How destinations stay seen as AI becomes the front door
AI already has a view of every destination. The question DMOs are working with now is a different one: how does a destination shape the story AI tells about it?
Every source the model can read
In this report
02The argument
03Tiki Envoy
04Oberösterreich and Peterborough
05Strategic recommendations
06Re-enter the zones
The argument
From being found to being cited
What changed
AI sources its view of a destination from everything it can read. The model reads the destination's own website alongside travel press, Wikipedia entries, Reddit threads, reviews and operator descriptions, and forms a picture of the place from all of them. That picture is what surfaces in answer to any question a traveller asks.
For years, ranking high on Google meant a destination was being seen, and the two systems walked together. That logic is breaking. The overlap between top Google rankings and the sources AI Overviews actually cite has fallen from around 70% to under 20%. A destination can rank well in classic search and still sit invisible in AI Overviews, because the model is looking at different signals.
The overlap is collapsing
Google links
AI sources
70%
Overlap
Top Google rankings vs AI Overview sources · Keynote, X. Design Week 2026
The signals that matter to AI are signals of authority. A 2025 study by the University of Toronto with Muck Rack found that 94% of AI citations come from earned media, third-party sources, press and reviews. Just 6% come from brand-owned content. The destination's own website is the weakest lever it has. What others say carries the weight.
94%
Earned media, third-party sources, press and reviews
6%
Brand-owned content
University of Toronto GEO study 2025, via Muck Rack
Three layers
This reshapes how visibility gets built. It works on three layers, each one resting on the previous.
The foundational layer earns the right to be read at all. Clean schema markup, JSON-LD, canonical URLs, a robots.txt that names the AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) and an llms.txt index at the site root that gives the models a curated map. The foundational layer is invisible to the visitor and central to the model. Most destinations are not on it yet.
The narrative layer is what gets said about the destination. The model reads owned content alongside third-party coverage and forms a picture. The DMO can write into that picture proactively, or wait to see what the model decides on its own. Either choice has consequences. Reactive work makes the destination legible to the view AI already holds, reinforcing whatever bias the training data carried. Proactive work introduces new themes and asks the model to learn them.
The ecosystem layer is the wider footprint. Operators, partners, travel press and the industry as a whole talk about the destination. When their language aligns with the DMO's narrative, the model picks up consistency and amplifies it. When the ecosystem says one thing and the DMO says another, the model averages them or ignores both. Industry alignment scales narrative.
How visibility gets built
Select a layer to see what sits inside it.
Ecosystem
The wider footprint
Narrative
What gets said
Foundational
The right to be read
The right to be read
Foundational
What this asks of measurement
Traditional measurement no longer tells the whole story. Tracking clicks, sessions and standard keyword rankings is no longer enough. What matters now is whether the destination appears in these new search answers, how accurately it is represented, and how much direct traffic the platforms send. These metrics are newer than the habits most teams have built. Organisations that adopt them early gain an immediate advantage, figuring out what works long before the rest of the field even knows what to count.
Two fundamental choices sit at the centre of this strategy. The first is about approach: do you passively accept what these platforms already say about the destination or actively work to change it? The second is about focus: do you try to capture every possible user query or commit to a few core narratives that the destination can defend and local businesses can echo?
Choosing to stay absent carries a compounding cost. If the destination does not provide the source material, someone else will tell its story. Small local businesses will slip behind well-funded commercial competitors. Factual errors regarding safety, accessibility or local culture will cause real harm, simply because the interface presents them with absolute authority. Ultimately, the destination's silence becomes someone else's claim.
Case study
Tiki Envoy
Building an ad unit for a world moving away from classic search.
10x
Google
harder over the past decade
750x
ChatGPT
harder, by the same measure
30,000x
Anthropic
harder still
Over the past decade, directional figures from Tiki's reading of the market
Tiki built Envoy as an ad unit that can act on a visitor's behalf, designed for a world where destination traffic is moving away from classic search. The data behind that move is sharp. Google has become roughly ten times harder to get traffic from than a decade ago. ChatGPT is around 750 times harder. Anthropic, by the same measure, sits at 30,000 times harder. The numbers show direction, and the shape of the change is clear. Open search is closing, and the field is reshaping.
