Google Gemini Gems: Create Your Own AI Assistants

Lisa Ernst · 31.05.2026 · AI Tools · 10 min read

Google Gemini Gems are customized versions of Gemini that keep your recurring instructions, tone, goals and role in one reusable assistant. Instead of writing the same long prompt every time, you create a Gem once and open it whenever you need a focused helper for writing, coding, research, planning, sales, learning or internal workflows.

This guide explains how to create your own Gem step by step, what to write into the instructions, which practical examples work best, and how Gemini Gems compare with ChatGPT Custom GPTs.

Source: Image sources: Pexels / Matheus Bertelli and Google Gemini logo via Wikimedia Commons

What are Google Gemini Gems?

A Gem is a custom AI assistant inside Gemini. You give it a name and a set of instructions, then use it like a specialized version of Gemini. A normal Gemini chat is flexible, but a Gem is better when the task repeats: it remembers how it should behave, what format it should use, what role it should take and what boundaries it should follow.

Typical Gems can act as a writing editor, interview coach, coding helper, learning tutor, meeting preparation assistant, product brainstorm partner or marketing reviewer. Google also offers premade Gems, but the strongest use case is building your own assistant around your actual workflow.

Source: Image source: Google Gemini logo via Wikimedia Commons

Gems are part of the Gemini ecosystem. They are designed for repeatable tasks where a saved instruction set is faster than starting from a blank prompt.

When should you create a Gem?

Create a Gem when you catch yourself reusing the same prompt structure more than a few times. The best Gems are not vague chatbots. They have a precise job, a clear output format and a repeatable decision pattern.

• Use a Gem for repeatable work: weekly content planning, SEO checks, meeting summaries, customer reply drafts, code reviews or study flashcards.
• Use a normal Gemini chat for one-off exploration: broad brainstorming, casual questions or tasks where you do not yet know the structure.
• Use a separate Gem per workflow: one assistant for blog outlines, another for customer emails, another for code review. A small focused Gem is usually better than one overloaded mega-assistant.

How to create a Google Gemini Gem step by step

The exact interface can change slightly, but the basic process is simple: open Gemini, go to Gems, create a new Gem, give it a name, write clear instructions, test it in the preview area and save it.

1. Open Gemini: Go to gemini.google.com while signed in with your Google account.
2. Open the Gems area: In the sidebar, choose Gems or Explore Gems. If you do not see it immediately, check the sidebar or the settings/help menu.
3. Create a new Gem: Select New Gem.
4. Name the Gem: Use a name that describes the job, for example SEO Blog Editor, Study Coach or Sales Email Reviewer.
5. Write the instructions: Define the role, target user, task, style, output format, rules and limitations.
6. Preview the Gem: Ask two or three realistic test questions and check whether the answer style fits your goal.
7. Save it: Save the Gem after previewing. Testing alone does not replace saving.
8. Improve it after real use: Add rules when the Gem makes repeated mistakes. Remove vague instructions that do not affect the result.

Source: Image source: Pexels / Matheus Bertelli

A useful Gem starts with a concrete workflow. Do not only tell it to be helpful; tell it exactly what job it owns, what it should ask, what it should avoid and how the final answer should look.

The instruction formula that makes Gems actually useful

The quality of a Gem depends mostly on the instruction text. A weak instruction says: You are a marketing expert. Help me with content. A strong instruction defines the job in operational terms.

A practical Gem instruction structure

Copy-ready example: SEO Blog Editor Gem

You are an SEO blog editor for practical AI tool tutorials. Your job is to turn a topic into a clear, useful and search-friendly article plan. Write for readers who want direct instructions, not generic AI hype. Always include search intent, a compact introduction, a step-by-step section, practical examples, a comparison table if useful, common mistakes, FAQ questions and a concise meta title and description. Keep the tone clear, helpful and realistic. If a claim depends on current product features, tell the user to verify it before publishing.

Copy-ready example: Meeting Preparation Gem

You are a meeting preparation assistant. Help me prepare for business meetings by turning rough notes into a short briefing. Ask for missing context only when essential. Structure every answer into: meeting objective, key people, likely questions, risks, talking points, decisions needed and follow-up tasks. Keep the wording professional and compact.

