AI and Critical Thinking: Are We Raising Thinkers or Prompt Writers?

When technology can generate answers in seconds, the greatest challenge facing education may no longer be finding information, but ensuring that students still know how to think.

By Eureka | Enlightening Minds. Elevating Education.

The Classroom Paradox: When the Answer Comes Before the Thinking

Imagine a Grade 5 classroom where students are exploring how people's beliefs and values influence their communities and cultures.

The teacher introduces an inquiry question:

"How do cultural beliefs and values shape the way people live?"

One student begins investigating different perspectives, examining sources, discussing ideas, and constructing an explanation.

Another student opens an artificial intelligence tool, types the inquiry question, and receives a sophisticated response within seconds.

The AI-generated explanation is well organized. It includes relevant terminology, examples, and a convincing conclusion.

The student copies the response, makes a few adjustments, and submits the assignment.

On the surface, both students have completed the task.

But have both students experienced the same learning?

The first student has engaged in questioning, investigating, analyzing, connecting ideas, and making judgments.

The second may have produced an equally impressive outcome without engaging in those processes.

This raises one of the most important questions facing contemporary education:

Are we developing students who can think critically, or students who have become increasingly skilled at obtaining answers without understanding them?

Artificial intelligence is not simply changing how students complete assignments.

It is challenging educators to reconsider what learning means, how understanding develops, and what schools should assess.

The concern is not that AI can provide answers.

The concern is what happens when receiving an answer begins to replace the intellectual effort required to understand it.

1. The Difference Between Producing an Answer and Developing Understanding

For generations, schools have used completed assignments as evidence of learning.

Students write essays, solve mathematical problems, produce research reports, and prepare presentations.

Teachers examine these products to make judgments about knowledge, understanding, and skills.

Generative AI complicates that relationship.

A student may now produce an accurate mathematical explanation without understanding the strategy used.

Another may submit a sophisticated persuasive essay without being able to defend the argument.

A third may generate a detailed scientific investigation without understanding the relationship between the variables being discussed.

The quality of the product can no longer be assumed to represent the quality of the student's thinking.

This does not mean that AI-assisted work has no educational value.

Rather, it means that educators must distinguish between successful task completion and genuine learning.

The OECD's Digital Education Outlook 2026 highlights this challenge. Its review of emerging evidence indicates that generative AI can improve task performance while, in some circumstances, reducing the cognitive and metacognitive engagement necessary for durable learning.

In other words, students may appear more successful while becoming less involved in the thinking that success is intended to represent.

This distinction is especially important in inquiry-based education, where the purpose of learning extends beyond arriving at correct answers.

Inquiry involves investigating uncertainty, examining evidence, challenging assumptions, recognizing different perspectives, and constructing meaning.

If AI performs these processes on behalf of the learner, the final product may be impressive, but the essential learning opportunity may have been lost.

2. Is AI Making Students Less Critical?

The relationship between AI and critical thinking is more complex than a simple argument that technology makes students intellectually passive.

AI can support critical thinking when learners use it to compare perspectives, challenge explanations, explore alternatives, and evaluate evidence.

However, it can also discourage independent thinking when learners accept its responses without question.

A 2025 study by researchers from Microsoft and Carnegie Mellon University surveyed 319 knowledge workers about their experiences using generative AI.

The researchers found that greater confidence in AI was associated with less self-reported critical-thinking engagement, while greater confidence in one's own abilities was associated with more critical-thinking engagement.

The study also identified a shift in how people think when using AI: from producing information themselves toward verifying, integrating, and overseeing AI-generated content.

These findings are important, but they require careful interpretation.

The research examined adults in professional settings, not schoolchildren, and relied on self-reported experiences. It does not establish that AI inevitably reduces students' thinking abilities.

Nevertheless, it raises a question educators cannot afford to ignore:

When learners increasingly trust technology to perform intellectual tasks, are they developing stronger judgment, or becoming less inclined to exercise it?

The educational risk is not the existence of AI assistance.

It is uncritical dependence on that assistance.

A student who asks AI to explain a difficult concept and then evaluates the explanation may be engaging in meaningful learning.

A student who asks AI to complete the task and submits the result without understanding it may be avoiding the very thinking the assignment was designed to develop.

The difference lies in how the technology is used.

3. The Hidden Danger of Cognitive Offloading

Human beings have always relied on tools to reduce cognitive demands.

Calculators support computation. Search engines help retrieve information. Dictionaries provide definitions. Digital applications organize tasks.

