Claude for DMOs. The Definitive Guide for Destinations That Want to Start Using Anthropic's AI.

How to configure Claude for a destination: Projects, Skills, Connectors, autonomous agents and chatbots. [long post: get comfortable]

The gap isn't technological. It's organisational.

Over the last eighteen months I've spoken with destinations, APTs (regional tourism boards), regional bodies and national DMOs all trying to figure out what to do with AI. The question was always the same: "Where do I start?"

It's not a trivial question. Neither is the answer.

Only 6% of travel companies are already actively scaling agentic AI across multiple domains and functions (Phocuswright, 2026). Hotels give their own AI systems an average reliability score of 4.7 out of 10 (h2c AI Study, 2025). Everyone's buying, few are actually using it, almost nobody trusts what they've installed.

Meanwhile, travellers aren't waiting. 56% of American leisure travellers already use AI to plan at least one trip — double compared to a year ago (Phocuswright, March 2026). Under 35, we're already past 70%. ChatGPT has 900 million weekly active users (OpenAI, February 2026). Google AI Mode reached 75 million daily users by late 2025. Google AI Overviews appear in roughly 48% of queries (BrightEdge/Ahrefs, March 2026). 63% of the destinations studied by Tempest saw year-on-year organic traffic losses in April 2025.

The technology is already in travellers' pockets. The organisations that are supposed to serve them are still debating whether it's worth it.

Claude doesn't solve the problem on its own. It doesn't replace people who know the territory, or who can read seasonal data. But for a small team with big expectations, it multiplies what you're able to produce.

---

The context every DMO needs to be clear on in 2026

Three things are happening simultaneously in destination marketing. Understanding them changes how you use Claude.

1. The funnel has contracted

The classic model (awareness, research, evaluation, booking) now plays out in a single conversational session. A traveller asks an AI agent to search, recommend and book. They get a shortlist. They book. Without having opened a single DMO website. Discovery happens inside the AI, before the traveller ever reaches you.

The destination website is no longer the point of discovery. It's the point of confirmation. If you get there at all.

2. Travellers are using multiple AIs in parallel

Amadeus calls this "Travel Mixology": LLMs for the big picture, Reddit and YouTube for real experiences, OTA AI assistants for personalisation. 64% more people are using generative AI tools to plan travel compared to last year (Amadeus 2026). But almost a quarter receive incorrect or outdated information, and fewer than half actually trust what they read.

This is where DMOs have an opening: those who structure their content well get cited by AI. Those who don't disappear from the conversation before the traveller even reaches the website.

3. AI search has upended who wins in search results

A study by Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi across 10,000 real queries in 25 sectors (GEO paper, ACM SIGKDD 2024) found that adding statistics to content increases visibility in AI responses by 41%. For content that already ranks low in traditional results, the "cite sources" technique reaches up to 115%. A small hotel or a well-structured secondary destination can appear above a large but poorly structured OTA in AI results. In traditional SEO this doesn't happen. It's probably the most important news of 2026 for European tourism, where the majority of operators are small and compete against enormous platforms.

---

The 7 real pain points of DMOs in 2026

Data from: 350+ global DMOs (Sojern/Dynata 2026), Tourism Tribe "State of AI in Tourism 2025" (Australia, UK), DTTT framework, Phocuswright 2026.

1. Shrinking budgets, growing expectations. 61% cite budget as the main digital obstacle (Sojern 2026). 31% of Australian operators name it as a barrier to AI adoption (Tourism Tribe 2025). Budgets won't grow in the short term. Expectations will.

2. Technical competence absent from the team. 40% of operators cite technical competence as their main barrier (Tourism Tribe 2025). 71% have started using AI, but only 19% feel genuinely capable. It's not laziness: accessible practical pathways simply don't exist.

3. ROI invisible to stakeholders. 72% of DMOs cite "demonstrable economic impact" as their primary KPI (Sojern 2026). Hotel room nights, average spend, conversions — not impressions. Those who can't tell this story with data lose trust and funding.

4. Content production is drowning the team. 51% struggle to keep pace with what the channels demand. 78% of Australian tourism organisations are run by a single operator or micro-team (Tourism Research Australia). VISIT FLORIDA produced 124 unique videos in 513 versions in one year. With small teams, without AI it's physically unthinkable.

5. Visibility in AI search is already at risk. 63% of the destinations studied by Tempest recorded year-on-year organic traffic declines in April 2025 — including many that were already in position #1 on Google (Tempest, 2025). Google AI Overviews appear in 48% of queries (March 2026). Perplexity handled 780 million monthly queries in May 2025, growing 20% month on month. If AI doesn't cite you, you don't exist at the moment the traveller decides.

6. The data is there, but nobody has time to read it. 45% struggle to analyse the data they already have effectively; 43% can't integrate it from different sources (Sojern 2026). The result: decisions made on gut feeling instead of evidence, despite the data being available.

7. Personalisation is stuck while demand rises. Travellers in 2026 expect personalisation (Amadeus: "Pick 'n' Stays", attribute-based selling). 45% of DMOs are still at "moderate" level, 22% at "basic" (Sojern 2026). The distance between expectation and ability to respond is the most underestimated risk.

---

The setup: how to configure Claude for a DMO

Step 1: Create a Claude Project for your destination

Go to claude.ai → "New Project". Name the project after the destination. Projects remember context across conversations: you don't have to re-explain who you are every time.

One project for the destination. Another for MICE work. Another for seasonal campaigns. Contexts stay separate.

