The Real AI Bottleneck in Travel Content Is Copyright, Not Hallucination

From the Travel Tech Podcast newsletter for episode 24, The Real AI Bottleneck in Travel Content Is Copyright, Not Hallucination.

There's a moment in this conversation where my guest says, flatly, "we were selling masks." He's not talking about PPE. He's talking about how his company kept an entire travel content business alive during COVID by tracking constantly-shifting entry restrictions for Delta, United, and Iberia, in near real time. If you're one of the people who read this because you work in aviation, travel, or AI, I wanted that story on the show.

He started as a C-level executive running a 500-person division of a coach and travel agency business, watched that company collapse, and then spent 13 years building something of his own from scratch. What began as a destination-events content company pivoted twice, first into COVID restriction tracking during the pandemic, and now into a broader "traveler support" platform covering visas, health rules, accessibility, and pet travel for airlines, cruise lines, and airports. Apple Podcasts, also streaming on Spotify and YouTube.

The tension underneath all of it: AI has made destination content free, but the parts of travel that actually can't afford to be wrong, vaccination rules, visa forms, image licensing, are exactly the parts AI still struggles to get right.

The gap between a free answer and a correct one

That tension has a structural cause. Cruise lines can have itineraries with over a thousand possible ports, and the order you visit them in can change what vaccinations you legally need, not just the destinations themselves. That's not a content problem you can solve with a general-purpose AI query; it's an operational data problem that has to be rebuilt for cruises, airlines, and airports separately.

AI is adding a specific new pressure here, and it cuts both ways. On one hand, generic destination content, "what to do in Madrid with kids," is now something anyone can generate for free, which is exactly why Smartvel stopped competing on it. On the other hand, AI image generation still can't reliably produce a real, licensable photo of anywhere that isn't already a globally famous landmark, which means companies still have to buy and manage image rights manually for the long tail. And there's a second shoe still to drop: AI token costs are currently heavily subsidized, and when that subsidy ends, the economics of AI-generated travel content change again.

Íñigo Valenzuela, founder of Smartvel (smartvel.com), walks through all of this: the COVID pivot, cruise line vaccination logic, the travel pets problem, image licensing, and why he thinks personalized dynamic pricing is the next uncomfortable shift for travelers.

Key Takeaways

• Order-dependent health rules: On multi-port cruise itineraries, the sequence of countries visited can trigger additional vaccination requirements beyond what any single destination would require.
• Upper-funnel repositioning: Smartvel deliberately stopped competing on point-of-need destination Q&A and moved its business toward inspiration and conversion tooling for airline and OTA marketing campaigns.
• The image licensing bottleneck: Commercial use of destination photos requires purchased IP rights, and AI-generated images remain unreliable for any location that isn't extremely well-known.
• Distribution beats hallucination as the real AI travel risk: Valenzuela argues new AI-native trip planners face a harder problem in customer acquisition against Booking and Expedia than in content quality.
• Personalized pricing is coming: Valenzuela expects airlines and OTAs to increasingly price based on inferred ability to pay, similar to postal-code-based insurance pricing, which could mean loyal customers see higher prices, not lower ones.

🔧 Two to Build

Íñigo Valenzuela's argument lands here: the hard part of travel AI is not generating content, it is getting trustworthy, legally clean source data in the first place.

Build a computer vision pipeline that detects changes on government travel pages the moment they go live.

Smartvel crawled official travel restriction pages hourly during COVID, then used document comparison to flag any change for a human editor. The system did not summarise or interpret. It detected a diff, surfaced it, and let a person decide what it meant. For airline content teams, that separation of machine detection from human judgement is what made it defensible and publishable.

Catch: optical character recognition on government PDFs and scanned notices will fail you more often than clean HTML will, so budget for a messy ingestion layer before you build anything downstream.

→ Computer vision for regulatory monitoring

Build a compliance monitoring system that maps regulatory text to a knowledge graph and alerts teams when the graph changes.

Natural language processing can parse a new regulation and link its clauses to existing rules, routes, or product features stored as nodes and edges. When a new document contradicts or extends an existing node, the system raises an alert rather than waiting for a manual review cycle. This is the infrastructure layer that makes AI generated content trustworthy, because it answers the question of whether the source itself has changed.

Catch: the graph is only as good as your entity resolution. Two documents using different names for the same route, carrier, or restriction type will silently create duplicate nodes and miss the conflict entirely.

→ Automated regulatory compliance monitoring

Alex Brooker, Host of the Travel Tech Podcast, Founder of Airside Labs