These tools can make learning more efficient and free mental resources for complex thinking.

Generative AI extends this capability considerably.

It can summarize, interpret, compare, argue, plan, explain, and generate creative responses.

This introduces an important distinction between two forms of cognitive offloading.

Productive cognitive offloading occurs when technology reduces unnecessary mental effort so learners can focus on higher-order thinking.

For example, using AI to help organize a large collection of research notes may allow a student to concentrate on evaluating sources and constructing an argument.

Unproductive cognitive offloading occurs when technology performs the very cognitive process the learner is supposed to develop.

For example, if the objective is to develop argumentation skills, asking AI to formulate the argument, select the evidence, and construct the conclusion may bypass the intended learning.

A 2025 field experiment published in the Proceedings of the National Academy of Sciences explored this distinction with nearly 1,000 high school mathematics students.

Researchers compared a general-purpose AI tutoring interface with a more structured AI tutor designed to guide students through hints rather than immediately providing solutions.

Students using the tools demonstrated improved performance during AI-assisted practice.

However, students using the less structured version performed worse on a subsequent unassisted assessment than students who had not received AI assistance. The structured tutoring approach substantially reduced this negative effect, although it did not produce a statistically significant improvement in unassisted performance.

The implications are important.

Access to sophisticated AI does not automatically translate into stronger learning.

The educational design of AI use matters as much as the technological capability of the tool itself.

4. When AI Becomes an Intellectual Shortcut

Consider a student asked to analyze the causes of climate change.

Without AI, the learner may need to locate relevant sources, compare explanations, identify relationships, evaluate evidence, and construct a justified conclusion.

With AI, the student can receive a complete response almost immediately.

The technology has reduced the time required to complete the task.

But has it improved the student's ability to reason?

Not necessarily.

The challenge becomes particularly significant when students begin to confuse the ability to generate information with the ability to understand and evaluate it.

Three risks deserve attention.

The Illusion of Understanding

An AI-generated explanation may sound clear and persuasive, giving students the impression that they understand a concept they have not yet meaningfully processed.

However, recognition is not the same as understanding.

A student may recognize the explanation as sensible while remaining unable to reproduce the reasoning independently or apply it to a new context.

The Decline of Productive Struggle

Learning frequently involves uncertainty, unsuccessful attempts, revision, and reconsideration.

These experiences are not necessarily signs of ineffective teaching. When appropriately scaffolded, they can contribute to deeper learning.

If students consistently turn to AI at the first sign of difficulty, they may lose opportunities to develop persistence, problem-solving strategies, and intellectual independence.

The goal is not to make learning unnecessarily difficult.

It is to preserve the cognitive effort that serves a meaningful educational purpose.

The Authority Illusion

Generative AI can produce responses that appear confident and convincing even when they contain factual inaccuracies, invented references, oversimplifications, or biased interpretations.

Students who do not understand these limitations may mistake fluent language for reliable knowledge.

This creates a new educational responsibility: learners must be taught not only how to obtain information, but how to question the systems generating it.

5. Critical Thinking in the AI Era: What Should Students Actually Learn?

Critical thinking has always involved more than identifying correct information.

It includes the ability to examine assumptions, evaluate evidence, recognize bias, consider alternatives, construct logical arguments, and revise conclusions when new information emerges.

AI does not make these abilities obsolete.

It makes them increasingly important.

In a world where convincing answers can be generated instantly, learners need to become more discerning about what deserves their trust.

They should be able to ask:

• What evidence supports this claim?

• Is this information accurate and verifiable?

• Which perspectives have been represented or excluded?

• What assumptions influence this explanation?

• Could a different interpretation be equally reasonable?

• Does this conclusion follow logically from the available evidence?

• What do I understand independently of the AI-generated response?

UNESCO's AI Competency Framework for Students (2024) emphasizes human-centered thinking, ethical responsibility, and critical judgment when engaging with artificial intelligence.

These competencies suggest that AI literacy should not be limited to learning how to construct effective prompts.

Students also need to understand the limitations, implications, and responsibilities associated with AI-generated information.

Prompt engineering may help students communicate effectively with AI. Critical thinking enables them to decide whether the response deserves to be accepted at all.

6. The Modern educational systems Perspective: Can AI Strengthen Inquiry Rather Than Replace It?

For modern educational systems educators, the relationship between AI and critical thinking is particularly relevant.

