Google's Gemini Spark AI agent uses personal data to plan trips, raising privacy concerns

Google's new AI agent Gemini Spark uses extensive personal data from Gmail, Google Calendar, Photos, and location history to generate detailed, personalized itineraries, according to a hands-on test by The Verge. The agent is rolling out to Google's $99/month AI Ultra subscription plan starting June 2026.

What Gemini Spark does

Spark is designed as an "always-on AI agent" that can use external apps and operate computers on behalf of users. In testing, the agent successfully:

• Organized Gmail inbox cleanup suggestions with unsubscribe links
• Scanned Google Docs for incomplete tasks
• Generated a comprehensive trip itinerary using unprompted personal information

For a family trip to Hershey, Pennsylvania, Spark created a multi-thousand-word itinerary that included:

• Driving directions from the user's home address (not provided in the prompt)
• The user's dog's name (Frida), sourced from veterinary emails
• Children's names and ages, noting one child's free admission eligibility
• The user's wife's dietary restrictions (no onions or scallions)
• Concert tickets from Ticketmaster confirmation emails
• Correct nap time scheduling

The agent also drafted and sent a Google Doc to the user's wife using her email address, which was not explicitly provided.

Technical limitations

Spark failed when attempting to book an Airbnb directly. After requesting permission to interact with websites, the agent navigated to Airbnb but was blocked by the platform's security policies. "Due to security and authentication policies on Airbnb, I am unable to log in, handle payment, or complete bookings directly on your behalf," Spark reported.

Data sourcing

The agent's capabilities rely on Google's Personal Intelligence feature, which aggregates data across:

• Gmail
• Google Calendar
• Google Photos
• Search history
• Location data

Google did not disclose the specific AI models powering Spark or provide technical specifications such as context window size or benchmark performance.

What this means

Spark represents Google's most aggressive deployment of personal data for AI assistance. The agent's ability to surface unprompted personal details—from pet names to children's ages—demonstrates the direct relationship between data access and AI utility. While Google's existing data trove gives it an advantage over competitors like OpenAI and Anthropic, who lack similar access to users' personal information, the implementation raises questions about data mining practices and user consent. Third-party platforms appear to be implementing blocks against automated AI agents, potentially limiting the practical utility of tools like Spark for completing transactions across the web.