Skip to main content
Lesson 11 of 14 · ChatGPT in Schools: Elementary to High School

Formative Assessment and Self-Assessment

Lesson 11 of 14 Updated November 6, 2026 Beginner
On this page
  1. Assess to Learn, Not to Punish
  2. Immediate Feedback: The Engine of Learning
  3. Creating Varied Formative Assessments
  4. Guided Self-Assessment
  5. Assisted Rubrics
  6. Two Pitfalls to Avoid
  7. Your Turn
  8. Key Takeaways

Assess to Learn, Not to Punish

Formative assessment rests on a simple idea: identify what students have mastered and where they struggle during the learning process, while there is still time to act, rather than at the end when a grade lands like a verdict. In practice, this approach always hits the same obstacle: time. Correcting thirty interim assignments every week is impossible. ChatGPT removes this bottleneck by speeding up assessment preparation and, starting in middle school (grade 6 and up), by providing students with immediate feedback.

One limit is non-negotiable: assessment remains your responsibility. The CNIL (France’s data protection authority) reminds us that teachers cannot delegate their power to evaluate, nor can they assign grades or sanctions based solely on an AI tool’s output. OpenAI similarly states that relying on the model for an assessment decision without a human in the loop is discouraged and violates its usage policies.

Immediate Feedback: The Engine of Learning

When a student works on an exercise with ChatGPT, they do not wait a week for a corrected copy. They receive feedback explaining why an answer is right or wrong, followed by remediation strategies, and they try again while their reasoning is still fresh. This cycle of attempt, feedback, and correction lies at the heart of learning. A student who understands a participle agreement error within a minute corrects it permanently, whereas the same annotation read eight days later leaves little trace. Study Mode, which asks questions one by one and checks for understanding, is ideal for this work.

Creating Varied Formative Assessments

For teachers, ChatGPT accelerates the production of assessments in different formats, each serving a specific function. A diagnostic quiz at the start of a chapter identifies missing prerequisites: before tackling fractions, verify that division is solid. Flash questions midway through a lesson check in two minutes whether a concept has landed. A case study asks students to apply knowledge to a concrete situation, revealing deeper understanding than a simple recall question. Finally, a concept map asks students to describe links between concepts; ChatGPT helps spot inconsistencies, which you then confirm.

A well-constructed prompt prepares all materials in one go:

Create a formative assessment of 8 questions on [topic].
Level: [grade].
Format: 4 multiple-choice questions + 2 short open-ended questions + 2 scenario-based questions.
For each question, prepare:
- The correct answer
- A pedagogical explanation
- A remediation strategy if the student makes a mistake
Order the questions by increasing difficulty.

Requesting remediation strategies distinguishes formative assessment from a simple quiz: every possible error already has a pedagogical response. Run this prompt for your current chapter, verify every correct answer, then intentionally answer two or three questions incorrectly on the side. The quality of the proposed remediation will tell you whether the assessment is usable.

Guided Self-Assessment

Self-assessment is a skill that must be taught, and ChatGPT can train this metacognition. Structure the dialogue in three phases. “What do I know?”: the student states what they think they have mastered, and the tool questions them to test that confidence. “What am I missing?”: targeted questions reveal unsuspected gaps. “How do I progress?”: the student defines a concrete action plan with specific points to rework, rather than a vague “review the chapter.”

This dialogue works best when ritualized in a learning journal, for which you prepare a template. At the end of each session, the student answers three questions:

  1. What did I learn today?
  2. What do I not understand yet?
  3. How will I work on this point?

The tool then helps transform vague answers into actionable formulations. “I didn’t understand equations” becomes “I get stuck when the variable appears on both sides of the equal sign.” The second formulation opens the door to precise work; the first opens nothing.

Assisted Rubrics

ChatGPT also generates criteria-based rubrics tailored to each activity: criteria aligned with lesson objectives, described levels of mastery, observable indicators for each level, and explicit weighting. Review them, then distribute the rubric before students begin. Knowing exactly how they will be judged shifts their energy from guessing to working.

The rubric enables another use: self-positioning. Before submitting their work, the student places themselves on each criterion, and the tool helps them stay lucid by questioning them: “You placed yourself at the highest level for structure; can you show me where the transition between your two parts is?” Few students maintain a flattering self-assessment after three questions like that.

Two Pitfalls to Avoid

The first pitfall is turning ChatGPT into a grading machine. Formative assessment aims for progress, and grading remains a human decision. The second is the risk of complacency: feedback that is too lenient leads students to overestimate their skills, resulting in a rude awakening during a test. Calibrate your prompts to obtain feedback that is honest and constructive, encouraging regarding progress but uncompromising regarding gaps.

Self-assessment in three questions Three phases: what the student thinks they know, which the tool questions; what they are missing, revealed by targeted questions; how to progress, with a precise plan rather than a vague "review the chapter." ASSESSMENT Self-assessment in three questions 1 I know Confidence questioned 2 I am missing Gaps revealed 3 I progress A precise plan Grading remains a human decision: do not grade based solely on an AI tool's output.
Lesson method; assessment based on the CNIL FAQ for teachers (June 2025) and the article "Getting started with ChatGPT as an educator" from the OpenAI Help Center (November 2026).

Your Turn

Prepare a complete formative assessment.

  1. Run the assessment prompt from the lesson on your current chapter
  2. Verify every correct answer and every remediation strategy
  3. Intentionally answer three questions incorrectly on the side and judge the remediation
  4. Have the tool generate the criteria-based rubric for the next assignment, and review it
  5. Prepare the three-question learning journal template

Official sources: Getting started with ChatGPT as an educator · Using study mode in ChatGPT (OpenAI Help Center) · Usage policies (OpenAI) · Teacher: How to use an AI system in your duties? (CNIL) · Artificial intelligence and its uses in education (éduscol)

Key Takeaways

  • Assessment remains the teacher’s responsibility: do not grade based solely on an AI tool’s output
  • Immediate feedback accelerates the attempt, feedback, correction cycle
  • Formative assessment includes remediation for every possible error
  • Criteria-based rubrics, reviewed and shared in advance, make assessment transparent
  • Calibrate feedback to be honest and constructive