilingual, specific, and continuously updated destination knowledge, personalization can connect niche visitor intent with places that mass tourism marketing rarely surfaces.
Part of future tourism competitiveness may therefore be decided not only in airports and hotels, but in how accurately destinations are described for machines—and how effectively AI can match them with the right traveller.
Key Findings
- AI is becoming a growing intermediary in travel discovery, but it can either concentrate or disperse tourist demand.
- The Barcelona study identified 215 ChatGPT-recommended locations versus 1,321 Instagram locations, a roughly 6.1-fold difference.
- Generic AI travel advice tends to favor established tourism hubs; specific and personalized prompts can broaden the option set.
- In Georgia, Tbilisi and Adjara represented 61.9% of inbound visits in 2025; the top three regions represented 73.0%.
- Reliable, multilingual, machine-readable destination information can become a competitive asset for regional tourism.
- Small tourism businesses may benefit when AI ranks relevance to a specific traveller rather than mass popularity.
- Tourism dispersion must be managed alongside carrying capacity, infrastructure, community impact, and information accuracy.
Why This Matters for Georgia
Georgia’s tourism value depends not only on how many visitors arrive, but where they stay, how long they remain, and which local businesses receive their spending. A wider geographic distribution can spread economic benefits beyond the dominant destinations.
AI creates a new channel for doing this, but only where regions are digitally visible and operationally ready. AI tourism strategy is therefore simultaneously a data, SME, regional development, and sustainable-tourism issue.
Conclusion
TikTok showed how algorithms can send everyone to the same place. AI could create a different model: different places for different people.
That outcome is not guaranteed. Generic questions will still produce generic landmarks. Diversification becomes more likely when destination information is strong, traveller intent is specific, and tourism managers deliberately use personalization as part of destination strategy.
For Georgia, the opportunity is not only to attract more visitors. It is to help existing visitors discover a larger map of the country.
Data and Main Sources
- The Economist – “Can AI save tourists from the TikTok herd?”, August 2026 – user-provided source.
- Daniel Paül i Agustí – The Concentrated City: Effects of AI-Generated Travel Advice on the Spatial Distribution of Tourists, Urban Science, 2025.
- Current Issues in Tourism – ChatGPT and the Tourist Trail: Pathway to Overtourism or Sustainable Travel?, 2025.
- McKinsey & Company / Skift – Remapping Travel with Agentic AI, 2025–2026.
- OECD – Tourism Trends and Policies 2026.
- National Statistics Office of Georgia – Inbound Tourism Statistics, 2025, Q1 2026 and Q2 2026.
- BTU research materials – tourism-market diversification, regional distribution and AI-mediated consumer choice.
Prepared by the academic team of Business and Technology University and the BTUAI Research Team, Tbilisi, Georgia.