In brief
- Gmail’s new “AI Inbox” uses Gemini 3 to prioritize messages and surface summaries instead of a simple chronological feed.
- Writing and creative tools expand, with AI drafting now available to all users and fast image generation added to Workspace.
- Google stresses privacy safeguards as deeper AI integration raises concerns over how personal inbox data is used.
Google unveiled a sweeping overhaul of its flagship email service today, integrating its most advanced artificial intelligence model, Gemini 3, directly into Gmail.
The move marks the company’s most aggressive attempt yet to transform the inbox from a passive repository of messages into an active personal assistant, escalating its rivalry with OpenAI and Microsoft.
The updates, announced alongside new features for the Gemini app, leverage the company’s latest large language model to automate daily tasks, sort communications by priority, and generate creative visual content.
The AI-powered inbox
At the center of the update is a redesigned Gmail interface powered by Gemini 3, the next-generation model Google introduced late last year. The new “AI Inbox” view, available starting today, departs from the traditional chronological list. Instead, it uses on-device processing to organize emails into priority clusters and offers a “Catch me up” summary of recent activity, such as shipping notifications, appointment reminders, and purchase receipts.
“This is us delivering on Gmail proactively having your back,” said Blake Barnes, a Google VP of product, in a statement Thursday.
The company positioned the update as a shift toward a “thought partner” capable of answering complex queries about a user’s digital life, such as “When does my flight land?”—all without requiring manual searches.
The update also expands the “Help Me Write” tool. Previously a premium feature, the AI drafting capability is now rolling out to all users with enhanced tone-matching abilities, allowing the software to mimic a user’s personal writing style more improved accuracy.
The rollout started Thursday for users in the United States. While some features (like email thread summaries and “Help Me Write”) are free for everyone, the most powerful “Assistant” features—specifically the ability to ask questions across your inbox (e.g., “What size shoes did I order?”)—are currently restricted to paid subscribers using Google AI Pro or Ultra. The new “AI Inbox” view (which sorts mail by priority) is still limited to a group of “trusted testers,” not the general public.
Gmail holds roughly 30% of the global email client market share, typically ranking second behind Apple Mail—which has a higher share because it is the default app on iPhones, often used to access Gmail accounts. Most major industry reports and data suggest the user base has stabilized around 1.8 billion, though some recent estimates project it crossing the 2 billion mark in 2025–2026.
‘Nano Banana’ and creative tools
Beyond productivity, Google is pushing its capabilities in creative media.The announcement highlighted the integration of its efficient image-generation model, codenamed “Nano Banana” (officially Gemini 2.5 Flash Image), into the broader Gemini ecosystem.
Originally previewed in late 2025, the model is designed for high-speed, on-device image creation and editing. Google confirmed today that the model’s “Pro” variant is now available to enterprise customers and is powering new creative tools across Google Workspace and the Gemini mobile app, allowing users to generate or remix images with natural language prompts.
Privacy and market stakes
The rollout comes as Google faces intensifying competition. The launch of Gemini 3 and its integration into core products like Gmail and Search has prompted competitors like OpenAI to accelerate their own product roadmaps.
However, the deep integration of AI into personal email raises persistent privacy questions. Google emphasized on Thursday that while Gemini 3 processes inbox data to provide summaries and answers, that data remains within a secure “engineering privacy” barrier. The company stated explicitly that user content from these personal workspace tools is not used to train its public AI models.
To further address privacy concerns, Google also highlighted the broad rollout of “Temporary Chat” in the Gemini app. Functioning similarly to an “incognito mode,” this feature allows users to conduct one-off AI conversations that are not saved to their history or used for model training, a feature aimed at users wary of their data leaving a trace.
Generally Intelligent Newsletter
A weekly AI journey narrated by Gen, a generative AI model.
Source: https://decrypt.co/353987/google-just-overhauled-gmail-gemini-3-ai-assistant
