Gemini Gems for Teachers: Build the AI Assistant You Wish You Had

Summary for Educators

Based on Med Kharbach, Ph.D.
Educators Technology, September 2026
“Gemini Gems for Teachers: Templates”

THE BIG IDEA

Most teachers who experiment with generative AI eventually discover a frustrating pattern: they repeatedly type variations of the same instructions.

“Create this for eighth-grade students.”
“Use this rubric.”
“Differentiate for three reading levels.”
“Ask questions rather than simply giving answers.”
“Use my preferred lesson format.”

Google’s Gemini Gems offer a different approach. A Gem is a customized version of Gemini created for a particular job. Teachers can establish the instructions, expectations, tone, and reference materials once and then return to that specialized assistant whenever the task arises. Educators Technology describes Gems as custom teaching assistants that can be tailored to specific instructions and materials rather than requiring educators to rebuild prompts from scratch.

The important shift is from prompting AI one conversation at a time to designing reusable AI workflows.

For teachers already overwhelmed by planning, differentiation, assessment, communication, and resource creation, that distinction could make AI considerably more useful.

KEY TAKEAWAYS

• A Gem is a reusable AI assistant. Teachers can create specialized versions of Gemini for recurring tasks instead of repeatedly entering long prompts.

• Gems can be grounded in teacher-selected materials. Google allows educators to create Gems using class materials and make those Gems available to students through Google Classroom.

• Google already provides education-oriented starting points. Classroom includes templates such as Study Partner, Quiz Me, Brainstorm Partner, and Real-World Connector that educators can customize.

• Teacher-created Gems can serve very different purposes. Potential applications include lesson planning, differentiation, assessment creation, feedback, parent communication, curriculum support, and student tutoring.

• The instructional directions matter more than the technology. A poorly designed Gem simply automates mediocre prompting. An effective Gem embeds the teacher’s goals, criteria, constraints, and expectations.

• Human judgment remains essential. AI-generated lessons, assessments, feedback, and instructional materials still require teacher review for accuracy, appropriateness, bias, and alignment with student needs.

WHY IT MATTERS

The real promise of Gems is not that teachers can generate more material.

Schools already have plenty of material.

The promise is reducing repetitive cognitive work.

Imagine that a middle school teacher routinely differentiates texts for three levels of readers. Instead of explaining the desired reading levels, vocabulary treatment, question format, and standards alignment every time, the teacher could build a Differentiation Gem containing those standing instructions.

A department chair might create a Common Assessment Gem trained to follow the department’s agreed-upon assessment structure.

An instructional coach might build a Lesson Reflection Gem that consistently asks teachers about learning targets, evidence of student thinking, misconceptions, differentiation, and next steps.

A principal could develop a Faculty Communication Gem that helps turn rough notes into concise staff communications while preserving the principal’s preferred tone.

And teachers could create student-facing Gems. Google now permits educators to assign Gems through Classroom, using teacher-selected instructional materials to create guided AI experiences. Google says these tools are designed to emphasize understanding and practice rather than simply providing answers.

That opens an intriguing possibility.

Instead of every student having unrestricted access to a general-purpose chatbot, a teacher might provide an AI assistant deliberately designed for one educational purpose.

A history teacher could create a source-analysis coach.

A mathematics teacher could create a problem-solving tutor instructed never to reveal the final answer immediately.

An English teacher could create a revision coach that asks questions about thesis, evidence, organization, and voice without rewriting the essay for the student.

That is a much more interesting use of classroom AI than simply asking a chatbot to “help with homework.”

LEADERSHIP ACTIONS

1. Start with recurring problems, not flashy technology.
Ask teachers: What task do you perform repeatedly that requires the same instructions every time? Those are promising candidates for Gems.

2. Build a small library of vetted school Gems.
Teachers should not all have to invent the same tools independently. Departments or instructional coaches could create and share reliable templates.

3. Include instructional guardrails.
A student-facing Gem should specify what it should—and should not—do. For example: ask guiding questions before offering solutions, request evidence, encourage revision, and avoid completing graded work.

4. Teach teachers to revise the instructions.
The first version of a Gem will rarely be the best version. Professional learning should include testing outputs, spotting weaknesses, and tightening the Gem’s directions.

5. Protect student information.
District leaders should make clear which information educators may enter into AI systems and ensure that use aligns with district privacy and data-governance requirements.

6. Share successful Gems across classrooms.
One teacher’s excellent rubric assistant or discussion-question generator can become a starting point for an entire team.

LEADER REFLECTION

The first generation of school AI use has largely revolved around individual prompts:

Write this. Summarize this. Generate this. Give me ideas.

Gems suggest a more mature stage.

Teachers begin asking:

What kind of assistant would help me do this work repeatedly—and what instructions would make that assistant behave like a thoughtful educator rather than a generic chatbot?

That distinction matters.

The goal should never be to automate teachers.

It should be to automate enough repetitive work that teachers have more time for the parts of education we cannot—and should not—delegate: knowing students, noticing misunderstanding, designing meaningful experiences, building relationships, exercising judgment, and inspiring young people.

For school leaders, the better AI question may therefore be:

Instead of asking teachers to learn hundreds of clever prompts, what if we helped them build a few dependable AI assistants that actually fit the way they teach?

Original Article

------------------------------

Prepared with the assistance of AI software OpenAI. (2026). ChatGPT (5.2) [Large language model]. https://chat.openai.com