1 of 27

Using AI Tools for Lesson Planning

Manas Gaur

https://kai2.umbc.edu

2 of 27

I would present a Mixed Opinion Talk

While 67% of teachers rated their AI expertise at 6 or higher on a 10-point scale,

Centre for Teacher Accreditation (CENTA)

Good Progress by AI

2025

2023

3 of 27

Focus of today’s talk at FDP

  • Why use AI tools for lesson planning?
  • What AI models and tools have been widely used in lesson planning?
  • How can I craft effective AI prompts for lesson planning?
  • NotebookLM
  • Comparison of the tools I have used
  • Ethical and Critical Perspective
  • Are there any issues with confabulation in lesson planning?
  • Is AI lesson planning safe to use in schools?

4 of 27

Why use AI Tools for Lesson Planning

  • Enhancing instruction with responsible AI
    • Improving efficiency while maintaining quality
    • Supporting differentiated and standards-aligned learning
  • Saves time on routine planning tasks
    • Generates standards-aligned outlines and activities
    • Supports differentiated instruction instantly

5 of 27

Enhancing Instruction with Responsible AI

  • Step 1: Define standards and measurable learning objectives
  • Step 2: Select the appropriate AI tool for the task
  • Step 3: Craft structured, standards-aligned prompts
  • Step 4: Generate and critically review AI output

  • Use AI to generate initial lesson outlines and activity ideas.
    • Request detailed reasoning for mathematical or technical concepts.
    • Verify equations, facts, and references.
    • Adjust tone, pacing, and examples to fit your classroom context.

6 of 27

Supported Differentiated Learning

  • Generate multiple reading levels for the same concept
    • Create scaffolded examples and guided practice
    • Design extension tasks for advanced learners
    • Produce varied assessment formats (MCQs, short answers, projects)

Maintaining Responsible AI Practices

  • Use FERPA/COPPA-compliant tools when required
    • Never upload identifiable student data
    • Maintain teacher oversight in all decisions
    • Continuously refine prompts based on classroom outcomes

7 of 27

School AI Powerups and Diffit

  • Generates presentations, flashcards, visuals, and activities
    • Supports interactive student workspaces
    • Designed for education with privacy compliance
  • Adapts reading levels and text complexity
    • Generates summaries, vocabulary, and questions
    • Exports to Google Docs, Slides, and Forms

8 of 27

School AI Powerups and Diffit

  • Generates presentations, flashcards, visuals, and activities
    • Supports interactive student workspaces
    • Designed for education with privacy compliance
  • Adapts reading levels and text complexity
    • Generates summaries, vocabulary, and questions
    • Exports to Google Docs, Slides, and Forms

9 of 27

How do I write AI Prompts for lesson Plans?

  • Effective AI prompts for lesson plans include four key elements: your specific learning standard with the exact code, the grade level and subject, relevant class constraints like time and available materials, and your preferred activity format.

  • The more context you provide upfront, the fewer revisions you'll need before the lesson is classroom-ready.

For example, instead of "plan a machine learning," write "create a 50-minute lesson on linear regression for undergraduate”, including a hands-on activity using paper strips and an exit ticket at two reading levels."

10 of 27

Prompt Engineering for Lesson Planning

  • LLMs assist curriculum development
    • Helps structure mathematical reasoning
    • Supports lab generation and coding exercises

  • Zero Shot Prompting
    • Relies on pre-trained knowledge
    • High Chances of Confabulation
    • Best for quick overviews

Teacher Prompt

LLM (Pretrained Knowledge)

Lesson Output

11 of 27

Few Shot and Chain of Thought

  • Provide example lesson structures
    • Model mimics the structure pattern
    • Improves consistency
  • Encourages step-by-step reasoning
    • Useful for derivations
    • Improves logical transparency

Example Lesson 2

New Topic Prompt

Structured Output

Define Objective

Derive Equation

Compute Gradient

Final Explanation

Example Lesson 1

12 of 27

ReAct Prompting (Reason+Act) and Structured Decomposition

  • Combines reasoning and action
    • Useful for ML lab design
    • Generates explanations + code

Reason About Concept

Generate Dataset

Produce Python Code

Create Exercises

  • Break large topics into submodules
    • Prevents shallow outputs
    • Ideal for Neural Networks lectures

Biological Intuition

Mathematical Formulation

Backpropagation

Implementation

13 of 27

Retrieval-Augmented Prompting

  • Integrates external sources
    • Improves factual grounding
    • Useful for research-based ML topics

Prompt

Retrieve Research Papers

LLM Reasoning

Cited Lesson Output

14 of 27

Practical Use of these Prompting Schemes

  • Teaching Linear Regression by prompting GPT-5.2 using a
    • CoT prompt on Step-by-Step Derivation
    • ReACT Prompting for Generating Lab and Code
  • Teaching Neural Networks using Decomposition Strategy:
    • CoT for backpropagation mathematics
    • Retrieval for Latest Architectures

  • Adaptive tutoring systems (e.g., AI Institute ALOE)
    • Automated curriculum generation
    • Human-AI collaborative teaching

15 of 27

Effectiveness of these Prompts when applied to GPT 5.2

16 of 27

How NotebookLM Can Be Used for Lesson Planning

  • AI tool grounded in uploaded sources
    • Generates summaries, outlines, and questions
    • Cites information directly from your materials
  • AI-Assisted Curriculum Design
    • Grounded in Your Own Teaching Materials
    • Instructor-Controlled & Source-Based
  • Upload PDFs, Google Docs, or course readings
    • AI analyzes only the provided sources
    • Generates lesson-ready outputs with references

