Home / Apps

Chrome Gets an Agent: Gemini 'Auto Browse' Clicks Through the Web for You

Google answers ChatGPT Atlas and Perplexity Comet by building multi-step browsing automation straight into the browser used by roughly seven in ten people.

AppsJan 28, 20264 min read2 diagrams

Key takeaways

  • Auto browse launched in preview on January 28, 2026 for U.S. Google AI Pro and AI Ultra subscribers on desktop Chrome.
  • It handles multi-step chores in a side panel, logs its actions, and has a daily limit on agentic actions.
  • Chrome held about 69% of worldwide browser share in August 2026, giving Google a distribution edge over AI-native rivals.

Google has put an AI agent inside Chrome. On January 28, 2026, the company began rolling out "auto browse," a Gemini 3-powered feature that carries out multi-step tasks on websites on your behalf, initially as a preview for U.S. subscribers to Google's AI Pro and AI Ultra plans.

What auto browse does

The feature runs in Chrome's side panel on desktop, covering Windows, macOS and Chromebook Plus. You describe a chore and Gemini works through it using the page's contents and the model's multimodal abilities. Google's examples included finding products that match a photo, staying within a budget, and applying discount codes. Coverage of the launch notes that the agent keeps a running record of what it does in the side panel and logs activity to Gemini Apps Activity and Chrome history, and that there is a daily limit on how many agentic actions a user can take. A later report adds that some sensitive actions require the user's confirmation.

Basic Gemini in Chrome, meaning the side-panel assistant that summarizes tabs and answers questions about a page, remains available to ordinary Chrome users. The agent is the part that sits behind a paid plan.

Worldwide browser share, August 2026 Worldwide browser share, August 2026 (%) Chrome 69.4 Safari 15.8 Edge 5.4 Firefox 3.0 Samsung 2.0
Statcounter figures for August 2026. AI-native browsers such as Atlas and Comet are not broken out. Source: Statcounter.

The challengers Google is answering

The move is a response to a new crop of AI-first browsers. OpenAI launched ChatGPT Atlas, a Chromium-based browser with ChatGPT built in, for macOS in October 2025, with an agent mode for Plus, Business and Pro users. Perplexity's Comet is a free Chromium browser with built-in AI search and automation, and a dedicated Android app followed in 2026. Reviewers have been mixed on how reliable these agents are, calling the approach novel but underdeveloped.

Google's advantage is distribution. Chrome already holds roughly 69% of worldwide browser use, so any agent feature it ships reaches more people than a rival must win over with a download. The trade-off is that Chrome is a mature product with a vast extension and enterprise footprint, so changes have to land cautiously. One reviewer expects Chrome to remain the most used option but to trail AI-native browsers on bleeding-edge agent behavior.

How an auto browse task runs in Chrome 1. You ask in the side panel 2. Gemini 3 plans reads the page 3. Acts on sites clicks, fills, compares 4. You confirm sensitive steps Every step logged to the panel, Gemini Apps Activity and Chrome history
Simplified flow of an agentic task based on Google's launch description. Illustrative.

Why it matters

Browser agents change who does the clicking, and that disrupts the economics of the web. Sites built around ad impressions, comparison traffic and checkout funnels tuned for people now face software that skips banners and compares prices in seconds. Merchants may welcome agents that finish a purchase, while publishers worry about agents that read pages without sending a human visitor.

There is also a security dimension. A tool allowed to click and fill forms can be steered by hostile text hidden in a page, a risk researchers have raised for agentic browsers generally. Google's activity logs and confirmation prompts are an early attempt at guardrails, but a daily action limit and a preview label signal that the company is moving carefully.

The subscription angle

Putting the agent behind AI Pro and AI Ultra is a business decision as much as a technical one. Agentic actions are costly to run, since each step involves model calls over a page's contents, and a daily limit keeps that cost bounded. It also ties Chrome to Google's paid Gemini plans, giving the company a new reason to push subscriptions at the same time its other Google services are bundling AI features. Free users still get the summarizing and question-answering sidebar, which keeps Gemini in front of the vast Chrome audience without the expense of running agents for everyone.

For the AI-native rivals, the arrival of a comparable feature in the default browser narrows the gap that justified downloading something new. OpenAI and Perplexity can still compete on speed, openness and how deeply the assistant is woven into the interface, but Google's reach means expectations for what a browser can do will be set by Chrome.

What to watch

  • Expansion beyond U.S. AI Pro and Ultra subscribers, including Android, where Gemini in Chrome later reached U.S. users with auto browse for subscribers.
  • Whether Google moves auto browse into the free tier or keeps it as a paid upsell.
  • How websites respond: new bot policies, agent-friendly checkout flows or blocking.
  • Independent testing of reliability and prompt-injection resistance against Atlas and Comet.

Sources

  1. Chrome Gemini auto browse — 9to5Google
  2. Google releases Chrome auto browse for AI Pro and Ultra users — TestingCatalog
  3. Google integrates Gemini 3 into Chrome, adds agentic AI features — Business Today
  4. Browser market share worldwide — Statcounter
  5. AI shakes up competition in the internet search market — The Irish Times

More in Apps