Step 2: Upload your key documents

In each Project you can upload files that become the permanent knowledge base. What to upload:

• Brand guidelines and voice & tone document
• Destination fact sheet (key data, main attractions, target audience)
• Annual and stakeholder reports from the last 2 years
• Current marketing strategy plan
• Examples of already-approved, high-performing content
• Visitor FAQs (from your email inbox or chat)
• Seasonal data (occupancy, average spend, origin markets)

Step 3: Write the Project's custom instructions

Each Project has a "Project Instructions" field. Write a briefing along these lines [everything in square brackets IS TO BE CUSTOMISED]:

You are the editorial and strategic assistant for [DESTINATION NAME],
a DMO responsible for [TERRITORY].

Our voice is: [e.g. authentic, direct, community-oriented]
Our primary target: [e.g. families from northern European markets, MICE from Germany and UK]
Our main KPIs: [e.g. hotel room nights, average spend, qualified visits]
Our key stakeholders: [e.g. local council, APT, hoteliers, associations...]
Our primary markets: [e.g. DACH, Benelux, UK, domestic market]

When writing content:
- Always use a [e.g. warm but professional, never salesy] tone
- Cite data whenever available
- Avoid unsourced superlatives ("the most beautiful destination in Italy")
- Use short sentences in social posts, more elaborate ones in strategic documents
- Write web content optimised for AI search: direct answer at the top,
  Q&A structure, citable data, FAQs
- If you're unsure about a fact, tell me explicitly

Step 4: Enable Web Search and learn the levels

In Claude Settings, enable web search. This allows Claude to verify current data, search trends, news about the destination, and what competitors are doing.

The three levels of use:

• Basic: you use Claude for general questions, copy the output without editing. It works, but wastes the potential.
• Intermediate: you've set up the Project with your brand voice, you use Claude for specific, repeatable tasks, you edit and review the output before publishing.
• Advanced: Claude knows your organisation in depth, works with your real-time data via Connectors, produces output that feeds directly into your workflows.

My advice is to start at Intermediate. Not Basic, not Advanced.

---

8 practical use cases (with ready-to-use prompts)

---

🗞️ 1. Content production at scale

The problem: you need a blog post, a newsletter, 5 social captions and a press release, this week. With a team of 3.

I need to communicate the opening of the 2026 summer season for [DESTINATION].

Create:
1. An SEO and AIO-friendly blog post of 700 words — open with a direct answer
   to the question "What to do in [DESTINATION] in summer?", use H2s as real
   visitor questions, include a closing FAQ with 4 questions
2. 3 Instagram captions (short, medium, long) for a beach sunset photo
3. A newsletter lead (3 lines) for the B2C contact list
4. An A/B email subject line test (2 variants)

Data to include: [paste your fact sheet here]
Tone: [define]
Main call to action: discover itineraries at [URL]

---

📊 2. Stakeholder report: turning data into narrative

The problem: you have a CSV with occupancy, sentiment and traffic data. You need to present it to the board of directors (or the local councillor) by Friday. Nobody has time to write it up.

I'm attaching/pasting this raw data from the summer 2025 season for [DESTINATION].

Write a 2-page executive summary for [STAKEHOLDER] including:
- 3 high-impact headline numbers (format: "X% / €X / X visitors")
- A narrative connecting the results to the promotional investments
- 2 issues that emerged, with a proposed action for 2026
- Tone: professional but readable by non-marketing experts
- Format: continuous text, no tables (will be projected on slides)

[paste data]

---

🔍 3. AEO content audit: is your site structured to be cited by AI?

The problem: you don't know whether your site is structured so that ChatGPT, Gemini or Perplexity cite it when a traveller asks about your destination. 63% of destinations have already lost organic traffic because of this.

Technical note: Claude cannot query ChatGPT, Gemini or Perplexity directly — they are systems belonging to different companies. What it can do, with Web Search enabled, is analyse your site and evaluate its structure against AEO/GEO best practices, which determine whether AI models cite you or not. To directly check what ChatGPT and Perplexity actually respond, open them manually and type the typical questions your visitors ask. For systematic monitoring over time, consider rankwit.ai.

(Enable Web Search)

Analyse the site [DESTINATION URL] and assess how well optimised it is
to be cited by AI search systems (ChatGPT, Gemini, Perplexity,
Google AI Overviews).

Retrieve the site content, then tell me:

1. RESPONSE STRUCTURE
   Does the site open each page with a direct answer to the main question,
   or does it start with generic introductory text?
   Are H2 headings phrased as real visitor questions
   (e.g. "How do I get to X?" vs "Getting there")?

2. DATA CITABILITY
   Are there specific opening times, prices, distances, phone numbers, addresses?
   Does each piece of data have a visible source or update date?
   Are there FAQs with questions and answers clearly separated?

3. CROSS-CHANNEL CONSISTENCY
   Is the core information (hours, contacts, description) consistent
   across the website, Google Business Profile and social channels?
   Look for any discrepancies that an AI model might read
   as a reliability signal.

4. GAPS RELATIVE TO AEO
   List the 5 types of content that are missing or poorly structured
   relative to Answer Engine Optimisation best practices.
   Reference: Princeton/Georgia Tech/IIT Delhi GEO paper (ACM SIGKDD 2024) —
   adding statistics with a cited source increases citability by 41%.

5. ACTION PLAN
   5 priority actions, ranked by estimated impact, with an indication
   of who can implement them (internal team / developer / agency)
   and estimated implementation time.

Use only data actually present on the site.
If a section is not accessible, tell me.

Recommended full workflow:

1. Run this prompt → get the technical content audit
2. Open ChatGPT and Perplexity → search "[DESTINATION] things to do in summer" in your primary market's language → see whether you appear
3. Compare the gap between what ChatGPT cites and what your site offers
4. For systematic monitoring → rankwit.ai

Important caveat: this is only a very first starting point. The GEO/AEO topic is "probabilistic" not "deterministic": each LLM has different citation criteria, which change over time and vary by language, market and query type. What works