Google has officially shut down Project Mariner, its experimental web-browsing agent, on May 4th, 2026. The Verge AI reports that the project’s landing page now carries a farewell note: “Thank you for using Project Mariner. It was shut down on May 4th, 2026 and its technology voyaged to other Google products.” The shutdown was first surfaced by Wired’s Maxwell Zeff, and according to The Verge AI, Google hasn’t responded to a request for comment.
This isn’t a quiet retreat from agentic browsing. It’s a consolidation play, and the timing tells the story.
What Project Mariner was
Google introduced Project Mariner in December 2024 as a research preview that could navigate the web and complete tasks on a user’s behalf. Last year, Google bumped its capabilities so it could juggle up to 10 tasks at once. Think of it as the lab where Google figured out how to make an AI actually click buttons, fill forms, and move between pages without a human babysitting every step.
That lab work didn’t get thrown away. It got distributed.
Where the technology went
Over the past year, Google quietly migrated Mariner’s capabilities into its broader product stack. The Verge AI notes the tech now powers:
- Gemini Agent: handles tasks like archiving emails or booking hotels.
- AI Mode: Google’s agentic search experience, which leans on the same browsing intelligence.
- Auto-browse in Chrome: a newer feature that runs multi-step jobs like comparing flight prices, though Google hasn’t explicitly said it’s Mariner under the hood.
The pattern is clear. Mariner was the prototype. Gemini Agent, AI Mode, and auto-browse are the production lines.
Why this matters
The agentic browsing race is heating up fast. OpenAI, Perplexity, and OpenClaw have all rolled out tools that let an AI take the wheel on the web. Google needs its agent story to stop looking like a research demo and start looking like a shipping product. Closing Mariner and routing its DNA into consumer-facing surfaces is how you do that.
There’s also a strategic angle worth flagging. Keeping a separate “experimental” brand alongside Gemini muddies the message. By killing Mariner, Google narrows the pitch to a single agent identity. Everything agentic now lives under Gemini or directly inside Chrome and Search. Cleaner story, easier to sell.
The I/O angle
Google I/O kicks off May 19th. Sunsetting Mariner two weeks before the event isn’t an accident. The Verge AI suggests Google may be clearing the deck for new AI features set to debut on stage. Expect a more polished agentic story: tighter integration with Gemini, deeper Chrome capabilities, and a unified pitch that doesn’t require explaining what “Project Mariner” was supposed to be.
What practitioners should take away
If you’ve been building on or testing Mariner directly, the migration path is already paved. Gemini Agent and AI Mode pick up the workload. The bigger lesson is structural: experimental AI products at the big labs have shorter half-lives than ever. The capability survives. The brand often doesn’t.
This is the same playbook OpenAI ran when it moved features from labs into ChatGPT proper, and what Anthropic does when research previews get folded into Claude. Watch for the capability, not the codename.
What’s next
The interesting question isn’t whether Google will keep building agents. It will. The question is whether agentic browsing becomes a default behavior in Chrome, the way password autofill or translation did. If I/O ships an auto-browse feature that’s on by default for hundreds of millions of Chrome users, the competitive math shifts hard. OpenAI and Perplexity have to fight for installs. Google just flips a switch.
More details on the shutdown and the migration are in the original report from The Verge AI.