IB education emphasizes inquiry, conceptual understanding, reflection, international-mindedness, and learner agency.

These principles provide a valuable foundation for determining how AI should be used in teaching and learning.

In March 2026, the IB published draft AI Design Principles for community feedback.

Among these principles was the expectation that AI should strengthen inquiry, critical thinking, creativity, and collaborative problem-solving while preserving educator agency and human relationships.

This raises a fundamental pedagogical question:

Does our use of AI deepen the learner's inquiry, or does it replace the inquiry process with a ready-made response?

Consider a PYP unit exploring identity, beliefs, and cultural values.

A superficial use of AI might involve asking students to generate descriptions of different cultural communities and reproduce those descriptions in presentations.

A more meaningful approach would require learners to begin with their own questions and investigate authentic sources.

AI might then be used to generate alternative interpretations or raise further questions.

Students would evaluate those responses against evidence from reliable texts, community perspectives, and other appropriate sources.

They might identify oversimplifications, challenge stereotypes, and refine their conceptual understanding.

In this situation, AI does not become the primary authority.

It becomes one resource within a wider process of inquiry.

The learner remains responsible for interpretation, judgment, and meaning-making.

This is consistent with the broader IB emphasis on students taking an active role in their learning and acknowledging AI-generated contributions appropriately.

In an inquiry-based classroom, AI should expand the questions learners can explore, not eliminate their responsibility for answering those questions thoughtfully.

7. From Answer Generators to Thinking Partners

The most productive educational question is not simply whether students should use AI.

It is how AI can be positioned to support meaningful thinking.

This requires a shift from treating AI primarily as an answer generator to using it, where appropriate, as a carefully guided thinking partner.

Consider the following examples.

In Mathematics

Answer-dependent use:

"Solve this division problem and show me the answer."

Thinking-focused use:

"I have attempted this division problem using partial quotients. Ask me questions that will help me identify where my reasoning might be incorrect. Do not give me the answer."

The second approach encourages reflection and self-correction while preserving responsibility for the mathematical reasoning.

In Writing

Answer-dependent use:

"Write a persuasive essay about protecting endangered animals."

Thinking-focused use:

"Here is my argument and supporting evidence. Identify two possible weaknesses in my reasoning and ask questions that could help me strengthen my position. Do not rewrite my essay."

The learner retains ownership of the ideas and the writing process.

In Scientific Inquiry

Answer-dependent use:

"Explain why biodiversity is important."

Thinking-focused use:

"I believe biodiversity supports ecosystem stability. Present a possible counterargument and help me identify what evidence I would need to evaluate both perspectives."

The second interaction creates an opportunity for deeper analysis rather than simple information retrieval.

In Research

Answer-dependent use:

"Give me five sources proving that my argument is correct."

Thinking-focused use:

"What evidence might challenge my current argument? Suggest questions I should investigate and help me identify reliable places to search. I will verify the sources myself."

This encourages learners to move beyond confirmation bias and engage with competing evidence.

The distinction is not about writing longer or more sophisticated prompts.

It is about preserving the student's responsibility for thinking.

8. The Eureka THINK Framework: A Practical Approach to AI-Supported Critical Thinking

To help educators integrate artificial intelligence without compromising cognitive engagement, Eureka proposes a five-stage instructional framework: THINK.

The framework is a practical pedagogical proposal, informed by research on AI-supported learning, critical thinking, and metacognition.

T — Think Independently First

Before interacting with AI, learners should attempt the task, formulate an initial explanation, or articulate their existing understanding.

This establishes a starting point for learning.

Students might record:

"I think..."

"My initial explanation is..."

"The evidence I already have suggests..."

The objective is to ensure that AI does not automatically become the starting point for every cognitive task.

H — Hypothesize and Question

Students identify what they are uncertain about and generate questions worth investigating.

Instead of asking AI for a complete answer, they consider what they need to understand.

For example:

"Which part of my explanation needs stronger evidence?"

"What alternative interpretation should I consider?"

I — Interrogate AI Responses

Learners examine AI-generated information critically rather than accepting it immediately.

They look for factual accuracy, missing perspectives, unsupported assumptions, and possible biases.

They distinguish between claims that can be verified and claims that require further investigation.

N — Navigate Evidence and Perspectives

Students compare AI-generated responses with credible sources, disciplinary knowledge, primary evidence, and relevant human perspectives.

They consider whether information is consistent, contradictory, incomplete, or context-dependent.

