Science classrooms are continuing to evolve as science teachers seek to balance infusing their classrooms with the excitement of discovery and wonderment with helping students understand the theoretical underpinnings and fundamental principles. Asking students questions such as “Why does the chair not go through the floor?” or “Is light a wave or a particle?” will, no doubt, spark a series of counter questions from us to help our students access the essence of the all-important, “Why?” of the subject matter. But at the same time, these questions pose a challenge since they can be given to AI.

The act of guiding scientific thinking is a skill that many veteran teachers may have cultivated through years of experience, but it can be challenging for novice teachers. This blog post aims to illustrate how the latest features of NotebookLM can guide this process, regardless of a teacher's level of experience. It is also important to clarify that while there may be many different tools that can help with this, we have chosen to focus on Google NotebookLM because many educators and students are familiar with Google products in the education setting.

New Features of NotebookLM Are Transformative

Unlike traditional AI chatbots, NotebookLM is designed for accuracy and transparency. It doesn’t invent facts or fill in blanks. Instead, it works only from the sources the user uploads: your lab readings, diagrams, data tables, or even microscope images. For science educators, this creates a secure and authentic environment for students to practice real-world scientific reasoning without misinformation clouding their conclusions.

NotebookLM uses MLLM, which is a Multimodal Large Language Model. This is an advanced AI system that can process and generate content across multiple data types, or "modalities," such as text, images, audio, and video. The following paragraphs describe the newly released features of NotebookLM that support three-dimensional learning even more, along with suggested strategies for use in the science classroom.

Custom Reports
Teachers or students can generate AI-driven reports in customizable formats and tones. Users can define structure, style, and even language (more than 80 options). The AI builds reports directly from uploaded sources, making it ideal for turning lab notes or readings into summaries, comparisons, or scientific analyses.

Flashcards
NotebookLM can automatically create flashcards from uploaded content, which can help reinforce key terms, equations, or concepts. These flashcards can adapt to learners’ responses, offering immediate feedback and deeper review. If printed, students can use them in card sort–type activities to build concepts as well.

Quizzes
The platform now converts content into quick, interactive quizzes that check understanding. Teachers can use these quizzes to assess comprehension of readings, experiments, or lecture materials.

Modular Audio Overviews
This upgraded version of the “podcast” feature produces short, AI-generated audio summaries of uploaded content. You can choose formats like deep dive, brief, critique, or debate—great for review sessions, flipped classrooms, or auditory learners. Each type of format has different features, from a simple overview to more in-depth discussion.

Moving from the typical use of these tools to uses that engage students in more three-dimensional thinking represents a different approach to using these tools. Science teachers have always been champions of curiosity, urging students to ask why, to seek patterns in data, and to see beyond the surface of phenomena. Now as artificial intelligence transforms education, tools like the new features in NotebookLM are becoming powerful allies in that mission. 

Practical Ways to Use These New Features

These new features don’t just make learning interactive: They have the potential to reshape how students think. When used thoughtfully, NotebookLM can scaffold critical thinking by making students’ reasoning visible, prompting reflection, and grounding every insight in evidence.

Consider, for instance, a classic physics challenge: the wave-particle duality of light. A teacher might upload excerpts from Einstein’s work on the photoelectric effect, a short article about interference patterns, and a diagram of the double-slit experiment. After students have thoroughly provided their explanation, guided by your expertise as a teacher, you can help students use NotebookLM’s Custom Report feature. Students could ask the AI to compare how each source explains the dual behavior of light. The AI would then highlight key contradictions and overlapping ideas, helping students evaluate which model best fits the evidence. Rather than accepting one explanation at face value, learners are guided to weigh competing models—the essence of critical thinking in science.

NotebookLM can also help students become more reflective experimentalists. In a biology lab on photosynthesis, for example, students can upload their lab reports (excluding their names), raw data, and even a photo of their setup. When they ask what might explain the variation in their results, the AI can analyze both text and images to suggest potential sources of error, such as uneven lighting or inconsistent measurement timing. At the very least, it provides points for discussion and consideration. Students can then use that feedback to refine their next experiment. This type of AI-assisted metacognition doesn’t replace scientific judgment; it strengthens it by connecting reasoning with real evidence while modeling how these points might be connected.

