Google Kills Gemini's Gems and Migrates Them to 'Skills'

Google will shut down Gems, the Gemini feature that let users build custom AI assistants for specific tasks, and automatically migrate them into a new format called "skills" starting on November 17, 2026.

The company disclosed the change in a message inside the Gemini app, which warns users that Gems will become skills on that date. Google said it will handle the migration itself, so users will not need to take any action to make the transition. Until November 17, 2026, existing Gems remain usable. The work users invested in building their Gems will not be destroyed; it will carry over into the new skills format.

9to5Google first reported the wind-down over the weekend.

Google launched Gems in 2024. The idea was simple: users could teach their AI to perform certain tasks without repeating the instructions every time. Google shipped a set of pre-made Gems, including a learning coach, a brainstorming assistant, a career guide, a coding partner, and an editor. Users could also build their own — a running coach, a nutritionist, a vacation planner. Gems could be shared with others, a feature Google hoped would make the Gemini app more popular.

Why the shutdown matters

The timing explains the decision. All-in-one AI agents like Meta's Muse and Instinct are taking off, and they work differently from the custom-assistant model Google bet on with Gems. Users type what they want into a chatbot, and the system figures out the rest. That is a lower barrier than asking someone to configure, name, and manage a bespoke assistant for each task.

The shift from Gems to skills is a bet on which interface wins. Google is consolidating bespoke assistants into a system that operates across different AI tasks rather than standing alongside them. It is also an implicit admission that the custom-assistant framing did not gain the traction needed to justify its own brand, icon, and placement in the app's navigation.

The interface problem

Even after the migration, the former Gems will not be as consumer-friendly as typing plain text into a chatbot like Meta's Muse. Google notes that users will have to enter a forward slash — "/" — in a task thread to select the skill they want to use. That is a user interface that engineers, not regular people, tend to prefer.

That detail may prove more consequential than the branding change itself. Slash commands are second nature to developers and power users, who have used them in tools from Slack to code editors for years. But Google is competing for a mass audience against Meta's agent products, which ask nothing of the user beyond a sentence of intent. If selecting a skill requires learning a command syntax, Google has traded one friction for another.

A familiar pattern for Google

The Gems shutdown is another example of a long-standing Google failing: giving every new feature its own brand name, icon, and prominent navigation placement, then shuffling things around and merging one feature into another over time. This predates the AI era. At one point, the company operated multiple messaging and communication apps simultaneously.

For users, the churn carries a real cost. Anyone who invested time building and sharing Gems now has to track a migration, learn a new invocation method, and adjust to a renamed feature. Google's promise that the transition will be automatic softens the blow, but it does not change the pattern that produced it.

For developers and enthusiasts who built audiences around shareable Gems, the question is what sharing looks like under the skills model. Google has not, in the announcement described here, detailed how shared Gems will behave once converted.

What happens next

The deadline is November 17, 2026. Existing Gems keep working until then. After that, they exist only as skills, invoked through slash commands inside task threads, and the Gems name joins the long list of Google product brands that were folded into something else.

The stakes go beyond one feature. Google is restructuring Gemini around task-oriented skills just as Meta pushes conversational agents like Muse and Instinct. Whichever interface ordinary users actually adopt — plain conversation or structured skill selection — will shape how the next generation of AI assistants gets built.