Over the past four years, numerous product demonstrations have touted AI's potential for trip planning. The promise has been straightforward: simply state your destination, and a chatbot or virtual assistant will meticulously explore travel options, gather local attractions, and create a comprehensive itinerary. However, in practice, I've found this capability to be somewhat limited—it typically caters to the most common tourist activities in any given city.
My recent experience with Spark, Google's innovative always-on AI assistant, showcased a different level of functionality. Spark represents an ambitious leap for Google, as it aims to serve as a central hub for using external applications and, in the future, could even control your computer. Currently available through Google's $99 monthly AI Ultra plan, I was fortunate to get an early test drive. I began with basic tasks, such as having Spark sift through my Gmail to identify emails to unsubscribe from and combing through Google Docs for outstanding tasks. In both instances, Spark delivered effectively, even organizing a document with links for quick unsubscribing from promotional emails.
I then tasked Spark with planning a weekend trip. I informed it, “I’ll be in Hershey, PA, with my wife, two kids, and our dog over the weekend of July 18th. Can you put together a plan that covers accommodations, dining, activities, and more?” I omitted important details, like concert tickets for that Saturday evening, hoping to kick-start the process with a few of Hershey's most recognized attractions.
A few moments later, Spark responded with a detailed reply, providing a comprehensive family and dog-friendly itinerary for our stay in Hershey, from July 17 to July 19, 2026. It included a link to a Google Document filled with extensive and practical travel suggestions.
This was the most personalized trip plan generated by an AI that I've experienced to date. Surprisingly, Spark included driving directions from my home, even though I hadn’t disclosed my address. It suggested a range of pet-friendly hotels and activities suitable for my dog, Frida—a fact I had not shared, likely gathered from communications with my veterinary clinic.
Spark also highlighted that my son Lewis would enter Hershey Park for free since he is under one year old, while noting that my three-year-old Arthur would need a ticket. Remarkably, it even accounted for nap time at 1:30 PM, seemingly based on an educated guess about my children's routines.
The itinerary provided an abundance of personalized details. It included my wife's preferences—specifically her aversion to onions and scallions—and referenced the Thomas Rhett and Niall Horan concert we planned to attend on Saturday night, inferred from Ticketmaster confirmations present in my email. Upon reaching the point about securing a babysitter for the night, I remembered that my parents would be joining us for that purpose, adding a note in our conversation.
Spark responded enthusiastically, addressing my parents by name, and adjusted its recommendations from a hotel to an Airbnb based on this new information. When I requested Spark to compile all this into a Google Document and share it with my wife, it effortlessly located her email, attached the document, and even crafted a note that resembled a correspondence between colleagues rather than a couple.
The only hiccup occurred when I asked Spark to book an Airbnb directly. It prompted me for permission to interact with web pages on my behalf, went to the Airbnb website, but encountered a blockade. Spark informed me that it could not log in or process payments due to security restrictions on Airbnb, while still providing a list of available accommodations for the appropriate dates and reminding me of the necessary booking details.
Overall, this experience marked one of the most remarkable demonstrations of AI capabilities I’ve encountered. Google’s strength in AI, coupled with the wealth of personal data it has through its Personal Intelligence feature, enabled the creation of a highly relevant and tailored itinerary. It compiled the information with a level of detail resembling an actual human assistant, considering our family's unique needs and preferences. Each time I review the plan, I’m impressed by the intricate details, suggesting we will likely follow it closely.
Yet, amidst this excitement is an unsettling sense of invasion. While Spark’s efficiency feels almost magical, it also raises privacy concerns. It’s disconcerting to see an AI addressing personal details such as my children's names and ages, as well as information it seemingly accessed without explicit permission. Intellectually, I understand that Google possesses vast amounts of my data—my emails, calendar entries, photos, and search history paint a comprehensive picture of me. However, the use of this information feels less like a safeguard and more like an intrusion.
This scenario exemplifies the current dilemma: the correlation between how much information we willingly share with AI systems and the level of utility we receive in return. Google holds a unique advantage due to its comprehensive data collection, while competitors like OpenAI and Anthropic are striving to gather similar insights. The most powerful AI tools will be those that can understand us intimately, make decisions in our stead, and act without our direct involvement. However, that comes at a cost of self-disclosure.
The old adage “If you're not paying for it, you’re the product” takes on new significance in the context of AI. In this case, we are indeed paying—our lives, correspondence, and images are the resources and products constantly being processed and reconfigured. Some outcomes may be astonishingly beneficial, but they require this level of personal trade-off. While I anticipate an enjoyable weekend in Hershey this summer, I cannot shake the feeling of being monitored, ostensibly for my own benefit.
