Most people who use AI tools for recurring tasks run into the same problem. A chatbot does not remember what it did last time. A user who needs Gemini to write in a specific tone, follow a particular research process, or stick to a coding framework has to explain all of that again at the start of every new conversation.
Google released a feature two years ago specifically built to solve that. It became one of the most relied-on parts of Gemini for users with repeating workflows. Now Google is pulling it and replacing it with something new, and users have until November 17 to figure out what that means for them.
What Gemini Gems are and why users rely on them
Google introduced Gems in 2024 to give Gemini a permanent set of custom instructions. A user could configure a Gem once and return to it any time that task came up, without repeating themselves.
The use cases ranged widely. One user might build a Gem for drafting emails in a specific voice. Another might set one up for coding in a particular framework. Others used them for research workflows, tutoring routines, or document formatting. A Gem could also be given a default tool, such as Google’s image-creation feature or Canvas, and files could be attached to provide Gemini with additional context.
More Google:
- Alphabet’s biggest AI fear may be fading
- Bank of America says Alphabet stock investors are missing the bigger signal
- Google stock price faces major AI test ahead of earnings
Gems were shareable through links. A team could distribute one Gem configuration to multiple members and run the same setup across different accounts. Users who organized their Gemini work around specific Gems had little reason to look elsewhere.
The feature is available to users on Google AI Pro and Google AI Ultra subscriptions. Those plans give access to Gemini’s more advanced capabilities, and Gems became a central reason many users chose to stay subscribed.
What skills are and how they differ
Google is replacing Gems with a new feature called skills. The underlying idea is the same. Users define custom instructions, and Gemini follows them. The change is in how those instructions are accessed and used.
Accessing a saved Gem requires opening the side panel in Gemini, locating the Gems manager, and selecting the correct configuration. Skills work differently. Users type a slash in the prompt box, and the skill activates directly inside the conversation.
Typing a slash in the prompt box opens the skills menu. Selecting a skill applies it to the current conversation. More than one skill can be active in the same session. That is not an option with the current Gems setup.
That last part is significant. A user who needed both a research workflow and a writing configuration previously had to choose one Gem per session. Skills allow both to be active together. For users managing complex or multi-step tasks, this could change how they work with Gemini.
When the migration happens
Google has set November 17, 2026, as the date it will begin automatically migrating existing Gems to skills. Users can continue using their current Gems until that date.
Skills are already available through the Gemini Spark tab for subscribers on Google AI Pro or Google AI Ultra. Users can create a skill there now and access it from the main conversation using the slash command.
When 9to5Google first reported the change, Google’s help article for skills and the “Create skills” button inside the Gems manager were not yet working. The migration notice appears to have surfaced before all the supporting tools and documentation were ready.
What Google has not answered yet
The announcement raises questions that Google has not fully addressed.
Gems supports file attachments, default tool assignments, and shareable links. None of those features is confirmed to carry over automatically when a Gem migrates to a skill. A user who attached reference documents to a Gem or shared a Gem configuration with colleagues does not yet know whether that setup will survive intact.
Google’s help documentation for skills was not live when the change was first reported. Until that material is available, users who built detailed Gem configurations are working without confirmation of what the migration will preserve.
For casual users, the shift may be straightforward. For users who built Gems around specific files or shared them across a team, those are not small details.
How does this fit Google’s push to make Gemini more useful
Google has been working to position Gemini as more than a chatbot. The move from Gems to skills is part of that effort.
A basic AI chat tool answers one prompt at a time. Customized configurations let users set rules that carry across sessions. Skills go a step further by making those configurations available on demand inside the normal chat experience and by allowing several configurations to work together at once.
OpenAI has taken a similar approach with custom GPTs, which let users build and share personalized versions of ChatGPT. The competition for users who rely on repetitive workflows is a real part of how AI companies differentiate their products.
Google is betting that moving custom instructions into the prompt itself, rather than a separate manager, will make them more practical for everyday use. The November migration will test whether that bet plays out, and whether the transition is smooth enough that users do not lose the configurations they spent time building.
Related: Google set to challenge Meta Muse in the Agentic AI race