Is AI Just Helping Teachers Keep Up, or Is It Helping Students Leap Ahead?

A new paper has just been published, and it’s significant for those interested in education and technology. The report, “Emerging Patterns of GenAI Use in K-12 Science and Mathematics Education,” shares findings from a nationally representative survey of 979 U.S. public school teachers of math and science. Conducted by the RAND Corporation for the University of Washington between April and May 2025, the study aimed to provide timely insights into how these educators are managing the integration of Generative AI (GenAI). 

It reveals an important fact: 50% have integrated Generative AI (GenAI) into their work. On the surface, this is a sign of ongoing tech transformation. However, a closer look shows a more complex story. Although teachers are increasingly using tools like ChatGPT, their methods suggest we're only making modest gains in efficiency and missing a much larger chance to fundamentally change learning.

(The image is from page 5 of the report)

Report’s Main Argument

The report argues that although GenAI is rapidly adopted by K-12 math and science teachers, its full potential to greatly improve teaching and learning remains largely unexploited. This gap is primarily caused by a significant disparity between educators' interest in the technology and the absence of clear district policies, approved tools, and sufficient professional development. As a result, teachers tend to use GenAI more for personal efficiency than for innovative teaching.

Premises

• Adoption is Widespread but Nascent: Half of all surveyed math and science teachers have used GenAI tools in their teaching. However, this adoption is still in its early stages. Most users are new to the technology, with 76% having used it for less than a year. Usage is also often infrequent, with 57% of users engaging with the tools only monthly or rarely. Despite this, adoption is trending upward, as 64% of teachers who use GenAI report that they have increased their usage over time.

• Use is Focused on Efficiency, Not Instructional Transformation: Teachers are strategically leveraging GenAI to improve their preparation and decrease their workload, rather than deploying it directly in the classroom. The most common uses are: instructional planning (76%) and creating student assessments (61%). In contrast, far fewer teachers use it for tasks like automating grading (13%) or teaching students how to use GenAI as a learning tool (10%). While a large majority (86%) agree GenAI helps them work more efficiently, a smaller portion (between 20% and 46%) report using it to create richer or more relevant learning experiences for students.

• Ambivalence and Concern Regarding Student Impact: Educators hold mixed views on the impact of GenAI on student learning. The survey respondents were almost evenly split, with 30% believing the impact is positive, 34% believing it is negative, and 36% holding a neutral view. Teachers' primary concerns are centered on academic integrity, with 40% citing plagiarism and 35% noting student over-reliance on GenAI for basic tasks.

• A Critical Lack of Institutional Support: A major barrier to deeper integration is the absence of clear guidance and training from school districts. Only 5% of teachers reported that their district has formal guidelines for GenAI in place, while 48% stated that no guidelines exist at all. This creates uncertainty, which 45% of teachers identified as a barrier to adoption. Moreover, there is a significant mismatch between the demand for and supply of training; 75% of teachers want training on GenAI basics, yet 66% have not received any formal training on this topic. The main challenges cited by teachers are the time required to learn the tools (61%) and the lack of sufficient training opportunities (54%).

The “Efficiency Trap”: AI as a Virtual Assistant

Take another look at what teachers are actually doing with GenAI. The data shows they’re leaning on it as a tireless prep assistant.

76% use it for instructional planning and creating lesson plans.

61% use it to generate quizzes, worksheets, and other assessments.

A whopping 86% of GenAI-using teachers say it helps them complete routine tasks more efficiently.

Who can blame them? Easing the immense workload of educators is a tremendous win. The problem is, this use is happening outside the classroom. Far fewer teachers are using AI to create richer, more dynamic learning experiences with their students. The focus is on making the job more manageable, not necessarily making the learning more meaningful.

A Divided Verdict on Student Impact 

When it comes to students, teachers are deeply ambivalent. The survey shows a near-perfect three-way split on GenAI’s impact on student learning:

30% think the effect has been positive.

34% believe it's been negative.

36%feel it’s had no impact either way.

The main worries? Can you guess it? Plagiarism is the top concern for 40% of educators, followed by student over-reliance on AI for basic tasks (35%). This “dual-use dilemma,” where AI can both aid and hinder learning, puts teachers in a challenging position, especially when they feel like they’re navigating it alone.

Wanted: Guidance Here

This brings us to the core of the problem: a massive support vacuum. Teachers are being left to figure this all out on their own.

An impressive 48% of teachers report that their schools have no guidelines for GenAI at all. Only 5% have established formal policies.

While 75% of teachers want training on GenAI basics and capabilities, 66% have received no formal training whatsoever.

The most significant barriers to adoption aren't that teachers are resistant; rather, it's that they lack the time to learn (61%) and training opportunities (54%). Without clear policies, approved tools, and effective professional development, AI’s potential will remain locked behind a wall of uncertainty.

The top challenges cited by teachers are the time required to learn the tools (61%) and the lack of sufficient training opportunities (54%).

Our View of the Possible Implications for the Next Five Years

The findings suggest several key trends that will likely shape math and science education over the next five years:

• Shift in Professional Development Focus: The overwhelming demand for training will compel districts to move beyond introductory sessions. Professional development will need to evolve to include advanced pedagogical strategies, such as utilizing GenAI for differentiated instruction, designing complex real-world problem sets, and providing personalized student feedback. There will also be a growing emphasis on training teachers how to guide students in the ethical and practical use of GenAI, a skill 65% of teachers want to develop.

• Formalization of Policies and Tools: The current ad-hoc approach is unsustainable. Within five years, most school districts are likely to establish formal policies for the use of GenAI to address teacher uncertainty and prevent student misuse. This will be accompanied by the endorsement of education-specific GenAI tools that offer better curriculum alignment and data privacy protections than general-purpose models, such as ChatGPT, which currently dominate with 88% usage.

