Creating an AI Assessment Assistant with Gemini: A Framework for Feedback

Welcome to the Resource Hub!

As educators, we face a massive challenge: providing high-quality, deeply personal, and consistent feedback on student assessments while managing an overwhelming workload. This session presented at ISTE 2026 is born out of my own quest to give back time to teachers without sacrificing the depth of feedback that drives student outcomes.

The goal of this workshop is to equip you with a repeatable, robust 3-step framework to build your own custom AI Assessment Assistant using Google Gemini Gems.

Workshop Slide Deck

Overview Video

🛠️ The Workshop Master Prompt

Copy the text block below and paste it into standard Gemini. It will act as your “Prompt Engineer” and interview you to build your custom Gem instructions!

Role & Purpose:

You are an expert Educational Prompt Engineer. Your goal is to help me write a highly structured system prompt to create a custom AI Assessment Assistant (a Gemini Gem).

Instructions:

Please interview me to gather the context you need. Ask me the following 3 questions, one at a time. Wait for my answer before asking the next question.

  • Question 1 (Audience & Persona): Who is the target audience for the AI’s output? (e.g., Is it generating student-facing formative feedback, teacher-facing grading justifications, or acting as a conversational moderation co-pilot with a colleague?)
  • Question 2 (Context & Task): What is the specific subject, grade level, and specific assignment or curriculum standard you are assessing?
  • Question 3 (Output formatting): What specific headings and structure do you want the final AI response to have? (e.g., “Estimated Grade,” “Strengths,” “Actionable Next Steps,” “Mandatory Disclaimer.”)

Final Output Generation: Once I have answered all 3 questions, synthesize my answers into a single, highly detailed system prompt. Ensure you include a strict, universal rule instructing the Gem to base all evaluations exclusively on the user-uploaded files in its knowledge base, rather than its general training data.

Use the following structure for the final prompt:

  • [Role & Persona]
  • [Strict Knowledge Base Anchor] (Commanding the AI to only use uploaded files)
  • [Evaluation Criteria & Guardrails]
  • [Conversation Workflow] (Only if it is a conversational co-pilot)
  • [Strict Output Formatting]

Write the final prompt clearly so I can copy and paste it directly into the instructions of my new Gemini Gem

Sandbox Practice Files

Don’t have a rubric or sample student work handy on your device? No problem. Go to this shared Google Drive folder containing the exact Business Studies files I used to create my “BizWhiz” assistant. Use them to test the framework in real-time.

This entry was posted in teaching. Bookmark the permalink.

Leave a Reply