In chemistry, the platform’s new flashcard and quiz features provide yet another layer of critical engagement. When students upload graphs showing how reaction rates change with temperature, NotebookLM can automatically generate adaptive questions like these: “What happens to the slope beyond 50°C?” or “Why does the rate double when the temperature rises from 25°C to 50°C?” These prompts require explanation, not memorization, helping students translate visual data into conceptual understanding.

NotebookLM also enhances students' work with complex, multimodal data, such as in environmental science projects that combine pollution maps, soil samples, and scholarly readings. Initially, students can use these materials to generate their own Claim-Evidence-Reasoning (CER), then students can upload all of these materials and ask the system to review the students’ own CER and build a CER outline for their argument. Students can compare and contrast the two, then engage in additional examination of the information and CERs. By comparing the strengths of different datasets, NotebookLM encourages students to justify why one piece of evidence supports a claim more convincingly than another. In doing so, it transforms the often mechanical task of writing lab reports into a deeper exercise in scientific argumentation.

Larger Shifts in Science Education

These applications reflect larger shifts in science education. AI is no longer just grading work. It’s diagnosing misconceptions, visualizing data, and prompting self-reflection. It’s multimodal, capable of interpreting graphs and text together, and it can provide real-time feedback on experimental design or data quality. Yet amid all these advances, the teacher remains the architect of the three-dimensional learning process. AI may assist with error analysis or hypothesis generation, but it’s the teacher who frames the question, guides the reasoning, and ensures that curiosity—not convenience—drives the learning process.

Multimodal AI in science contexts
Because science teaching involves diagrams, graphs, images (microscope views, spectra, apparatus photos), and experimental data, AI models that can handle multi-mode inputs and outputs are especially valuable. The use of MLLMs is an important frontier. 

Explaining complex phenomena, connecting scales, and analogies
AI can help bridge multiple representations (mathematical, narrative, graphical, microscopic) in science, and generate analogies or explanations at different levels of abstraction. For example, an AI might explain diffusion using a particle model, an equation, and a graphical simulation, and help students map between them.

Error diagnosis and feedback in student reasoning
In science, student misconceptions or flawed reasoning (e.g., misunderstanding forces, energy, chemical equilibrium) are common. AI can help detect misconceptions by analyzing responses (explanations, sketches, diagrams) and provide targeted feedback or prompt corrections.

Importance of Continued Teacher Monitoring of AI Outputs

Still, as we embrace tools like NotebookLM, science teachers must proceed with thoughtful caution. AI is only as reliable as the data it’s given, and errors in uploaded sources can cascade into flawed reasoning. Teachers should model healthy skepticism, showing students how to verify AI-generated insights against trusted scientific literature. Privacy and data ethics also demand attention; student lab notes, photos, or local datasets should be handled securely. Moreover, while NotebookLM’s scaffolds can reduce cognitive load and enhance scientific thinking, overreliance on it may dull students’ initiative if not balanced with open-ended questioning. Teachers must frame AI as a partner in reasoning, not a shortcut to answers—guiding students to question, critique, and ultimately own their scientific thinking.

NotebookLM isn’t about automating science learning: It’s about amplifying it. When teachers use it to prompt inquiry, reflection, and argumentation, it becomes more than a tool; it becomes a catalyst for cultivating young scientists who not only know facts, but also understand how to think about them.

 

Valerie Bennett, Ph.D., Ed.D., is an Assistant Professor in STEM Education at Clark Atlanta University, where she also serves as the Program Director for Graduate Teacher Education and the Director for Educational Technology and Innovation. With more than 25 years of experience and degrees in engineering from Vanderbilt University and Georgia Tech, she focuses on STEM equity for underserved groups. Her research includes AI interventions in STEM education, and she currently co-leads the Noyce NSF grant, works with the AUC Data Science Initiative, and collaborates with Google to address CS workforce diversity and engagement in the Atlanta University Center K–12 community.


 

Christine Anne Royce, Ed.D., is a past president of the National Science Teaching Association and currently serves as a Professor in Teacher Education and the Co-Director for the MAT in STEM Education at Shippensburg University. Her areas of interest and research include utilizing digital technologies and tools within the classroom, global education, and the integration of children's literature into the science classroom. She is an author of more than 140 publications, including the Science and Children Teaching Through Trade Books column.

This article is part of the blog series From Chalkboards to AI, which focuses on how artificial intelligence can be used in the classroom in support of science as explained and described in A Framework for K–12 Science Education and the Next Generation Science Standards.

The mission of NSTA is to transform science education to benefit all through professional learning, partnerships, and advocacy.