AI-powered travel planning is everywhere right now. Booking.com, Google, and every major travel brand is racing to embed artificial intelligence into search, recommendations, and trip planning. Ask an AI where to stay, what to book, or how to plan a weekend, and you get polished, confident answers in seconds.
At first glance, this looks like a breakthrough for travelers.
But there’s a growing problem hiding behind the excitement: one we’ve been thinking about a lot as we build tools for fitness travelers.
AI Is Only as Good as the Data It Receives
Here’s the thing about AI: it doesn’t actually know anything. It processes what’s already out there.
And when it comes to hotel gyms, what’s already out there is fundamentally broken.
AI models don’t visit hotels. They don’t step into fitness rooms. They don’t check whether the dumbbells go up to 10 kg (22 lbs) or 45 kg (100 lbs). They don’t know if the squat rack exists or if it was removed three years ago.
They reuse what already exists online:
- Marketing photos taken years ago
- Images designed to look attractive, not informative
- Descriptions written for sales, not accuracy
- User reviews written from wildly different fitness expectations
From an AI perspective, all of this looks like valid input. From a fitness traveler’s perspective, it often leads to disappointment.
Why Hotel Gym Information Is Uniquely Unreliable
I’ve stayed in dozens of hotels over the past few years, and one pattern keeps repeating: the gym photos rarely match reality.
Hotel gyms suffer from a problem most other hotel features don’t. Photographers are incentivized to make gyms look appealing, not useful. A scenic treadmill view looks great in a listing. A clear photo of the dumbbell rack, weight range, or machine layout doesn’t get the same attention.
The result:
- Mirrors make gyms look larger than they are
- Old equipment stays visible long after upgrades or removals
- Critical details like max dumbbell weight, presence of barbells, and functional training space are missing entirely
- Cardio equipment is overrepresented, strength training is under-documented
AI can’t fix this. It can only repeat it.
When Opinions Aren’t Enough
Customer reviews are valuable. But they’re subjective by nature. A guest who jogs twice a week and a guest who trains seriously can both describe the same gym as “good” or “bad” and mean completely different things.
AI systems treat these opinions as signals of quality. But without structured, objective criteria, they can’t distinguish:
- Adequate from excellent
- Decorative from functional
- Marketing claims from reality
This isn’t a failure of AI. It’s a failure of the underlying data.
Why Major Platforms Can’t Solve This Alone
Booking.com is exceptional at scale, availability, pricing, and logistics. That’s their strength.
But they don’t measure gyms. They don’t audit fitness facilities. They don’t define what makes a gym suitable for different training needs.
Any AI feature they deploy will inevitably rely on hotel-provided descriptions, uploaded photos, and general guest reviews. That produces impressions, not assessments.
And impressions aren’t enough when fitness is a deciding factor in where you stay.
If you want to understand how this data problem plays out on the hotel side, not just the traveler side, read our piece Your Hotel Gym Is Losing You Guests. The Data Proves It — the numbers make the cost of vague gym data concrete.
We Tested the MCPs. Here’s What We Found.
We didn’t want to make these claims on theory alone. In May 2026, we ran a direct test using the MCP (Model Context Protocol) connectors that Booking.com, Expedia, and Tripadvisor now expose to AI systems like Claude. These are the live data feeds that AI travel tools pull from when generating recommendations. We used Caesars Palace in Las Vegas as our test case, a hotel with a well-documented fitness facility.
Here is what each platform returned.
Booking.com returned property details, pricing, and a flat list of facility tags. “Fitness center” appeared as a single text label in a list of 60+ amenities. No images were returned at all. No equipment detail. No photo of the gym. Just the tag.
Expedia returned a geofencing error and no data. Their MCP connector is blocked for users outside the United States. An AI tool used outside this country, gets nothing from Expedia’s API at all.
Tripadvisor returned the most data by a significant margin. The property came back with an amenity confirmed as “Fitness Center with Gym / Workout Room,” plus approximately 50 traveler photos and 10 recent reviews. Tripadvisor’s AI-generated review summary, drawn from over 33,000 guest reviews, made no mention of the gym at all. Of the 10 reviews returned, none mentioned fitness, equipment, or the gym facility.
The photos were the most telling result. Tripadvisor returned full-resolution images with dynamic URL templates, meaning the images themselves are accessible and usable. But not one photo carried a category label. There was no way to identify which images showed the gym without running a vision model across all 50 photos individually.
That is exactly the problem in practice. The data infrastructure exists. The image URLs are there. But the structured layer that would let an AI answer “what does the gym at Caesars Palace actually look like, and what equipment does it have?” does not exist anywhere in these MCPs.
A “Fitness center” tag and a pool of uncategorized photos is not actionable information for a fitness traveler. It is the same vague signal that has frustrated travelers for years, now served through a faster pipe.
What We’re Building Instead
This is exactly why we started working on structured gym data. Not based on vibes, adjectives, or star ratings, but on criteria that actually matter for training:
- Equipment variety and quality
- Strength vs cardio balance
- Functional training capability
- Space and usability
- Consistency across locations
We use AI extensively in our process. But we don’t trust it blindly. Our GymFactor is a complex system requiring many different data points to produce a serious rating.
Images get checked for age and relevance. Descriptions are reviewed to avoid misleading information. Visual tricks like mirrors and wide-angle lenses are accounted for. When data is unclear, it gets flagged and reviewed with the hotel directly.
This hybrid approach is slower than scraping and summarizing. But accuracy matters more than speed when trust is at stake.
The Bigger Picture: AI Needs Domain-Specific Foundations
Here’s what worries us beyond just hotel gyms.
If AI struggles this much with fitness facilities, something relatively straightforward to evaluate, what about other specialized travel needs?
- Family-friendly hotels: is that pool actually suitable for toddlers?
- Accessibility features: does “wheelchair accessible” mean the entire property or just one room?
- Pet policies: what does “pet-friendly” really include?
- Work-friendly spaces: reliable wifi vs a “business center” claim
In each case, the same problem applies. If the underlying data is vague, inconsistent, or marketing-focused, AI won’t magically have better information. The MCP (Model Context Protocol) test confirmed it at the infrastructure level: the APIs reflect exactly the data that already exists, nothing more.
AI tools will continue to improve. But without structured, verified data in niche domains, they’ll keep failing travelers who care about specific details.
Moving Forward
We’re not anti-AI. We’re building for a future where AI can actually help fitness travelers, by creating the data foundation it needs to give accurate answers.
Our browser extension GYMR is a first step toward making fitness infrastructure visible, comparable, and honest. But the real work is in building structured data that doesn’t exist yet anywhere in the travel stack, as our MCP test confirms.
AI doesn’t replace domain expertise. It amplifies it, but only if the data exists. And we need your help. Add your review, share your experience, and help others stay fit while they travel.
Together, we’re building that data.
