Case Study: How a Bali Boutique Hotel Went from Invisible to AI-Recommended in 60 Days

When the owner of a 24-room boutique eco-lodge in Ubud, Bali reached out to palmtree.ai in early 2026, the situation was frustrating but familiar: a beautifully curated property with five-star guest reviews, yet completely invisible to every major AI travel engine. Eighty-five percent of their bookings came through Booking.com at a 22% commission rate. Not a single guest had ever arrived through an AI recommendation.

Sixty days later, the property was being actively recommended by ChatGPT, Perplexity, and Google Gemini. Direct bookings had increased 340%. Monthly OTA commission savings hit $4,200.

This is the story of how we made it happen.

The Challenge

The property (we’ll call it “Sawah Eco Lodge” to protect client identity) had been operating for six years. Nestled among rice terraces just outside central Ubud, it offered everything the modern conscious traveler wants: sustainable bamboo architecture, farm-to-table dining, daily yoga sessions, and immersive Balinese cultural experiences.

On paper, the lodge should have been a natural fit for AI recommendations. When travelers ask ChatGPT for “sustainable boutique hotels in Bali” or “best eco-lodge near Ubud,” Sawah Eco Lodge should appear. But it didn’t.

The owner had invested heavily in traditional SEO and maintained an active OTA presence. Google organic traffic was decent, social media following was modest but engaged. Yet the property remained locked in the OTA dependency cycle: paying 22% commissions on 85% of all bookings while watching revenue erode year after year.

The fundamental problem wasn’t visibility in general. It was visibility to AI.

What We Found

When we ran our initial AI Travel Score audit, the results told a clear story.

Overall AI Travel Score: 12/100

Here’s the breakdown across all five major AI engines:

Zero mentions across all five engines. When we prompted each AI with queries like “best eco-lodge in Ubud,” “sustainable hotel Bali,” and “boutique hotel near rice terraces Ubud,” Sawah Eco Lodge appeared in none of the responses.

The 12 points came from minimal schema markup already on the site and a handful of third-party mentions that AI models could theoretically reference but weren’t prioritizing.

Root Cause Analysis

We identified five critical gaps:

1. No llms.txt file. The property had no machine-readable summary optimized for AI crawlers. Without an llms.txt file, AI engines had to piece together fragmented web data instead of receiving a structured, authoritative summary of the property.

2. Thin schema markup. Basic Hotel schema existed, but it was missing critical fields: amenities, sustainability certifications, geographic coordinates, check-in procedures, pricing ranges, and guest review aggregation.

3. Zero GEO content. The blog section had three outdated posts from 2023. No content was structured for Generative Engine Optimization, meaning nothing on the site answered the types of questions AI engines process when generating travel recommendations.

4. Minimal social signals. Instagram was active but unoptimized. No presence on Pinterest, TikTok, or X. Google Business Profile was incomplete and had unanswered reviews from months prior.

5. No entity authority. The property lacked the kind of cross-platform entity consistency that AI engines use to build confidence in recommendations. Business name variations existed across platforms, and there was no unified knowledge graph presence.

The Strategy

We designed a 60-day implementation plan with three parallel workstreams: technical foundation, content engine, and social amplification.

Phase 1: Technical Foundation (Days 1-14)

llms.txt Implementation

We created a comprehensive llms.txt file that gave AI crawlers everything they needed in a single, structured document: property overview, unique selling points, room categories with pricing, sustainability credentials, location context, guest experience details, and booking information.

The file was deployed at the root domain and structured following the latest llms.txt specification, with clear markdown formatting optimized for language model parsing.

Schema Markup Overhaul

We rebuilt the schema from scratch using LodgingBusiness as the primary type, nested with:

• amenityFeature arrays covering all 47 amenities
• geo coordinates pinpointing the exact location relative to Ubud landmarks
• aggregateRating pulling live review scores
• hasOfferCatalog with room types and seasonal pricing ranges
• sustainabilityPolicy (a newer schema property) documenting their eco-certifications
• containsPlace linking to on-site restaurant, yoga shala, and spa

Google Business Profile Optimization

We completed every available field, uploaded 85 new high-quality photos with AI-optimized alt text, responded to all 43 unanswered reviews (both positive and negative), and established a weekly posting cadence.

Phase 2: Content Engine (Days 7-50)

24 GEO Blog Articles

We produced 24 articles specifically engineered for AI retrieval. These weren’t traditional SEO blog posts optimized for keyword rankings. They were structured to answer the exact questions AI engines process when generating travel recommendations.

