Hey Siri, Take Me to Santorini
Animation by Sora
TL;DR
Experts believe Search is at risk of becoming the 21st century yellow pages. The theory is that Google Ads, for 25 years the beating heart of travel customer acquisition, is risking irrelevance as generative AI reshapes how travellers discover and plan. But—as you’d expect—Google has a plan.
Agentic AI is the next act: bots that don’t just suggest but reason, decide, and act. The travel industry, with its tangled mix of logistics, preferences, and human emotion, is ground zero for this transformation. But the thing is, it doesn’t quite work—yet.
Investors are backing AI-native startups. Lots of them. BMPD shows the future now—proactive, problem-solving, invisible AI. But most agents still overpromise and underdeliver.
For travel brands, the choice is stark: build and govern your AI, or risk becoming invisible in someone else’s ecosystem. The agentic era has arrived.
From Party Trick to Power Tool: The Shift to Agentic AI
Generative AI’s first act was flashy but shallow—chatbots that wowed us with eloquence but lacked substance. Ask a question, get a grammatically correct answer, even if it was dead wrong. We called them assistants, but they couldn’t really assist. They mimicked knowledge, but didn’t reason. It was parroting with a diploma.
Act Two changes that. Today’s frontier models pause, reflect, and compute. They can chain steps together, apply logic, cross-reference constraints, and adjust their approach in real time. They can also act. They have ‘agency’ hence the name. That capability—reasoning at inference time and acting—is what turns a chatbot into an agent. It’s the difference between “find me hotels in Tokyo” and “plan me a week-long cultural trip to Japan with a mix of city and countryside, some time for hiking, and I hate big group tours—oh, and make the bookings for me.”
Travel is the perfect test case. It’s multi-dimensional, highly personal, emotionally charged, and financially significant. It’s also chaotic—prone to last-minute changes, cross-border variables, and messy data. The AI that can navigate that minefield with grace becomes more than a tool. It becomes a trusted advisor. It is also an industry where cooperation across the supply chain is both historical and essential, and where margins are often squeezed, making any new efficiencies attractive.
Is This the Death of Search?
For over two decades Google Ads has been the plumbing behind the travel industry’s demand engine. Want bookings? Bid on keywords. Need scale? Buy more clicks.
A debate is raging over whether the Google Ads model is broken. Protractors argue that generative AI is inserting itself between customer and click, and that travel search click-through rates are down as much as 70%.
Organic traffic isn’t safe either. If AI can summarise twenty web pages into one answer, the value of being on page one plummets. Instead of paid or optimised results, visibility now comes from being quotable, credible, and structured. That means brands must publish structured data, speak naturally, and design for AI digestion—not for search spiders.
And then there’s the shift in behaviour. With users now giving full-blown prompts: “I want to spend ten days in Fiji in November, somewhere quiet but not boring, near good snorkelling and a spa, budget around $4,000,” AI agents should theoretically be able to deliver tailored answers in order to win the traveller’s trust—and get the booking. A fine theory indeed—but with the proof being in the pudding, there are still lots of proof points missing.
For brands still addicted to Google Ads, this could well be the cliff. If you're not quoted in an AI response, you may as well not exist. In a world of ambient discovery and AI-first planning, the demand funnel has moved—upstream, conversational, and beyond keyword control.
Article continues below.
Like this article? Join the thousands of tech founders, board members and investors who subscribe to our free monthly newsletter, Tech Round-Up. Sign up below!
The Empire Strikes Back
In May 2025 Google released a suite of initiatives—‘AI Mode’ (infusing a ChatGPT-like search into the traditional search page) and ‘Google Shop with AI Mode’ (allowing users to discover and purchase products directly from Google Search—no need to visit a retailer’s site). Powered by Gemini and Google’s Shopping Graph, Shop with AI Mode handles product discovery, comparisons, price alerts, and even checkout via Google Pay.
This innovation this has potential to blow apart the e-commerce scene by disintermediating retailers and e-commerce platforms. It collapses the customer journey into a single Google interface, with Google’s AI assistant assuming the central role in purchase decisions and customer relationships.
The effect? Google moves from gateway to commerce to storefront. The implications for the travel industry are profound, to say the least.
So, Who’s Building What in Agentic AI?
Let’s start with Booking.com. Their AI Trip Planner, live in the US, UK, and now Australia and New Zealand, allows you to have a conversation with your travel app. Want boutique hotels with rooftop bars and canal views in Amsterdam? Just say so. It’ll fetch suggestions, refine based on feedback, and guide you to a shortlist—all without menus or filters.
This isn’t a gimmick. CTO Rob Francis says their long game is to evolve from a booking engine to a travel partner—helping users not just transact, but dream, explore, and personalise from the first spark of inspiration to the final boarding pass.
