Written by Jessyca Lucero-Flores
September 2026
One of the defining questions schools, districts, educators, and families are grappling with is whether artificial intelligence is strengthening or weakening students’ thinking. I believe this is one of the most important questions of our time—and a genuine catch-22. AI can expand access, feedback, creativity, and possibility. It can also allow students to bypass the productive struggle through which understanding develops.
The emerging research reflects this tension. In a recent national survey, 90 percent of college faculty said they believed generative AI would diminish students’ critical-thinking skills. Soundar-Shah and Kestigian (2026) responded with a provocative question: “If a new tool can erode critical thinking this quickly, were we ever really teaching it in the first place?”
That question is worth sitting with.
Research from Microsoft found that greater confidence in AI was associated with less reported critical-thinking effort. At the same time, AI shifted critical thinking toward new responsibilities, including verifying information, integrating responses, and overseeing the quality of the final product. In other words, AI does not eliminate the need for critical thinking; it changes where and how that thinking must occur (Lee et al., 2025).
But I also wonder whether critical thinking alone is too narrow a goal. Students need to evaluate information, but they also need to generate possibilities, recognize patterns, make connections, develop explanations, anticipate consequences, and reflect on their own reasoning. They need opportunities to practice divergent, systems, analogical, creative, and metacognitive thinking.
Adam Grant (2024) captures this shift well:
“The hallmark of expertise is no longer how much you know. It’s how well you synthesize. Information scarcity rewarded knowledge acquisition. Information abundance requires pattern recognition. It’s not enough to connect facts. The future belongs to those who connect dots.”
AI can provide students with more dots than ever before. The instructional challenge is teaching them which dots matter, how they are connected, and whether the resulting pattern is accurate, meaningful, or misleading.
Even the familiar direction to “Google it” may now lead students first to an AI-generated summary rather than an original source. Students must learn to move beyond that summary, locate the underlying evidence, evaluate its credibility, and decide whether to accept, revise, or reject the AI’s conclusions.
This brings us back to backward planning and intentional lesson design. Before deciding how students will use AI, educators must identify the standard, determine the thinking students must demonstrate, and decide what evidence will reveal that thinking. Only then should we determine where AI can support learning without completing it for students.
Drawing on Grant’s emphasis on independent thought and rethinking, I use a simple structure: Think–AI–Rethink Instructional Cycle.
Students first develop their own questions, explanations, models, claims, or possible solutions. They then use AI to expand, challenge, or critique those ideas. Finally, they return to the evidence, revise their thinking, and defend their conclusions.
Schools could reinforce this process through a monthly thinking theme. One month might emphasize metacognitive thinking, another divergent thinking, and another critical or systems thinking, and so on until you cover all twelve (that I can think of). Each month could follow the same four-week cycle:
- Week 1: Introduce and model the thinking habit.
- Week 2: Apply it through standards-based instruction.
- Week 3: Use AI strategically to expand or challenge thinking.
- Week 4: Reflect, demonstrate, and transfer the learning.
Consider a middle school science class studying plant and animal cells. During Week 1, students practice divergent thinking by observing cells, generating questions, and developing several possible explanations before receiving answers. In Week 2, they connect that thinking to NGSS MS-LS1-2 by developing models showing how the nucleus, mitochondria, chloroplasts, cell membrane, and cell wall contribute to the cell as a whole (NGSS Lead States, 2013). During Week 3, students use AI to generate additional analogies, explanations, or possible misconceptions—but then evaluate which ideas are accurate, incomplete, misleading, or unsupported. In Week 4, they revise their models, explain which AI suggestions they accepted, changed, or rejected, and demonstrate what they can now explain independently.
A science PLC could strengthen this work by identifying a common standards-aligned unit, agreeing on what students must do before consulting AI, and examining student work together. Teachers could determine where AI may expand students’ possibilities, what thinking must remain student-owned, and what evidence will demonstrate independent understanding.
Without an intentional framework, students may use AI primarily to retrieve polished answers without questioning their accuracy, evidence, assumptions, or limitations. That is the central conundrum: AI can expand human thinking, but when it repeatedly replaces the work of generating, evaluating, connecting, and revising ideas, we risk the gradual erosion of the very capabilities students need most.
AI can either expand student thinking or quietly replace it. The difference lies in the design. When educators begin with the standard, identify the thinking students must do, and use AI only to extend, challenge, or refine that thinking, the technology becomes a tool for deeper learning rather than a shortcut around it. Our responsibility is not simply to teach students how to use AI, but to ensure they can still question, connect, create, revise, and think independently because of how we ask them to use it.
Citations
Grant, A. (2024, October 13). What builds resilience—and where to find your niche. Granted. https://adamgrant.substack.com/p/what-builds-resilienceand-where-to
Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery. https://doi.org/10.1145/3706598.3713778
NGSS Lead States. (2013). MS-LS1-2: From molecules to organisms: Structures and processes. Next Generation Science Standards. https://www.nextgenscience.org/pe/ms-ls1-2-molecules-organisms-structures-and-processes
Soundar-Shah, P., & Kestigian, A. (2026, June 23). We have never taught critical thinking: AI just makes those failures evident. Inside Higher Ed. https://www.insidehighered.com/opinion/views/2026/06/23/we-have-never-taught-critical-thinking-opinion
About the Author
Jessyca is a full-time Associate with Creative Leadership Solutions who partners with schools and districts to strengthen instructional leadership, professional learning communities, and systems that improve student learning. A former principal, instructional coach, and Director of Learning, she has supported educators across the United States and internationally in public and private school settings. Grounded in her background studying ecosystems, Jessyca brings a systems-thinking perspective to school improvement and is committed to helping teams build the capacity, structures, and practices that make meaningful change sustainable.