ChatGPT's user base now sits at around 14% of Google's. The site is the fifth most visited on the internet and the user base is growing roughly every three months. AI Overviews are pulling traffic away from organic search across every category, with around a 30% drop in organic search traffic from the highest performing sites. The decline applies across every category that depended on search.
Generative Engine Optimisation
Tiki's response is to treat Generative Engine Optimisation as its own practice. The principles read like a rewritten editorial brief. Write to user intent. Use natural language and conversational structure. Build a clear narrative with segments that stand alone. Answer the question the prompt is asking. Cite authoritative voices and back claims to source. Produce detailed explanations and how-to guides, which are the content types the models read in depth.
Measuring what matters now
Measurement shifts to match. Using Tiki, teams can track AI Answer Share to see the exact percentage of travel responses that explicitly reference or recommend them. This sits alongside AI Citation and Attribution Frequency, which logs how often the destination’s original content is linked, cited or paraphrased. The system tests messaging through a Narrative Alignment Score, checking the degree to which those descriptions actually match the brand's intended pillars. Meanwhile, Prompt Coverage Across the Journey maps visibility from early inspiration right through to planning and comparison. Rounding this out is the Zero Click Influence Index, which solves the hardest modern tracking problem: measuring the sway an answer holds over a traveler's decisions without them ever visiting the website.
Tiki's five new metrics
01AI Answer Share
02AI Citation & Attribution Frequency
03Narrative Alignment Score
04Prompt Coverage Across the Journey
05Zero Click Influence Index
GEO as a channel
The wider point in Tiki's argument is that GEO is a channel, with its own budget, its own measurement and its own discipline. Destinations that treat it as one more line of marketing spend will overcrowd it. Destinations that integrate it into the editorial operation will produce work that does the foundational, narrative and ecosystem work at the same time.
Case study
Two Destinations, Two Visibility Strategies
Building cited authority through memorable content.
These case studies show how two different destinations tackled the visibility problem. Both arrived at the same conclusion: the best way to get referenced by an algorithm is to create work that sticks with a human. Their approaches prove what strong source material and clear storytelling look like in practice once the technology becomes a standard part of the daily workflow.
Oberösterreich Tourismus
1.5MYouTube views
Oberösterreich Tourismus showcased In Unserer Natur, a campaign built around an AI-generated deer that raises awareness about respectful behaviour in nature. The campaign reached 1.5 million views, becoming the most-liked content the destination had ever published on YouTube. Nevertheless, it is important to recognise that channel match matters as much as the content itself. While the campaign performed strongly on YouTube, TikTok and the destination's own website, the video did not work on Instagram, where creator audiences are currently sceptical of AI content.
Throughout 2025, the team tested Google Veo, learning how to better direct the software while bypassing its eight-second limit by stitching clips together in CapCut. ChatGPT handled the base character images that Veo then animated. A major breakthrough followed with Google Flow, which finally allowed them to keep a subject looking the same from one shot to the next. Audio proved harder. Off-the-shelf voices struggled with the local Austrian dialect, forcing staff to record the voiceovers themselves until the technology caught up. By late 2025, the system could replicate the regional accent well enough to use directly.
Peterborough Tourist Board
Dinosaur City
Discover Peterborough described a parallel approach with the Dinosaur City campaign, where AI is being used to bring the destination's Jurassic heritage to life. Nano Banana was used to create AI-generated dinosaurs, which were then merged with real photographs of the city using Veo to create a hero video. Palaeontologists from the Natural History Museum reviewed the imagery for anatomical accuracy, resulting in improvements to how the dinosaur's front feet were visualised. This acted as an important moderation process to keep AI-generated content credible.
Suno generated the campaign music, while short social snippets were developed with Kling AI. Alongside AI discoverability, experience development was also considered. Claude was used to design floor decals to place around the fossils located in the city's shopping centre. These link via QR codes to the vibe coded campaign website, featuring detailed educational information about dinosaurs and the related experiences in Peterborough.
build
Shared takeaway
The shared lesson across both campaigns is that no single tool does the job. Workflow integration matters more than tool selection. Each team built its own pipeline, with editorial discipline running through every stage. Both treated AI as a production capability, with human judgement firmly at the gate. Accuracy checks and brand voice oversight stayed with humans, particularly on the parts of voice that the models still got wrong.
