Google Retires Gemini's Gems, Bets on 'Skills' Instead

Google is winding down Gems, the customizable mini-chatbots inside Gemini, and replacing them with a new feature it's calling "skills." The move marks a notable shift in how the company thinks about AI agents, trading purpose-built bots for a more modular system that plugs specialized capabilities directly into a single assistant.

Gems let users build their own task-specific versions of Gemini, complete with custom instructions, uploaded files, and a defined persona, similar in spirit to OpenAI's custom GPTs. Someone could spin up a Gem to act as a coding tutor, a résumé editor, or a brainstorming partner, then return to that same configured bot whenever the task came up again. It was Google's answer to a broader trend of letting everyday users assemble their own narrow AI helpers without writing code.

Skills take a different approach. Rather than spawning a separate, standalone agent for each task, Google is baking specialized capabilities into the core Gemini experience so the assistant can call on them as needed within a single, continuous conversation. Instead of hopping between different Gems for different jobs, users would interact with one Gemini that flexibly draws on whichever skill fits the moment.

The change lines up with where the rest of the industry appears to be heading. Meta recently introduced Muse and Instinct, all-in-one agents designed to handle a sweeping range of tasks without requiring users to configure separate bots for each one. OpenAI, Anthropic, and others have also been pushing toward more capable general-purpose assistants that lean on tools and extensions rather than proliferating standalone personas. Building a library of narrow, user-configured bots increasingly looks like a transitional step rather than the destination, as companies bet that a single, more capable agent that knows when to reach for a specific capability is the more durable product.

For Gemini users who invested time building out custom Gems, the transition raises practical questions about whether existing configurations will carry over into the new skills framework or need to be rebuilt from scratch. Google has not detailed a full timeline for retiring Gems or specified exactly how skills will be created, shared, or discovered, but the direction is clear: fewer standalone bots, more capabilities folded into one assistant.

The shift also reflects a broader rethink of what "agentic" AI should look like to mainstream users. Early experiments across the industry, including Gems and custom GPTs, asked people to essentially become bot designers, picking instructions and toggling settings to get useful, repeatable results. Skills suggest Google is betting that most users would rather have one smart assistant that adapts on the fly than a shelf of custom-built ones they have to remember to open.