At this stage, students practise making reasoned judgments rather than relying exclusively on automated explanations.

K — Keep Ownership of Learning

Learners synthesize their findings, construct their own conclusions, acknowledge AI assistance where required, and reflect on how their understanding has developed.

They should be able to explain their reasoning without depending on the tool to speak for them.

A useful closing reflection is:

"What can I now explain, justify, or apply independently that I could not do before?"

The THINK framework positions AI as a possible contributor to learning while keeping intellectual responsibility with the learner.

9. Rethinking Assessment: What Are We Really Measuring?

Perhaps the greatest challenge presented by AI is not the possibility of students using it dishonestly.

It is that many traditional assessments are vulnerable to being completed successfully without demonstrating the intended learning.

If students can generate sophisticated responses without developing the associated knowledge or skills, schools need to reconsider how understanding is demonstrated.

This does not mean abandoning written assignments, research reports, or extended projects.

It means designing assessment processes that make thinking more visible.

Educators can strengthen assessment through several approaches.

Assess the process as well as the product.

Invite students to demonstrate how their ideas developed, which evidence influenced their conclusions, and how they responded to feedback.

Include opportunities for explanation and justification.

Short conferences, oral discussions, classroom demonstrations, and reflective conversations can provide additional evidence of understanding.

Use appropriate independent checkpoints.

Some tasks should be completed without AI assistance so teachers can assess what learners understand and can apply independently.

Design authentic opportunities for transfer.

Invite students to apply their understanding to unfamiliar situations, new examples, or meaningful problems.

Clarify acceptable AI assistance.

Students should understand when AI use is permitted, what type of assistance is appropriate, and how to acknowledge that assistance.

Teachers should also recognize that no single assessment method provides a complete picture of learning.

Oral explanations, for example, must be designed accessibly for learners with communication, language, or additional learning needs.

The purpose is not to create an atmosphere of suspicion.

It is to gather credible evidence of learning while maintaining fairness, accessibility, and student trust.

The central assessment question is shifting from "What did the student produce?" to "What does the student understand, and how do we know?"

10. The Role of Teachers: More Important, Not Less

As AI becomes more capable, some may question whether the traditional role of the teacher is becoming less significant.

In reality, the need for thoughtful instructional decision-making becomes even more important.

AI can generate explanations, suggest activities, draft materials, and provide feedback.

However, educators remain responsible for determining whether these contributions are developmentally appropriate, pedagogically meaningful, accurate, equitable, and aligned with learning intentions.

Teachers must decide:

When should learners work independently?

When might an AI-generated explanation support understanding?

When is productive struggle educationally valuable?

When could AI assistance undermine the skill being assessed?

How can students be supported to question and evaluate automated responses?

These decisions require pedagogical expertise, professional judgment, and an understanding of learners that technology cannot simply replace.

AI may also help teachers develop differentiated materials, explore instructional alternatives, and reduce certain administrative demands.

However, these benefits should be considered alongside accuracy, privacy, and the need to maintain professional oversight.

UNESCO's AI Competency Framework for Teachers (2024) recognizes the importance of equipping educators with the knowledge and skills required to use AI responsibly and effectively.

The future of teaching should therefore not be framed as a competition between teachers and technology.

The more useful question is how technology can strengthen teachers' capacity to support deeper learning.

AI can generate educational content. Teachers create the conditions in which that content becomes meaningful learning.

11. A Whole-School Responsibility: Why AI Literacy Cannot Depend on Individual Teachers

Responsible AI integration requires more than individual experimentation.

Schools need coherent expectations, appropriate safeguards, professional learning, and shared educational principles.

Without these conditions, students may encounter contradictory messages.

One teacher may prohibit AI entirely.

Another may encourage extensive use.

A third may allow AI-assisted research without clarifying expectations for verification or acknowledgment.

Such inconsistency can create confusion about academic integrity, acceptable support, and the purpose of learning tasks.

School leaders should therefore consider several priorities.

Develop age-appropriate AI guidelines.

Expectations should reflect students' developmental stages, local requirements, approved tools, and safeguarding considerations.

Younger learners may require adult-mediated experiences rather than independent interaction with general-purpose AI platforms.

Prioritize data privacy and student protection.

Schools should assess tools before approving them, avoid unnecessary disclosure of student information, and establish clear expectations for handling personal or sensitive data.

Provide meaningful professional development.

Teachers need opportunities to investigate AI's educational implications, evaluate tools, redesign learning experiences, and share effective practices.