Strategic recommendations
Key Questions and the Way Forward
This section pairs the primary concerns DMOs brought to the room with the practical solutions agreed upon by the group. Five clear recommendations emerged.
01
How do we measure success in a zero-click world?
The most effective step is moving two-thirds of the weight away from website visits and toward citation metrics. This means tracking your AI Answer Share (how often the software explicitly recommends you) alongside Narrative Alignment (whether those generated answers actually get your brand story right). When the platforms do generate direct visits, treat them as a distinct acquisition channel with its own dedicated reporting line. The most urgent task is setting a baseline today. Because this landscape offers no historical data, the goal of your first quarterly audit isn't to hit a target but simply to draw the starting line. If you fail to map your footprint now, any progress you make six months down the road will be impossible to prove.
02
How do we balance GEO with paid media and social?
expand_more
Treat GEO as a distinct third pillar alongside paid and social, complete with its own budget, headcount and reporting. The core strategy still relies on knowing the audience and maintaining a steady brand voice. The writing itself, however, has to move at two entirely different speeds. Social feeds demand instant hooks. AI platforms demand exhaustive, highly structured authority. Destinations must either train their editorial teams to operate at both of these extremes or bring in specialised partners who can.
03
How do we maintain accuracy with AI-generated content?
expand_more
Introduce a mandatory review stage for any material touched by AI before it goes live. Factual claims must be verified by relevant subject-matter experts. Tone requires a standard editorial sign-off. Generated imagery needs a human eye to vet it for local and cultural accuracy. This strict internal discipline is the only reliable defence against publishing the software's fabricated details as absolute fact.
04
Which tools fit which task?
expand_more
Tool selection must always follow process design. Start by mapping out your production stages, from initial research and drafting through to final editing, publishing and measurement. The destinations managing this successfully do not rely on a single platform. Instead, they combine specialised AI applications for video or audio with standard language models and traditional editing software. How seamlessly these systems connect with each other matters far more than the specific tools you select.
05
How do we influence the recommendations agents will make on behalf of visitors?
expand_more
The automated AI assistants that plan trips for travellers rely on the exact same data as today's AI search results. The work you do to improve your visibility right now is precisely what will shape those automated recommendations in the future. Destinations establishing their presence today will be the ones these platforms confidently suggest over the coming years.
Recommendations to take into Q3 and Q4 2026
01
Audit your current AI footprint
Select five core search queries that matter most to your destination and run them through the main AI platforms. Document exactly what they say, what they miss, and which external sources they credit. This gives you an immediate, realistic baseline.
02
Fix the technical foundations
Ensure your website's back-end code and access permissions are updated so these systems can easily scan and index your data. This is the least expensive work to execute, but it delivers the highest long-term value.
03
Commit to three or four core themes
Choose the specific topics your destination genuinely owns and build deep, thorough content around them. Focus entirely on quality and authority, and resist the urge to try and answer every single type of question online.
04
Coordinate with local partners
Brief your local operators, businesses and the travel press on these core themes. When multiple independent websites and articles repeat the same consistent facts, it builds massive authority across the wider web.
05
Build the measurement habit immediately
Set a strict quarterly schedule to review your performance and track how things change over time. Destinations that master their numbers today will enjoy a massive head start over competitors who delay their tracking.
Each session approached the challenge of AI visibility from a distinct angle. The summaries below outline the discussions and the practical conclusions reached by each room, while the links provide access to the full digital environments built during the event.
science
The Lab
What happened
The Lab took the discoverability question into hands-on practice. The session moved through three angles. The technical layer covered schema markup, how to write it and how to read what AI is currently picking up from a destination site. The content layer worked through reverse prompt engineering, which means asking AI systems what they currently say about a destination and tracing back to the sources that shaped that answer. The tooling layer brought in browser-based AI assistants (Claude in Chrome, ChatGPT agents) and showed how they can be used to analyse website performance from an AI perspective. A live exercise had participants run their own destination sites through the analysis. The session surfaced a debate about whether browser-based agents are allowed inside the organisation, with several legal teams having already restricted them on internal devices. The room recognised the friction between practical exploration and current governance posture, with no easy resolution available from either side of the conversation.
