Artificial intelligence has moved from “future trend” to everyday classroom reality. Tools like ChatGPT can generate explanations, outlines, counterarguments, and feedback in seconds. For educators, that speed can feel like both a breakthrough and a threat: Will students become stronger thinkers, or will they outsource thinking altogether?
A 2025 systematic review published in Educational Process: International Journal (Melisa et al., 2025) helps clarify what we currently know. The authors analyzed 19 peer-reviewed studies (from 2023–2024) drawn from Scopus and ERIC to understand two core questions:
- How does ChatGPT affect the development of critical thinking skills in higher education?
- How does ChatGPT affect students’ ability to evaluate information and form independent judgments?
While the research focuses on higher education, the implications matter for K–12 systems too—especially because the habits students build now will shape how they learn, write, and reason later. For school leaders, special education teams, and student support professionals (including online therapy providers like TinyEYE), the message is practical: AI can be a powerful learning support, but only when paired with explicit instruction in critical thinking and verification.
Why critical thinking is the skill AI can’t replace
Critical thinking is commonly described as the ability to evaluate information, reason logically, and make sound judgments. The review highlights well-established components of critical thinking—skills such as interpretation, analysis, evaluation, inference, explanation, and self-regulation (often associated with frameworks like Facione’s).
In plain terms, critical thinking is what helps students:
- Spot weak evidence or missing context
- Compare multiple viewpoints fairly
- Explain “why” they believe something, not just “what” they believe
- Recognize bias (including their own)
- Make decisions based on verified facts
These are exactly the skills students need when AI can produce polished, confident-sounding answers that may be incomplete, biased, or incorrect.
What the research says: ChatGPT can strengthen critical thinking—under the right conditions
Across the 19 studies, many findings point to real benefits. ChatGPT can support critical thinking development by making it easier for students to explore ideas, test arguments, and access diverse perspectives quickly.
In practice, this often shows up in three ways:
1) Faster access to multiple perspectives
Several studies in the review describe how ChatGPT can broaden students’ viewpoints. When students can ask for counterarguments, alternative explanations, or examples, they may engage in deeper comparison and analysis than they would with a single textbook passage or a quick web search.
2) Support for analysis and argument construction
ChatGPT can help students structure their thinking. For example, in writing contexts (including EFL writing), AI feedback and idea generation can help students refine clarity, organization, and reasoning. In STEM and problem-solving contexts, some studies report that students can use AI outputs to compare solution paths and reflect on errors.
3) Increased engagement through interactive “dialogue”
Unlike static resources, ChatGPT can simulate a back-and-forth conversation—prompting students to clarify, expand, or defend their thinking. When teachers design the activity well, this can encourage inquiry and reflection.
From a school-system perspective, these benefits are especially relevant for students who need additional scaffolding—such as learners with language-based learning needs, executive function challenges, or anxiety around initiating tasks. With the right guardrails, AI can reduce “blank page” paralysis and increase students’ willingness to start.
The risk side: over-reliance can reduce reflection, motivation, and verification habits
The review is equally clear about the downsides. The biggest concern isn’t that AI exists—it’s how students may use it when expectations and instruction are unclear.
Common risks highlighted across studies include:
- Reduced self-reflection: If students accept AI responses immediately, they may skip the mental work that builds durable understanding.
- Weaker critical evaluation: Students may trust confident-sounding outputs without checking sources or accuracy.
- Motivation loss: If AI becomes the default “answer machine,” students may feel less need to practice writing, reasoning, or problem-solving skills.
- Academic integrity concerns: Plagiarism, authenticity of assessments, and unclear boundaries can create confusion for students and staff.
- Bias and inaccuracies: AI may generate misleading or biased content, and students may not recognize it without strong media literacy.
- Privacy and data concerns: Some studies note the need to consider data privacy, especially when students use third-party tools.
For K–12, these risks can be amplified because younger students are still developing metacognition (thinking about their thinking). That means schools can’t assume students will “just know” when to doubt AI. They need to be taught.
Independent judgment: AI can help students decide—but only if they stay in the driver’s seat
A major theme in the review is that ChatGPT may support independent judgment when students use it as a tool for exploration rather than a replacement for decision-making.
Some studies reported that students felt more confident making academic decisions after using ChatGPT to gather and compare information. However, other findings suggest that if students use AI mainly to finish tasks quickly, they may not develop the self-regulation required for independent judgment.
A useful way to frame this for students is:
- AI can suggest options.
- You decide what’s true, what’s relevant, and what you believe.
What schools can do now: practical guidance to build “AI-ready” critical thinkers
Melisa et al. (2025) conclude that educators must provide guidance for responsible use—especially around verification and critical judgment. Translating that into school practice, here are concrete steps leadership teams can implement.
1) Teach verification as a required step (not an optional “extra”)
Make “check it” part of the assignment rubric. For example:
- Require students to cite and summarize two non-AI sources that confirm or challenge the AI output
- Ask students to highlight one claim from the AI response and verify it with evidence
- Include a short reflection: “What might be wrong or missing here?”
2) Redesign tasks so critical thinking is visible
If an assignment only rewards the final product, students will be tempted to outsource it. Instead, assess the process:
- Draft checkpoints with teacher feedback
- Oral defenses (“Explain why you chose this evidence”)
- Comparison tasks (“Here are two answers—evaluate which is stronger and why”)
3) Set clear boundaries for acceptable AI use
Students do better when expectations are specific. Schools can define categories such as:
- Allowed: brainstorming, outlining, generating practice questions, rewriting for clarity with student-provided content
- Allowed with citation: AI-assisted feedback, AI-generated examples used for critique
- Not allowed: submitting AI-generated work as original, using AI during closed assessments (unless explicitly permitted)
4) Support educators with training, not just rules
Teachers are being asked to manage fast-moving change. Professional learning can focus on:
- How to design AI-resilient assessments
- How to teach prompt literacy without encouraging shortcut behavior
- How to spot common AI errors (hallucinations, overconfidence, vague sourcing)
5) Don’t forget student wellness and learning supports
AI adoption isn’t only an academic issue—it’s also a support issue. Students may use AI to cope with stress, workload, or language barriers. Schools should pair AI guidance with human support systems that build confidence and skills.
This is where services like TinyEYE can align with school goals: online therapy and student support can reinforce self-advocacy, executive functioning strategies, and communication skills—so students rely less on “quick fixes” and more on sustainable learning habits.
Bottom line: the goal isn’t to ban AI—it’s to teach students how to think with it
The systematic review by Melisa et al. (2025) paints a balanced picture. ChatGPT can enhance critical thinking by helping students access perspectives, analyze information, and build arguments. But it can also weaken critical thinking when students over-rely on it, fail to verify content, or use it to bypass reflection.
The most important takeaway for schools is straightforward: AI literacy must include critical thinking instruction. When educators explicitly teach verification, self-regulation, and ethical use, AI tools can become a catalyst for deeper learning rather than a shortcut around it.
For more information, please follow this link.