• Emergence of an “AI Skill Gap” Among Educators: A new divide may emerge between teachers who have mastered integrating GenAI into their instructional practice and those who only use it for surface-level efficiency tasks. This gap will be influenced by the quality of professional development and institutional support provided, potentially leading to disparities in student learning experiences across classrooms and schools.

• Integration of AI Literacy into Curriculum: To combat concerns about plagiarism and over-reliance, math and science curricula will increasingly incorporate “AI literacy.” This will involve teaching students not only how to use these tools but also how to critically evaluate their output, understand their limitations, and leverage them as a partner in problem-solving rather than a substitute for learning.

• Evolution of the Teacher's Role: As GenAI becomes more adept at handling routine tasks, such as creating lesson plans, worksheets, and assessments, the role of the teacher will continue to evolve. Educators will be able to dedicate more time to facilitating complex, collaborative projects, providing one-on-one student support, and fostering critical thinking skills—areas where the human element remains irreplaceable.

The Path Forward

The message from this report is clear to us. For GenAI to be more than just a fancy lesson planner, districts and school leaders must take the lead. Teachers need clear policies, training that goes beyond the basics to focus on effective instructional strategies, and guidance on how to help students use these powerful tools responsibly.

The AI wheel is here, but its direction is not yet set. The question for all of us is: how do we shift the focus from mere efficiency to true educational transformation?

Some potential advances remain to be thoroughly explored based on the report's findings.

1. Dynamic Assessment and Personalized Feedback

• Currently, GenAI's role in assessment is minimal. Only 13% of teachers use it for assessing student work, and 50% report it has had no impact at all on their ability to track student progress.

Example: Imagine a student struggling with a multi-step physics problem. Instead of just marking the final answer wrong, a future GenAI tool could analyze their step-by-step work. It could identify the exact conceptual error—for instance, confusing velocity with acceleration—and instantly provide a targeted explanation, along with a new, similar problem to reinforce the concept. This shifts from simple grading to a real-time, analytical tool that helps teachers understand why students are struggling.

2. True Differentiation for Diverse Learners

• While 32% of teachers use GenAI to support students with learning differences, there is a high demand for more training in differentiation techniques (69%). This suggests the current use is still basic.

Example: A teacher planning a unit on the periodic table could utilize an advanced GenAI to develop a comprehensive, multimodal learning path. For visual learners, it might create interactive 3D models of atoms. For English Language Learners, it could provide scaffolded texts with vocabulary support. For advanced students, it could generate real-world chemistry problems involving stoichiometry. This goes beyond making a single worksheet to building a full adaptive learning environment.

3. Fostering Student Agency and Inquiry-Based Learning

• The report indicates that GenAI is rarely used to directly empower students. Only 7% of teachers who reported changed interactions said GenAI gives students more agency to pose their own mathematical or scientific problems.

Example: A middle school science class could use a GenAI tool as a "research brainstorming partner." A student could input a broad interest, such as "ocean pollution," and the AI could help them narrow their focus, develop a testable hypothesis, outline experimental steps, and even identify potential variables to control. This would shift the student's role from a consumer of information to an active participant in scientific investigation.

4. AI as a Subject of Learning, Not Just a Tool for It

• Very few teachers (10%) report using GenAI to teach students how to use it as a learning tool. However, 65% of teachers want professional development on how to guide students in their use of GenAI, showing a clear need in this area.

A high school math teacher could present the class with a complex word problem and the flawed, step-by-step solution generated by an AI. The students' task would be to "audit the AI"—identifying where it went wrong, explaining the mathematical error, and correcting the process. This teaches the curriculum while simultaneously building critical thinking and digital literacy skills, helping students learn to use AI as a "learning partner instead of a learning substitute".

Ultimately, the most significant untapped potential lies in shifting GenAI's primary function from a teacher-centric tool for efficiency to a student-centered tool for deep, personalized, and inquiry-driven learning.

References

The 2025 Survey on Teachers' Use of Generative AI in Math and Science Instruction was conducted on behalf of the University of Washington by the RAND Corporation. Some sources cited in the report are listed below.

Diliberti, M. K., Grant, D., Keskin, A. K., Setodji, C. M., Hunter, G. P., DiNicola, S. E., & Schwartz, H. L. (2025). More districts are training teachers on the use of artificial intelligence. RAND Corporation. https://www.rand.org/pubs/research_reports/RRA956-31.html

Diliberti, M., Schwartz, H. L., Doan, S., Shapiro, A. K., Rainey, L., & Lake, R. J. (2024). Using Artificial Intelligence Tools in K-12 Classrooms. RAND Corporation. 

Gallup & Walton Family Foundation. (2025). Teaching for tomorrow: Unlocking six weeks a year with AI. https://www.gallup.com/analytics/659819/k-12-teacher-research.aspx

RAND American Educator Panels. (2025). American School Teacher Panel, 2025 Survey on Teachers' Use of Generative AI in Math and Science Instruction, Technical Document, RAND Corporation. 

A group of friends from “Organizational DNA Labs,” a private network of current and former team members from equity firms, entrepreneurs, Disney Research, and universities like NYU, Cornell, MIT, and UPR, gather to share articles and studies based on their experiences, insights, and deductions, often using AI platforms to assist with research and communication flow. While we rely on high-quality sources to shape our views, this conclusion reflects our personal perspectives, not those of our employers or affiliated organizations. It is based on our current understanding, which is influenced by ongoing research and review of relevant literature. We welcome your insights as we continue to explore this evolving field. A major contributor, Prof. Nilza Cruz.