Sample titles from the content calendar:

• “What Makes Ubud Different from Other Bali Destinations for Eco-Conscious Travelers”
• “A Complete Guide to Sustainable Accommodation Options Near Ubud’s Rice Terraces”
• “Farm-to-Table Dining in Ubud: What to Expect at Boutique Eco-Lodges”
• “How to Choose Between Ubud’s Boutique Hotels: Price, Sustainability, and Experience Compared”
• “Cultural Immersion Programs at Bali’s Boutique Properties: What’s Worth It”

Each article was 1,500-2,500 words, heavy on specific details, comparisons, and practical traveler guidance. Every article included proper FAQ schema, internal linking, and entity mentions that reinforced the property’s authority in its niche.

Review Response Automation

We implemented an automated review monitoring and response system across Google, TripAdvisor, and Booking.com. Every new review received a personalized, on-brand response within 4 hours. Positive reviews were amplified with additional context about the experiences mentioned. Negative reviews received empathetic, solution-oriented responses.

This wasn’t just customer service. Review velocity and response patterns are signals that AI engines factor into recommendation confidence.

Phase 3: Social Amplification (Days 14-60)

Daily Multi-Platform Posting

We launched a daily posting schedule across Instagram, Pinterest, TikTok, and X. Content was designed to create the type of cross-platform entity signals that AI engines use to validate property quality and relevance.

Instagram focused on visual storytelling with detailed, keyword-rich captions. Pinterest targeted aspirational travel planning boards. TikTok showcased authentic guest experiences and behind-the-scenes sustainability practices. X shared industry insights and travel tips that positioned the property as a thought leader in sustainable Bali tourism.

Every post linked back to relevant GEO content on the property website, creating a reinforcing loop between social signals and on-site authority.

The Results

After 60 days, the transformation was measurable across every metric.

AI Travel Score: 12 → 71/100

Booking Impact

The direct booking increase didn’t just save commission costs. Direct bookers also spent 26% more on average, likely because they arrived with higher intent and better understanding of the property’s premium positioning, exactly what the GEO content communicated.

Content Performance

• 24 GEO articles published, 19 indexed by AI engines within 45 days
• Google Business Profile views increased 180%
• Social media following grew 2,400 across all platforms
• Review response rate went from 12% to 100%
• Average review score improved from 4.6 to 4.8 (driven by response quality, not manipulation)

Financial Summary

The return on investment in the first 60 days exceeded the cost of the entire engagement by a factor of 4.2x.

Key Takeaways

1. AI visibility is a technical problem, not a marketing problem. Sawah Eco Lodge had great reviews, beautiful photos, and loyal guests. What they lacked was the technical infrastructure (llms.txt, schema, entity consistency) that AI engines need to confidently recommend a property.

2. GEO content works differently than SEO content. Traditional blog posts optimized for Google keyword rankings don’t necessarily perform well for AI retrieval. The 24 articles we created were designed to answer conversational queries with specific, authoritative, detail-rich responses, the format AI engines prefer.

3. The OTA commission savings compound fast. A $4,200 monthly savings sounds modest for a 24-room property. But combined with higher direct booking values and the trajectory of increasing AI visibility, the annual impact exceeds $150K. That’s transformative for a boutique operation.

4. Social signals reinforce AI recommendations. Cross-platform entity consistency and active social engagement create the validation signals AI engines use to build recommendation confidence. Understanding how ChatGPT recommends hotels reveals that social proof and entity authority are major ranking factors.

Frequently Asked Questions

How long does it take for a hotel to appear in ChatGPT recommendations?

Based on our experience with boutique properties, initial AI visibility typically appears within 30-45 days of implementing a comprehensive GEO strategy. However, consistent mention across all five major AI engines usually takes 60-90 days. The timeline depends on factors like existing web authority, review volume, and the competitiveness of the destination.

Can a small boutique hotel compete with large chain hotels for AI visibility?

Yes, and in many cases small properties have an advantage. AI engines prioritize relevance and specificity over brand size. When a traveler asks for “sustainable eco-lodge in Ubud,” AI engines prefer recommending a property that perfectly matches that description over a generic chain hotel. The key is making sure AI engines can find and understand your property’s unique positioning through proper technical setup and GEO content.

What is an AI Travel Score and how is it calculated?

The AI Travel Score is palmtree.ai’s proprietary metric that measures how visible and favorably a travel business appears across the five major AI engines (ChatGPT, Perplexity, Google Gemini, Claude, and Microsoft Copilot). It evaluates factors including mention frequency, recommendation context, source attribution, entity accuracy, and sentiment. Scores range from 0-100, with most travel businesses currently scoring below 20.

Ready to find out your property’s AI Travel Score? Get your free AI visibility audit at palmtree.ai and discover exactly where you stand across all five AI engines, plus a custom roadmap to start capturing direct bookings from AI-powered travel planning.