355
Audio & Video Production344
Automation & Workflow224
Software Development250
Marketing & Growth192
AI Infrastructure & MLOps173
Writing & Content Creation203
Data & Analytics140
Design & Creative169
Customer Support130
Photography & Imaging156
Sales & Outreach125
Voice & Speech135
Operations & Admin87
Education & Learning131
Google is testing Gemini Spark, a new tool inside Gemini that can read your Gmail and Calendar, draft emails, and try to complete online tasks, with user approval.
In short: Google is rolling out a beta version of Gemini Spark, a new “always on” assistant inside Gemini that can use your personal Google data to plan and carry out tasks.
Google introduced Gemini Spark at its I/O developer conference and began rolling it out this week in beta. It is available to subscribers of Google’s AI Ultra plan, which starts at $100 a month. Spark appears as a new tab inside the Gemini chatbot and works on both desktop and mobile, including iPhone.
Gemini Spark is designed to connect to a user’s personal data, like Gmail, Google Docs, and Google Calendar. Instead of typing “prompts,” users give Spark “tasks,” which are basically instructions. Google says Spark can draft emails and create calendar events, but it asks for approval before sending messages.
In a hands-on test reported by Wired, Spark scanned a user’s emails and calendar and found an existing karaoke reservation for a birthday party. It then produced a detailed multi page plan, including a suggested guest list, nearby restaurants, and draft invitation emails. The test also showed limits, since Spark tried and failed to book a dinner reservation online, even after triggering a verification code sent by text.
Google is positioning Spark as its answer to “always running” AI agents like OpenClaw, which became popular in early 2026 but also caused public mishaps for some users.
Tools like Spark can feel like a personal assistant who can read your notes and messages (like handing someone your planner and inbox). That can save time, but it also raises privacy and security risks. Google warns about “prompt injection attacks,” which is when a bad instruction tricks the agent into doing something harmful, like leaking private email details.
Source: Wired