Google is retiring one of Gemini’s most recognizable customization features as the company moves toward a more agent-oriented approach to artificial intelligence, folding Gems into a broader “skills” system rather than maintaining them as standalone AI assistants.
The change will take effect on November 17, 2026, according to a notice displayed in the Gemini app. Google said users will not need to recreate their custom assistants because existing Gems will be automatically migrated into skills. The current Gems will remain available until the transition.
The move comes as the consumer AI market increasingly shifts from chatbots that respond to individual prompts toward AI agents that can perform multi-step tasks with less user direction. Meta has been developing products such as Muse and Instinct around that model, raising the competitive pressure on Google to make Gemini more capable as an integrated assistant rather than a collection of separately branded features.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Gems were introduced in 2024 as a way for Gemini users to create customized assistants with persistent instructions for specific jobs. Google also offered pre-built versions, including a learning coach, brainstorming assistant, career guide, coding partner and editor.
Users could create their own versions for tasks such as fitness coaching, nutrition planning and travel. Gems could also be shared with other users, giving Google a potential mechanism for expanding Gemini through user-created AI experiences.
That model is now being absorbed into the broader skills framework.
The change suggests Google is moving away from treating customized AI personalities or assistants as individual products and toward treating them as reusable capabilities that can be invoked inside different tasks.
A standalone Gem effectively required the user to decide which assistant they wanted to use. A skill can instead function as a capability that is called when it is relevant to a particular task.
Google’s decision to migrate Gems rather than simply discontinue them also indicates that the company wants to preserve the underlying user-created instructions while changing how those instructions are accessed.
For existing users, the practical benefit is that their work should not disappear.
Therefore, the bigger issue is discoverability.
Google says users will be able to invoke skills by entering a forward slash, “/”, in a task thread and selecting the relevant skill. That creates a more structured interface, but it also introduces an additional step between the user and the AI capability.
For mainstream consumers accustomed to simply typing what they want into a chatbot, that may feel less intuitive.
The Agent Race is Changing The Interface
The change comes as the competitive focus in consumer AI is moving toward agents that can handle tasks rather than simply generate responses. A traditional chatbot requires the user to formulate a request, interpret the answer, and decide what to do next. An agent can potentially handle a larger sequence of actions, using specialized capabilities along the way.
That makes the underlying architecture more important than the branding of individual assistants.
Google’s skills approach appears designed to support that model. Instead of creating a separate interface for every specialized assistant, Google can maintain a common Gemini environment in which different capabilities are invoked as needed.
But there is a trade-off.
The simpler the underlying architecture becomes, the less visible individual features may be to users. Gems gave users a clear concept: create an assistant, give it instructions, and use it for a particular purpose. Skills are more abstract.
That could make the system more powerful while making it harder for consumers to understand what they can actually do with it.
Google’s Recurring Problem With AI Branding
The retirement of Gems also highlights a broader issue with Google’s product strategy.
Google has frequently introduced new products and features under distinct names, only to later merge, rename, or discontinue them. The company’s history with messaging and communication services is an earlier example of the difficulty of maintaining a coherent product architecture when features and services evolve rapidly.
AI has accelerated that problem.
Google has introduced a large number of Gemini-related products and capabilities as it attempts to compete across consumer applications, search, productivity software, and enterprise services.
The risk is that users end up having to understand Google’s internal product structure before they can take advantage of the technology.
Meta’s approach with products such as Muse represents a different direction. The user does not necessarily need to understand which underlying capability is being activated. The emphasis is on describing the desired outcome and allowing the AI system to determine how to accomplish it.
That approach is expected to be increasingly used as AI agents move into everyday consumer applications.
However, the end of Gems does not necessarily mean the end of the functionality users created with them. By automatically converting Gems into skills, Google is preserving the underlying customization while changing the interface through which it is accessed.
That could ultimately make the feature more useful if skills become deeply integrated across Gemini’s different task environments. The challenge is making that power accessible without forcing users to learn a new workflow.
However, Google currently appears to be betting that a common skills architecture will serve Gemini better than maintaining a separate ecosystem of custom assistants.