When a German traveler asks ChatGPT where to find the best seaside towns in Italy for August, the answer takes shape in seconds. Whether your destination is inside or outside that answer is decided there, before any click, before any campaign.
This is no longer a niche behavior. Over 60% of travelers now use AI tools to plan trips, and a Klook survey of 11,000 travelers puts the figure at 9 in 10 among people who plan online. Even the most cautious segment, the experienced travelers tracked by Global Rescue's quarterly surveys, more than doubled in nine months. Meanwhile AI-generated answers appear in more than half of Google's travel searches, and DMOs are reporting organic traffic declines of 20 to 40% year over year.
Generative Engine Optimization (GEO) for destinations is the practice of making a destination visible, cited and accurately described inside the answers of AI engines like ChatGPT, Perplexity, Gemini and Google AI Overviews. It doesn't replace SEO: it works on the layer that comes before the click.
What actually changed (it's not the traffic)
The traffic drop is the symptom. The disease is different: the narrator changed.
For twenty years the DMO was the destination's official voice. Today the story is compiled by a language model that has read Wikipedia, Reddit, TripAdvisor, blogs and newspapers, and that grants the DMO no privileged channel. When CNN Travel asked ChatGPT to write guides to its own cities, the result was plausible, confident and full of the small mistakes only a local would catch.
The problem has two faces, and they are worth keeping separate:
- Inaccurate model answers: outdated information, wrong attributions, lost nuance. The model doesn't know the museum changed its opening hours or that the beach is swimmable again.
- Synthetic content polluting the story: AI-generated videos and articles inventing attractions. A couple drove 400 km in Malaysia for a cable car that existed only in an AI-generated video; in Tenerife, tourists went looking for a church in Taganana they had seen in a viral video. It never existed. A study in the Journal of Consumer Behaviour shows how much these errors erode traveler trust.
Neither problem fixes itself. And neither has an owner in most DMOs today: 71% say they are uncertain about AI's impact on their digital activity. With 6 in 10 travelers already on the other side of the chat, that gap has a cost.
The disintermediation misunderstanding
You hear that AI is "cutting DMOs out" of the booking. But the DMO never had the booking: transactions always ran through OTAs and properties, and today Expedia and Booking.com operate as apps inside ChatGPT, with OpenAI walking back its own checkout in March 2026 to focus on discovery.
The DMO's currency was always something else: influence over the choice of destination. That is what's at stake. The consideration set ("where should I go?") now forms in chat, and a destination missing from the answer isn't rejected. It simply isn't considered.
Why move now
- The audience is already there. With over 60% of travelers planning with AI and AI Overviews on half of travel searches, the AI's version of your destination already has more readers than any campaign you run. The only question is whether you are reading it too.
- Value per visitor. According to industry data from Noble Studios, a visitor arriving from AI search is worth about 4.5 times a traditional organic visitor: they arrive with options already compared and stronger intent.
- The competitive window. Princeton's GEO research (Aggarwal et al., KDD 2024) shows that targeted optimizations lift visibility in generative answers by up to 40%. The methods are public, but most destinations don't apply them: the advantage goes to whoever starts first, because models consolidate the sources they trust.
The DMO's real advantage
A language model doesn't care about public mandates. Wikipedia outweighs any institutional tourism portal. What the mandate does give is something no other player has: the position to coordinate. Only the DMO can bring hotels, museums, consortia and events together into one coherent, machine-readable source, the one local operators reference and models learn to trust. Authority isn't claimed: it's built. But the DMO is the only one sitting in the right seat to build it.
What to do, in practice: five moves
- Measure your baseline, by market and by language. Query ChatGPT, Perplexity and Gemini with 20 or 30 real prompts ("best beach towns in Italy for August", "wo in Italien im August ans Meer") in the languages of your source markets. Answers change from language to language: a destination selling in five markets has five different stories to check. Note who gets cited, from which sources, with which mistakes.
- Rewrite for questions, not brochures. Destinations International reports that the average AI prompt is about 7 times longer than a search query: build content on real questions ("how many days do I need?", "what's it like in November?") with direct, extractable answers.
- Publish the data only you have. Visitor flows, seasonality, occupancy, a verified events calendar. Models favor original information over recycled content, and proprietary data is the one thing no blog can copy before you.
- Make the destination machine-readable. Schema.org on attractions and events, an llms.txt file, open data that consortia and operators can reference. Every local operator pointing at your data strengthens your entity in the models' eyes. The Visit West Highlands case shows results in AI Overviews and ChatGPT within about three months; timelines vary, but that's the order of magnitude.
- Patrol errors the way you patrol press coverage. When a model gets opening hours wrong, invents attractions or picks up an outdated story, you need systematic detection and a published, authoritative correction. No DMO would let a mistake run in a national newspaper; AI answers have more readers.
Where to start
The first move, measurement, is the one DMOs skip most often: you fix what you can see. RankWit Destination does exactly that: it monitors every day how AI engines describe and recommend your destination in the markets and languages that matter, benchmarks it against competing destinations, and flags gaps, errors and citation opportunities to act on.
Website traffic was the metric of the old story. The citation is the metric of the new one. Book a demo and see what AI engines say about your destination today. Before they decide it for you.
FAQ
Do travelers really use AI to choose destinations?
Yes, and they are already the majority of people planning online: over 60% according to industry data, up to 91% in Klook's survey of 11,000 travelers. Even the most cautious segments doubled in nine months, and AI Overviews already touch more than half of travel searches on Google.
Does GEO replace SEO for a DMO?
No, it stacks on top. SEO defends the click, GEO defends the answer that comes before the click. The techniques partly overlap (structure, schema, authority), but the goal is different: being the source, not just the result. See our breakdown of GEO vs SEO.
How long before results show?
Documented cases show first citations in about three months for well-structured new content. It depends on your domain's starting authority and the competition on the query. Anyone promising guaranteed citations in two weeks is guessing: models update on their own schedule.
What should a DMO measure instead of sessions?
Three things: which AI answers your destination appears in (share of voice), what exactly they say and from which sources (narrative accuracy), and how that changes over time and against competitors. Those are the KPIs RankWit Destination tracks for every market.