Takeaway
The technical work is achievable today. Schema, llms.txt and crawler permissions can be implemented in a sprint. The governance work, around which AI tools sit inside the organisation's boundaries, needs to catch up before practical exploration becomes everyday practice.
The Strategy Room examined how destinations can take a more deliberate and structured approach to AI visibility, moving beyond ad hoc content creation toward a position they actively design and sustain. The group spent time understanding how AI overviews actually pull and weight information, recognising that the algorithms are opaque, subject to frequent change and shaped by signals, such as authority, recency and indexing frequency, that destinations can influence but never fully control.
The question of where to focus effort surfaced consistently throughout the discussion. Chasing every query AI systems might generate is a path that does not end and the session drew a clear distinction between reactive content, which answers the questions people are already asking and proactive strategy, which shapes the narrative around the destination at a higher level. Both have value, but the higher-value work sits in the latter: curating a coherent story across the destination's own content, its industry partners, its PR activity and its creator relationships so that the same core themes are triangulated from multiple credible sources.
The session closed on data infrastructure, with a discussion of how machine-readable destination data, structured and accessible to AI systems through mechanisms such as MCP, represents the next stage beyond content optimisation. Rather than answering every granular question through articles, destinations can build an intelligence layer that allows AI systems to answer precise, real-time questions directly. That shift raises a question the group left open: at the point where a DMO is building data infrastructure for AI access, the work starts to look less like marketing and more like destination operations.
Takeaway
Destinations that lead with a clear strategic narrative and distribute it consistently across owned, earned and partner channels are better placed than those optimising content in isolation. The shift toward machine-readable data infrastructure will matter increasingly and destinations with well-organised data assets will find themselves ahead of those still treating discoverability as a content problem alone.
The Debating Room focused entirely on hard trade-offs, acknowledging that the ideal approach outlined in the morning keynote requires resources most destinations simply do not have. The discussion split into three distinct schools of thought on where to spend a limited budget first.
The technical argument stated that without proper back-end code and access permissions, any subsequent investment is completely wasted. The editorial argument countered that building genuine authority takes much longer than fixing infrastructure, meaning the storytelling has to start immediately. Meanwhile, a third group focused on the local network, suggesting that since businesses and the travel press are already publishing anyway, simply aligning these existing voices offers the highest return for the lowest spend.
While the room never settled on a single winner, the debate forced an honest look at tight financial realities, with several participants sharing candid stories of having to make these exact choices and what happens when you get them wrong.
Takeaway
This choice is a practical one that depends entirely on each destination's financial reality. The real internal debate is deciding which of those three strategies gets the first dedicated role or funding line and what needs to be put on hold to pay for it.
The Advisory Clinic worked through direct DMO submissions one by one. Several themes recurred across the submissions. Reverse engineering the destination's current AI presence was the foundational starting point, because no DMO can plan a visibility strategy without first knowing what AI is currently saying. The SEO and GEO divergence emerged as a central thread, with the room recognising that the two practices aligned through 2025 but are pulling apart now, with keyword-based work giving way to contextualisation and narrative depth. Net information gain, the principle that each piece of content should add something the model could not find elsewhere, became a shorthand the room returned to. Schema markup, llms.txt and crawler permissions came up as the technical baseline most submissions had not yet addressed. Diversifying into Reddit and other social forums where AI now reads heavily was a thread several DMOs picked up, raising follow-up questions about how to engage these platforms without losing brand voice.
Takeaway
One specific tension stayed in the room without easy resolution. Destinations promoting responsible tourism and dispersal narratives want strong GEO performance, which rewards depth on well-known anchors. The two pulls work against each other. The room landed on treating the dispersal narrative as its own owned theme and writing it with the same depth as the anchor narratives, which preserves the responsible tourism intent while still earning the AI authority signals.