Google is launching the ability to share Gemini “Gems,” or special custom AI assistants designed for a single task. The feature enables people to hand off a Gem — complete with its instructions, tone, and examples — to co-workers, friends, or family members the way they would share a file in Google Drive. It’s an incremental UI change with significant implications: reusable assistants mean less duplication, greater consistency, and quicker onboarding for all of those who may one day adopt AI at work or at home.
How Gem sharing works across Google Workspace
Gems are written commands that define how Gemini should react. You could define a “brand voice editor,” a “Socratic study coach,” or a “Python pair programmer,” and include sample prompts and files to direct actions. To share one, you open the Gem manager on the web, click the Share button, and select permissions that should be familiar to most Google Drive users — who can view a Gem and who can edit it.
Google had earlier launched Gems for Gemini Advanced, Business, and Enterprise subscribers in over 150 countries before making them available for everyone else to build and use, including with file uploads. Now, sharing enables the custom assistants to be portable. The receiving user sees the same setup, too, so you don’t have to work backwards from a set of instructions or guess what prompt recipes were employed.
Nothing is for sale here; Google is framing this as link-based or access-controlled sharing, not a public directory. That fits with the way teams now share templates and playbooks within Drive — private by default, public when helpful.
Why Gem sharing matters for teams and individuals
Templatizing how an assistant behaves is usually the difference between “pretty good” and “production-ready.” Shared Gems cut down on prompt drift — “the way we word something can have a critical difference in the outcomes.” For teams, it’s one canonical research brief writer, one coding helper who adheres to those same lint rules, or a sales email coach that enforces tone and compliance.
Imagine a marketing team that constructs a Gem pre-trained on brand guidelines, approved phrases, and product FAQs. When it is published and shared across regions, every copy of it that a writer has access to will have the same constraints. A support team could write a triage Gem that parses uploaded policy PDFs and generates recommended actions with links to internal docs. Families can share a travel-planning Gem set to keep track of spending and how much each person owes, or a weekly meal schedule bound by dietary rules from an uploaded spreadsheet.
Analysts have long identified reuse and standardization as critical for AI adoption at scale. Big industry research published by the likes of Stanford’s AI Index and a bunch of consulting firms points to the role that template-based workflows play in moving pilots into everyday use, because they minimize variance and training overhead. Shared Gems provide that with Gemini as the natural place to reuse such files.
Enterprise governance, security and admin angles
Since sharing mirrors Drive, it complements established collaboration habits inside Google Workspace. That’s important for admins who like predictable permission models. “Can use” can be the setting for most people on teams, while “can edit” is reserved for a few maintainers who control updates. Let’s have some version discipline here: freeze editing on canonical Gems and document changes so that users know what to expect.
Data processing must stay at the center. Also remember that the files and context you include might contain sensitive information, even if it’s a shared Gem. Companies should provide clear guidance about what can be embedded (public docs, scrubbed examples) and what should reside in a secure environment. These are all safe defaults that provide least-privilege sharing, as well as internal-only access to these locations (and your friends’ files), and they should be combined with periodic permission reviews.
For regulated teams, this also exposes a positive: standardized Gems can bake in compliance language, safe responses, and escalation rules — reducing the risk of ad hoc prompting off-piste policy-wise.
How it compares to OpenAI GPTs and Microsoft Copilot
OpenAI made custom assistants popular with GPTs, then let us share and publish them, complete with an app store–like experience. Microsoft provides Copilot Studio for creating and sharing custom copilots with other individuals in the organization. Google’s approach is somewhere in between, emphasizing private, Drive-like sharing without the public marketplace; a design that could woo enterprises wary of app stores but looking to standardize their internal tools.
Feature-wise, the competitive battlefield is less about raw model power and more about having excellent tooling: permissions, templates, analytics, connectors, and lifecycle management. Shared Gems make Gemini a stronger player in that tooling race.
What to watch next for shared Gemini Gems features
And here are some awesome additions that could make shared Gems even more valuable:
- Version labels
- Usage analytics
- Bulk deployment to groups
- Curated template libraries by function (support, finance, engineering)
More comprehensive integrations — for example, governed access to Drive folders, Sheets, or third-party knowledge bases — could help harden these repeatable workflows.
For now, sharing is an answer to shared pain: everyone rebuilding the same assistant from scratch.
By making Gems portable and permissioned, Google transforms custom AI from something like a personal productivity hack to a team-ready asset.