Reconsider assessment and academic integrity.

Policies should distinguish appropriate support from misrepresentation of work while maintaining reasonable, transparent expectations.

Monitor impact rather than adoption alone.

Schools should not assume that increased technology use demonstrates educational improvement.

Leaders need to investigate whether AI integration supports greater understanding, independence, engagement, accessibility, and critical thinking.

A school-wide AI strategy should therefore begin with educational purposes, not purchasing decisions.

12. Five Questions School Leaders Must Ask Before Integrating AI

For schools developing AI policies or professional learning plans, five questions deserve particular attention.

1. Are we using AI to enhance thinking or merely accelerate task completion?

Efficiency has value, but faster production does not automatically indicate deeper learning.

2. Can students explain their learning independently of the technology?

If learners cannot justify their conclusions or demonstrate their understanding without AI, the school may need to reconsider how tasks are designed and assessed.

3. Are teachers equipped to evaluate AI-generated content critically?

Responsible implementation depends on educators who understand the possibilities, limitations, and risks of the tools they use.

4. Does our AI strategy promote equity, accessibility, and learner agency?

AI should not widen existing inequalities or undermine students' confidence in their own intellectual abilities.

5. What evidence will demonstrate that AI is improving learning?

Schools need to evaluate educational impact, not simply the frequency of AI use or the sophistication of AI-generated products.

These questions can provide a foundation for professional inquiry, collaborative planning, and school improvement.

Final Reflection: The Future Belongs to Those Who Can Think, Not Simply Prompt

Artificial intelligence presents education with an extraordinary opportunity.

It can expand access to explanations, support personalized learning, introduce alternative perspectives, and help educators respond to diverse learning needs.

However, it also presents a fundamental challenge.

When students can obtain almost any explanation instantly, will they still develop the patience to investigate, the confidence to question, and the judgment to determine what is true?

When technology can construct persuasive arguments, will students still learn to build their own?

When AI can generate solutions, will learners still understand how to solve problems independently?

The answer depends partly on the decisions educators and school leaders make today.

We should not respond to AI by rejecting technological progress or expecting students to learn in a world that no longer exists.

Neither should we embrace AI so enthusiastically that we unintentionally remove the intellectual effort through which understanding develops.

Instead, we must be deliberate.

We must design learning experiences in which students question before accepting, investigate before concluding, and reflect before relying on automated responses.

We must create assessments that value reasoning, understanding, and intellectual ownership.

And we must support teachers in developing the professional judgment needed to integrate AI responsibly.

Because the ultimate purpose of education has never been simply to produce answers.

It is to develop human beings who can reason, question, create, evaluate, and make informed decisions.

AI may change how answers are generated. It must not diminish our commitment to developing the minds that evaluate them.

The defining question for the next generation of educators is therefore not:

"How can we teach students to use AI?"

It is:

"How can we ensure students continue to think deeply, independently, and critically in a world where AI can do so much for them?"

That is the challenge.

And perhaps it is also one of the greatest educational opportunities of our time.

About Eureka

Eureka | Enlightening Minds. Elevating Education.

Eureka supports teachers, school leaders, and educational institutions through research-informed professional development and educational consultancy services.

We believe meaningful educational transformation emerges when innovative technologies, educational research, professional inquiry, and thoughtful leadership come together to enhance human learning.

Our commitment is to help educational communities navigate change while preserving what matters most: critical thinking, learner agency, intellectual curiosity, and meaningful understanding.

Website: https://eureka-edu.org

References and Further Reading

1. OECD. (2026). OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education. OECD Publishing. https://doi.org/10.1787/062a7394-en

2. Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26). https://doi.org/10.1073/pnas.2422633122

3. Lee, H.-P., et al. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3706598.3713778

4. UNESCO. (2024). AI Competency Framework for Students. https://www.unesco.org/en/articles/ai-competency-framework-students

5. UNESCO. (2024). AI Competency Framework for Teachers. https://www.unesco.org/en/articles/ai-competency-framework-teachers

6. UNESCO. (2023). Guidance for Generative AI in Education and Research. https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research

7. International Baccalaureate Organization. (2026). Shaping Our Approach to AI as a Community. https://ibo.org/news/news-about-the-ib/ib-ai-principles/

8. International Baccalaureate Organization. Artificial Intelligence in Learning, Teaching, and Assessment. https://ibo.org/programmes/artificial-intelligence-ai-in-learning-teaching-and-assessment