Google’s Quiet September Surprise
On September 24th, Google unveiled an MCP server for their Data Commons. You’re forgiven for being too deeply engrossed in the announcements of Taylor Swift’s new album release or the Baywatch reboot to notice, but once the noise quiets down, you’re going to want to think about its implications for travel. By the way, MCP stands for Model Context Protocol and it’s been described as the equivalent of a USB-C plug for agents, meaning that an AI agent can use it to easily access the tools and capabilities it needs without having to program a direct (API) connection. It also has another superpower in that it can help the agent find what it needs, sort of like a concierge at a hotel. Here’s what we have, and here’s how to get it. Agents love MCP.
But first, you say, what is this Data Commons thing you’re talking about? And why should I care? Data Commons is a commerce-neutral “knowledge graph” of public data. It includes things like census stats, climate, health, education, safety, transit, economic data, and many more compiled statistics. It’s designed to be neutral because it’s aggregating from verifiable public sources, not commercial inventory. The reason you might be interested is because it can be really helpful to GenAI agents in creating highly personalized trip plans. The MCP server technology makes it far simpler for a buyer agent to understand the data available and to access it.
How a Travel ‘Agent’ Uses It: A Real Example
Unless you’ve worked with Data Commons before, an example might be the best way to describe it. Let’s say you ask your buyer-side agent (e.g., ChatGPT Agent, Perplexity Comet, Gemini Agent, or some other app that uses an agent to do travel search for you) to find you a reasonably priced four-day family weekend, someplace warm in the second half of October, with lots of cultural attractions, hiking and cycling activities, and preferably within 5 flight hours of the Atlanta airport.
The agent first has to turn that into some kind of measurable criteria for a first-cut search. For example:
- Time window: the target weekend, possibly flexible ±1 day.
- Warm: daily highs between, say, 72–88°F, low chance of rain, acceptable humidity, decent air quality.
- Culture: museums and cultural venues per capita, festival density, UNESCO sites nearby, performing arts calendars.
- Outdoors: trail density and average trail rating, bike lane miles per capita, elevation profiles, bike share presence.
- Family fit: safety, walkability, transit, kid-friendly venues.
- Trip practicality: flight duration and frequency from your home airport, on-time performance, general (but no shopping here—yet) price ranges.
Your agent then hits the Data Commons MCP server to fetch comparable, trusted public data to build a list of candidate destinations:
- Climate normals and recent anomalies to validate “warm” for the specific weekend.
- Air quality, safety, walkability, transit reach to score family comfort and ease.
- Population and tourism seasonality to avoid crush periods or recommend out of the way destinations.
- Education/culture proxies the graph already aggregates (museum counts, venues, universities as proxies for cultural density), with citations.
From Data to Discovery: How It Changes Trip Planning
The output is a ranked short-list of destinations that actually meet the weather and “family-friendly culture + outdoors” brief, with sources attached to cut hallucinations. Reductions in hallucinations are a big feature of Google’s announcement because of the high quality, and neutrality, of the data. Nobody is paying for placement here!
The Data Commons won’t have everything in, so your agent has to look for additional data sources:
- Trails and cycling: pull trail density and difficulty from Open Street Map-based or parks datasets, bike lane mileage, bike share coverage.
- What’s happening: events calendars for your dates (performing arts, festivals, exhibits).
- Logistics: average flight time and frequency from your home airport, historic delay rates, typical fares.
Finally, the agent blends in what it knows about you already, to create something more personalized. This is not part of Data Commons access per se, it builds on it from the previous steps.
- Loyalty status, preferred airlines and hotel brands.
- Kids’ ages and (family) tolerance for hikes, museum time, and transfers.
- Budget preferences and room (e.g., 1 vs. 2) needs.
- Dietary, mobility, or other restrictions, etc. where relevant.
- A collective memory of previous trips you’ve taken, things you liked (or didn’t), what you did when you got there, the kinds of places you stayed, etc.
Based on all of the above, the agent is ready to surface recommendations for you. Note that this is just in the discovery phase of travel: shopping and booking are separate, and they happen after you’ve chosen the alternatives that you find the most attractive. In the example above, it might surface destinations like Santa Fe, NM; Austin, TX; San Diego, CA; Asheville, NC; or Charleston, SC.
That’s three flight destinations and two that are more likely drive destinations. It might also put the list in a table showing how each rates on the various components of your request.
This is radically different from traditional OTA experiences where you’d need to:
- Pick a destination first, then discover it doesn’t have what you want
- Open multiple tabs to cross-reference weather data, cultural calendars, and activity options
- Manually compare flight times and connections from your home airport
- Start over when you realize your first choice is too expensive or booked up
Instead, you get a curated shortlist of places that already match your criteria, backed by verified data rather than marketing copy. Now you’re ready to select a few and dive into the details of transportation, climate, hotels, activities, etc.
Let’s pause for a beat to consider that searches of the data commons hit way up in the funnel as part of the research/exploratory first cut, and quickly crystallize what might take hours to tens of hours of research with conventional web search. Based on your specific criteria, it might surface destinations you’d never thought of…but are exactly what you want. (Can you hear the over-traveled destinations breathing a collective sigh of relief?) It saves you tons of time AND brings you to cool new places!
Why This Matters to the Travel Industry
Could OTAs and metasearch access the Data Commons search? Yes, and they’re likely already working on it. What’s different here is that by attaching an MCP server to Data Commons, it’s now much easier for anyone with a buyer-agent to do it as well. It’s one more thing that’s no longer the purview of the travel titans, and it’ll allow your buyer agent to more fully customize the search to your specific needs.
Think about how this differs from the way an OTA or even a supplier works today: the first thing you see on their website is: “where do you want to go?” Using a GenAI agent to do your vacation planning doesn’t require you to know your destination up front, or visit any websites at all. Like a skilled and familiar executive assistant, it searches and evaluates options on your behalf and presents you with a curated set of alternatives…while you get a cup of coffee and scroll through Instagram looking for puppy videos.
But what if the OTA or Metasearch offers an AI search engine of their own? We’ll, it won’t know that you really love infinity pools or that your kids have allergies and can’t go to certain kinds of locations or that you texted your friend last month about how great the vacation scuba dive you recently took was. All of these things—and many, many more—can be part of the personalization your agent offers you when selecting a destination because it knows you and your family. And just because an OTA has a GenAI model that can select destinations and subsequently offer you hotels and flights, there’s no guarantee it’ll offer you what’s best for you rather than the options with the highest commission profit potential. Like they do today.
I hope at this point you can see that easy access to Data Commons by buyer agents can have a big impact on the way in which travelers can find the most attractive destinations and set the stage for personalized recommendations for their itineraries. With the introduction of the Google’s new MCP server, it’s also going to eliminate any reliance on travel titans for getting those great recommendations. It might not rise to the level of that Baywatch reboot, but it just might reboot some more adventurous vacations.