Expedia’s doing the same with an AI chat layer, integrated with itinerary planning. Hopper leans on predictive algorithms to forecast airfares and now offers fintech-style products like Price Freeze and flight disruption protection. Even OpenAI has joined the fray, with plugins for Kayak, Expedia, and others.
Each of these implementations aims to be the bridge between search and action—making it easier to not just find, but book, with AI steering the process.
Show Me the Money: VCs Go All-In
It’s not just Big Tech flexing here. Venture capital is pouring into AI-native travel startups at a pace we haven’t seen since the first OTA boom.
Berlin-based Tourlane raised €25m from Sequoia to build AI-generated itineraries from natural language prompts. A user says, “adventure honeymoon in Patagonia next March, some hiking, some hot tubs,” and Tourlane’s AI builds options, scores permutations, and adapts to changes.
Corporate travel platform Navan, formerly TripActions, which has raised over $1.5bn, has integrated GPT-4 into every part of its platform. Ava, its virtual assistant, doesn’t just book travel within corporate policies; it pulls receipts, analyses spend, and answers finance questions instantly.
TravelPerk, another corporate player, has leveraged AI to automate post-booking changes, cancellations, and rescheduling—raising $200m on the back of a 75% gross margin leap, credited to AI, with humans now focussing on edge cases, while machines handle the boring 80%.
Meanwhile, Nuitée is tackling the murky world of hotel inventory, bundling fragmented supplier data into a clean API. It raised $48m to power AI agents behind the scenes. Picks and shovels for the AI gold rush.
BMPD: More Than Just a Bot
While most AI travel tools are still learning to walk, founder-funded agentic AI tool BMPD is sprinting.
Developed by US startup Rebook, BMPD (short for “bumped”) lives in the background, scanning your upcoming flights for disruption risk. It draws from a vast set of data sources—weather forecasts, historical airline ops, airport congestion, aircraft rotations, even crew logistics—to model the likelihood that your flight will be delayed or cancelled.
And it doesn’t just warn you. It acts.
If BMPD predicts a high risk of disruption (with claimed accuracy of 94–99% within 48 hours), it proposes alternatives: other flights you can actually catch given your location, booking rules, and travel profile. Crucially, it doesn’t suggest flights you’d miss. It computes travel time, check-in requirements, gate proximity—then serves realistic options. No fantasy fixes.
Then comes the magic: one-click rebooking. No call centres. No chatbot queues. No waiting behind 200 angry travellers. Just you, your phone, and a proactive solution delivered quietly, before the chaos begins.
This is agentic AI at its most useful. BMPD doesn’t try to be everything. It picks one painful travel moment—disruption—and solves it better than any human ever could. It’s what Siri wants to be when she grows up.
Currently in US beta, Rebook plans to expand BMPD internationally. If it does, expect frequent flyers and travel managers to cheer. It’s not just sizzle. It’s steak—and it’s medium rare, ready to serve.
But Wait, Who Owns the Customer?
All of this innovation raises one big question: who owns the traveller?
In the old world, airlines, hotels, and OTAs all fought to ‘own’ the customer relationship. Now, AI agents are inserting themselves as gatekeepers. If Apple builds a brilliant AI travel agent into Siri (no matter how unlikely that seems today!), or Amazon into Alexa, or Google into Maps, customers may never return to individual airline or hotel apps.
Whoever controls the interface controls the relationship—and possibly the booking.
This creates a new disintermediation threat. If you don’t build your own agent, you risk becoming a line item in someone else’s. Suppliers that once feared OTAs now face the possibility of being commoditised by voice assistants.
The solution? Build or integrate deeply. Either your brand’s AI talks directly to the customer, or it talks to someone else’s AI—and gives up the relationship in the process.
Reality Check: Limitations and Legal Landmines
The reality today is that most AI travel agents are helpful...but hardly omniscient.
They’re good at narrow tasks: suggesting hotels, watching prices, pinging you about changes. But they stumble with complexity. Ask one to plan a round-the-world trip with airline points and specific routing, and you’ll likely be back to Excel by lunch.
Personalisation? Still mostly manual. If you like boutique properties and loathe chain hotels, you have to tell them—every time. AI doesn’t yet remember your style, your loyalty accounts, or your past trips in any meaningful way.
Worse, when things go wrong, AI agents show their immaturity. The Air Canada chatbot fiasco—where a bot misled a customer about bereavement fares and the airline tried to dodge liability—ended with the court holding the company responsible. That precedent matters. If your AI misguides a customer, it’s on you.