Copy-ready example: Code Review Gem

You are a pragmatic code reviewer. Focus on bugs, security risks, maintainability and unnecessary complexity. Do not rewrite the whole file unless asked. First list the most important issues, then suggest precise changes. Prefer simple solutions and explain trade-offs shortly. If the code touches authentication, database writes, file uploads, payment logic or external API calls, check security more strictly.

Examples of strong Gemini Gems for daily work

Here are practical Gem ideas that work well because they have clear input and clear output.

Source: Image source: Pexels / AI25.Studio

For teams, Gems are most valuable when they standardize recurring work: the same review logic, the same writing style and the same checklist can be reused across similar tasks.

Google Gemini Gems vs ChatGPT Custom GPTs

Gemini Gems and ChatGPT Custom GPTs solve a similar problem: they let you create a specialized assistant instead of repeating the same prompt every time. The difference is mainly ecosystem, depth of configuration and distribution model.

Source: Image source: Pexels / Matheus Bertelli

Custom GPTs and Gemini Gems are both useful. Choose based on where your work already happens: Gemini and Google Workspace for Google-centered workflows, or ChatGPT Custom GPTs when you need the broader GPT configuration model.

Where Gemini Gems are especially strong

Gems are a good fit when speed and repeatability matter more than complex automation. They help people who want consistent AI help without maintaining a full app or integration. The more specific your instructions are, the more predictable the Gem becomes.

• Google-centered work: If your files, planning and communication already live around Google products, Gemini Gems fit naturally into that environment.
• Simple team standardization: A shared Gem can help multiple people use the same briefing format, writing rules or review checklist.
• Personal productivity: A Gem is ideal for personal routines such as language learning, weekly planning, CV improvement or writing support.
• Low setup effort: You can build a useful Gem in minutes without code.

Where ChatGPT Custom GPTs may be stronger

Custom GPTs are often better when the assistant needs more advanced configuration, uploaded knowledge, GPT Store-style distribution or external API actions. For example, a GPT that checks live data from a business system, creates structured API requests or connects to a custom backend is usually closer to the Custom GPT model.

For simple repeatable guidance, a Gem is often enough. For a tool-like assistant with custom capabilities, ChatGPT Custom GPTs may be more flexible. For more AI tool ideas and practical automation topics, you can also explore Zerlo.

Source: Image source: European Union / Lukasz Kobus via Wikimedia Commons, CC BY 4.0

Gemini is one of Google's central AI products. Gems build on that ecosystem by letting users turn general AI assistance into repeatable, task-specific assistants.

Common mistakes when building Gems

Most weak Gems fail because the instruction is too broad. A custom assistant should not only describe a personality. It should define a workflow.

• Too vague: “Help me with marketing” is not enough. Say what type of marketing, for which audience and in which format.
• Too many jobs: One Gem should not be your writer, lawyer, designer, accountant and coder at the same time.
• No output format: Without a clear structure, answers vary too much.
• No test prompts: Always test the Gem with realistic tasks before relying on it.
• No boundaries: Tell the Gem when it should ask follow-up questions, when it should warn you and what it should not invent.

A simple checklist before saving your Gem

Before you save a Gem, check whether it passes this practical quality test:

1. The Gem has one clear job.
2. The audience is defined.
3. The preferred tone is defined.
4. The output format is defined.
5. The Gem knows what to avoid.
6. The Gem asks follow-up questions only when needed.
7. You tested it with at least two realistic prompts.
8. You know when not to use it.

Source: Image source: Arthur Petron via Wikimedia Commons, CC BY-SA 4.0

Gemini is closely connected with Google DeepMind's AI work. For users, Gems are the practical layer: they turn a broad AI model into a repeatable assistant for a specific task.

Final recommendation

Start with one Gem for one painful recurring task. Do not try to build a perfect assistant on the first attempt. Create a simple version, test it with real work, then improve the instructions. The best Gems become valuable because they are used repeatedly and refined over time.

If your workflow is mostly inside Google products, Gemini Gems are an easy and practical starting point. If you need deeper assistant configuration, public distribution or external API actions, compare the same use case with a ChatGPT Custom GPT before deciding.

FAQ