Upload Course Materials

NotebookLM Processes Sources

Ask Structured Prompt

Receive Grounded Lesson Output

17 of 27

Use-cases with NotebookLM

  • Upload a textbook chapter on Linear Regression
    • Ask for a 60-minute structured lesson
    • Generate objectives, examples, and activities

  • Summarize complex topics
    • Extract key definitions
    • Generate review questions
  • Adapt explanations to different reading levels
    • Generate simplified summaries
    • Create extension questions for advanced learners

Source: Chapter PDF

Prompt: Create 60-min lesson

Output: Structured Plan

Upload Lecture Notes

Prompt: Create Study Guide

Output: Definitions + Questions

Original Text

Simplified Explanation

Advanced Extension

18 of 27

Effective Prompting with NotebookLM

  • Specify lesson duration and objectives
    • Ask for step-by-step reasoning when needed
    • Request citation-based answers
  • Source-grounded responses reduce hallucination
    • Keeps instruction aligned to the course material
    • Supports research-based teaching

19 of 27

Connection to Research & Attribution

  • When Large Language Models (LLMs) are used to generate lesson plans, they synthesize information based on patterns learned during training — not from live, verifiable databases. Without explicit citation:
    • The origin of facts, statistics, and pedagogical claims is unclear
    • Teachers cannot verify academic accuracy
    • Students may unknowingly rely on

unsupported or outdated information

    • Transparency and academic integrity

are compromised

  • Referencing ensures:
    • Accountability
    • Verifiability
    • Alignment with curriculum standards
    • Ethical educational practice

20 of 27

Tool Comparison (1/2)

Feature / Tool

NotebookLM

ChatGPT

Diffit

TeachBetter.ai

Primary Strength

Source-grounded AI using uploaded materials

Flexible generative assistant

Text differentiation & leveled materials

Teacher-focused lesson & resource generator

Lesson Plan Creation

Yes – grounded in uploaded docs

Yes – prompt-driven generation

Partial – supports adapted content

Yes – full plans, slides, quizzes

Grounded in Source Material

Built-in source grounding

Optional via uploads (Edu versions)

No (focuses on adaptation)

No automatic grounding

21 of 27

Tool Comparison (2/2)

Feature / Tool

NotebookLM

ChatGPT

Diffit

TeachBetter.ai

Differentiation Support

Moderate – based on sources

Strong – generate variants

Excellent – leveled text & questions

Moderate – template-based

Assessment & Questions

Yes – from uploaded content

Yes – quizzes, rubrics, tests

Yes – comprehension checks

Yes – built-in generators

Best Use Case

Reliable curriculum-based lessons

Creative brainstorming & explanation

Inclusive & differentiated reading

Complete classroom toolkit

22 of 27

Ethical and Critical �Perspectives

  • AI outputs require teacher validation for Accuracy
  • AI may reflect “lesson” bias or outdated information without grounding
  • Teachers ensure cultural relevance and alignment with standards. Some courses have prerequisites, but how AI sees it is different.
  • Human-in-the-loop Instructional Model

23 of 27

Is AI Lesson Planning Safe to Use in Indian Schools?

  • Shiksha Copilot is used in Karnataka government schools: teachers co-create lesson plans with AI, blending AI drafts with local expertise.
  • In these deployments, researchers highlight that teachers need structured collaboration with AI and active human oversight to ensure quality and contextual relevance.

24 of 27

Factual integrity and student learning outcomes

  • Journalistic reporting notes that in many Indian classrooms where
    • Unregulated, generic AI tools are used without pedagogical safeguards,
    • Teachers have produced lessons, quizzes, or facts that were fabricated or misleading due to AI hallucinations.

57% could correctly identify a basic AI misconception, highlighting a gap between perceived and actual understanding.

25 of 27

Safety in AI Lesson Planning

  • AI can safely assist in lesson planning when governed by teacher oversight and policy review.
  • India’s Digital Personal Data Protection (DPDP) Act exists, but education-specific AI regulation is still evolving.
  • As a Faculty, I would emphasize on:
    • Cautious Optimism
    • Recognition of Regulatory Gaps
    • Emphasis on human oversight

26 of 27

27 of 27

References

  • [Microsoft and Cornell]: Teacher-AI Collaboration for Curating and Customizing Lesson Plans in Low-Resource Schools
  • Albrecht-Crane, C. (2025, October). Thinking Smarter, not Harder? Google NotebookLM's Misalignment Problem in Education. In Proceedings of the 43rd ACM International Conference on Design of Communication (pp. 121-127).
  • Dornburg, Alex, and Kristin Davin. "To what extent is ChatGPT useful for language teacher lesson plan creation?." arXiv preprint arXiv:2407.09974 (2024).
  • Saxena, Yash, Deepa Tilwani, Ali Mohammadi, Edward Raff, Amit Sheth, Srinivasan Parthasarathy, and Manas Gaur. "Attribution in scientific literature: New benchmark and methods." arXiv preprint arXiv:2405.02228 (2024).
  • https://www.unite.ai/scispace-review/
  • Diffit: https://www.post-gazette.com/business/tech-news/2025/12/27/ai-teachers-students-magicschool-diffit/stories/202512150048
  • https://indiaai.gov.in/article/how-to-harness-ai-in-schools-opportunities-and-challenges-ahead?