More generally, as a recent swing through San Francisco and Silicon Valley has revealed, big tech’s agentic bender is following the familiar move-fast-and-break-things motif, but it’s moving much faster than ever.
And the industry’s belief that all the broken things can be fixed later feels a little like hope over expedience.
There is still a huge gap between the promise of agentic ecosystems—billions of tiny bots independently acting and cooperating seamlessly and harmlessly, and reality. There remain huge holes in the agentic scaffolding—regulation and standards, for instance. There’s a lot of excitement about a protocol from anthropic called MCP, but cyber security purists have concerns, and even Anthropic’s partners fear worry about the business case implications of letting LLM providers write the road rules.
Then there’s the immaturity of orchestration, observability, and evaluation tools. Even at a fundamental level, there is currently no equivalent to the Internet’s domain name registry (which is how we find websites just by typing a URL into a browser).
Then, of course, there’s the small matter of trust.
Building Trust: Governance, Compliance, and the Human Touch
As agentic AI grows more powerful, governance becomes critical. Explainability, auditability, and accountability aren’t just nice-to-haves. They’re legal necessities.
Companies will need policy frameworks for AI: “Don’t auto-book connections under one hour”; “Don’t recommend hotels without verified safety ratings.” Gartner projects that AI governance platforms will become mandatory by 2028. Those who adopt early will earn trust—and avoid lawsuits.
Human-in-the-loop systems will be essential. Today’s best agents still allow override or supervision. The brands that combine AI efficiency with human empathy will win in the long run. Think co-pilots, not autopilots.
From the Front Line
No one quite has the answer. Opinions diverge wildly, depending on which camp you’re in. Read these and decide for yourself!
The Search Expert
“Search is changing—quickly. I would not be surprised if you find zero traditional ‘blue links’ on Google by the end of 2025. The ‘messy middle’ of travel will become more opaque, and attribution will become even more difficult than it is today. Our industry is accelerating towards an inflection point and the playbook is changing. Established players will suffer as agile, new entrants claim market share. This is a moment where brands can prioritize a pivot in their customer acquisition strategy -- or fade away.”—Brennen Bliss, Propellic
The Agentic AI Founder
“Most travel AI tools are still in nappies—we built BMPD to wear running shoes. It doesn’t just predict disruption; it fixes it before you even know there’s a problem. Real intelligence isn’t chatty, it’s quiet, proactive, and gets you on the next flight while everyone else is still queuing.”—Timothy Kasbe, Founder of Rebook
The OTA
Agentic commerce isn’t just another tech trend—it’s the new battleground for trust and attention. The winners won’t be the ones with the most inventory, but the one with the most unique content delivered with the most intelligence and trust. That means differentiated content with deep personalisation across the purchasing journey, and bots that don’t just chat—they act. The bot is always going to need a brand that a consumer trusts, so the brands that build that stack will own the traveller relationships. Everyone else risks becoming background noise in someone else’s interface, that is, irrelevant.—Katrina Barry, Managing Director, Webjet
“Sure, we’re in a fight. But in the words of Mark Twain, ‘rumours of my death are greatly exaggerated’. We’re systematically integrating AI search into Search. The collective impact of LLM searches is only 2-3% of our total search volumes, and revenues remain strong.” (paraphrased).
The Guru
“Travel founders should pay attention. This shift won’t stop with consumer goods. The same dynamic will hit tours, activities, hotels, and flights; anything transactional that can be indexed and surfaced by an agent. Once Google owns both discovery and conversion, the value shifts from branded experiences to structured data and AI visibility. What matters is how clean, comprehensive, and machine-readable your product feed is. The Google Merchant Center may soon be your new homepage.”—Mauricio Prieto, Travel Tech Essentialist
The Publisher
“The search engine now wants to position itself as a travel concierge that could outmaneuver OTAs and comparison sites. It does this by automating complex tasks and delivering hyper-personalized results. That, at least, is the aim... The fate of this new AI Mode, though, depends heavily on whether users come to trust it, how smoothly it operates in practice, and, perhaps significantly, whether Google can break free from its past struggles with launching successful new projects.”—TourismReview
Final Boarding Call: This Is Not a Drill
The age of agentic AI is upon us. Not in a sci-fi way, but in a real, operational, increasingly commercial sense. The early wins—BMPD’s disruption handling, Booking’s conversational planning, Navan’s admin automation—are showing what’s possible. Even if they can’t yet do the complex stuff, the bots are coming.
The next few years will define who leads, who adapts, and who disappears from view. Those who build intelligently, govern ethically, and act decisively will thrive. The rest could just be flattened by something from left field.
And if your AI does get you to Santorini? Make sure it booked the hotel with the infinity pool—and not during school holidays. We’ll see you on the beach. Hopefully with your luggage.