AI can generate an answer.

Students still need to decide whether the answer is any good.

That distinction is becoming increasingly important in higher education.

Generative AI can help students brainstorm, explain concepts, summarize information, revise writing, and produce polished content.

But critical thinking requires something different.

Students need to compare ideas, recognize weak reasoning, evaluate evidence, make decisions, explain their judgment, and improve their thinking based on feedback.

A 2026 systematic review of 67 empirical studies on ChatGPT and higher-order thinking found that AI was more likely to support critical and creative thinking when it was embedded in structured, inquiry-oriented learning experiences. When AI use was less structured, researchers identified risks such as cognitive offloading and reduced critical engagement (Li et al., 2026).

The challenge, then, is not simply deciding whether students should or should not use AI.

It is designing learning experiences where students still have to think, evaluate, explain, and make decisions when AI is available.

That is why AI transparency and peer assessment can be particularly powerful when used together.

VisibleAI helps instructors understand how students created their own work.

Kritik360 gives students repeated practice judging the quality of someone else's work.

Together, they make more of the learning process visible.

VisibleAI: How Was the Work Created?

VisibleAI is designed around a simple question:

How was this work developed, and what role did AI play?

Students complete their work inside a transparent writing environment.

Instructors can review information including:

• The writing process
• Original student text
• Pasted content
• AI-generated content
• AI-assisted revisions
• AI prompts and responses
• Assignment-level AI rules

The purpose is not simply to identify AI.

It is to understand the role AI played in producing the work.

That shift toward process evidence is becoming increasingly relevant as the final submission alone provides less information about how a student actually developed an assignment.

Recent research on writing-process transparency in GenAI-affected courses examined an approach in which students provided evidence such as outlines, drafts, revision traces, and literature-search records. The researchers found that transparency depended not only on what evidence students provided but also on how instructors reviewed that evidence and whether students had opportunities to explain their work (Zhang et al., 2026).

The final document shows what a student submitted.

The process can provide more context about how they got there.

Kritik360: Can the Student Judge Quality?

Kritik360 asks another question:

Can the student tell strong work from weak work?

Students typically evaluate 3 to 4 classmates using an instructor-defined rubric.

That means they cannot simply produce their own answer and move on.

They have to look closely at other approaches, apply criteria, make a judgment, and explain why.

They also receive feedback from multiple peers and rate how helpful that feedback was.

This creates accountability on both sides of the process.

Students learn by creating, but they also learn by judging.

There is research behind that idea.

A meta-analysis of online peer assessment and higher-order thinking found a significant positive overall effect on higher-order thinking, with particularly strong effects for areas including critical thinking, reasoning, and reflection (Zhan et al., 2023).

That matters in the age of AI because students increasingly need more than the ability to produce an answer.

They need the judgment to recognize whether an answer is actually good.
‍

Why Those Two Processes Belong Together

Consider a student who uses AI to help develop an assignment.

VisibleAI can help provide context around questions such as:

How much of the work was original?

What did the student ask AI?

Did they paste generated content?

How did the document change over time?

But authorship is only one dimension of learning.

The student should also be able to look at another person's response and determine:

Is the argument strong?

Does it answer the question?

Is the evidence convincing?

What is missing?

What would make it better?

That is where peer assessment adds another layer.

VisibleAI makes creation more transparent.

Kritik360 makes judgment more visible.

One Assignment, Start to Finish

A combined activity can follow a simple workflow.

1. Write

Students create their assignment in VisibleAI under the instructor's AI policy.

2. Show the Process

Typing, revisions, AI prompts, and other authorship information remain connected to the work.

3. Evaluate Peers

Students move into Kritik360 and assess classmates using the instructor's rubric.

4. Receive Feedback

Each student receives several perspectives on their work.

5. Grow Judgment

Students compare feedback, reflect on different approaches, and become better at recognizing quality.

The instructor remains in control throughout the process and can review or adjust grades when necessary.

The goal is not simply:

Write something. Review something. Submit it.

It is to create a cycle of creation, evaluation, feedback, reflection, and improvement.
‍

NC State: AI as a Guide, Not a Ghostwriter

A Fall 2025 graduate engineering course at NC State University provides a useful example of this combined approach.

The course included 18 graduate students from four engineering programs studying physics modeling with neural networks.

The instructor wanted students to learn how to work with AI without letting it replace their understanding.

Students completed two 500-word research proposals, one near the beginning of the term and another near the end.

VisibleAI's AI helper was constrained so it could ask questions, redirect students, clarify concepts, and point toward resources.

It was not intended to simply write the proposal.

Three classmates evaluated each proposal in Kritik360.

Internal course data showed a noticeable change in the writing process from the first proposal to the second:

• ‍Original writing increased from 73% to 91%.
• AI-written content decreased from 20% to 6%.
• Pasted content decreased from 7% to 3%.

Peer-assessed scores remained high and increased slightly from 91.8 to 92.2 out of 100.

Students also reported that the AI's questions helped them brainstorm and communicate their ideas more clearly.

This course example involved only 18 students and did not include a comparison group, so these results should not be interpreted as evidence that VisibleAI or Kritik caused the changes observed.

Instead, the example illustrates what an AI-ready assessment design can look like.

Students were allowed to use AI.

They could not simply hand the thinking over to it.

And they still had to judge the work of their peers.
‍

AI Should Support Thinking, Not Make Thinking Optional

The distinction between guided and unguided AI use is important.

The review by Li et al. (2026) found that positive critical-thinking outcomes were more consistently associated with learning environments that included structured inquiry, reflection, argumentative reasoning, and metacognitive engagement.

In other words, access to AI alone is not an instructional strategy.

What students are asked to do with the output matters.

Do they accept the first response?

Or do they question it?

Do they verify information?

Can they identify weak reasoning?

Do they compare alternatives?

Can they explain why they accepted one suggestion and rejected another?

Can they improve what AI produced?

These are the activities that keep students intellectually involved.
‍

Critical Thinking Happens When Students Judge

One of the strongest reasons to use peer assessment in the age of AI is that evaluation itself can be a learning activity.

Students learn what good work looks like by seeing different examples.

They become more familiar with the rubric because they have to apply it.

They encounter mistakes, ideas, and approaches they may not have considered in their own submission.

And when they return to their own work, they have a stronger reference point for quality.

The meta-analysis by Zhan et al. (2023) supports this connection. Across the studies included in the analysis, online peer assessment had a significant positive effect on higher-order thinking, with stronger effects reported for convergent forms of thinking such as critical thinking, reasoning, and reflection.

That is particularly relevant when AI can generate work that sounds polished.

Students do not only need the ability to produce an answer.

They need enough judgment to recognize whether that answer deserves to be trusted.
‍

What Kritik Students Report

Student feedback collected by Kritik between Winter 2021 and Winter 2025 also reflects this experience.

86% agreed that Kritik helped them learn the course material.

83% agreed that their critical thinking improved.

These are student-reported perceptions, not evidence that Kritik caused those outcomes.

But they align with the broader research finding that peer assessment can give students repeated opportunities to evaluate quality, apply criteria, reflect, and engage with perspectives beyond their own.

Transparency Without Policing

The combined workflow can also change the academic integrity conversation.

VisibleAI gives instructors greater insight into how work was created before students begin evaluating one another.

This allows AI expectations to become part of the assignment design rather than something addressed only when there is a suspected violation.

Students can know:
‍

• ‍What AI use is permitted.
• What parts of the process are visible.
• What they are expected to do themselves.
• What they may need to explain.

Research on GenAI-related writing transparency reinforces the importance of this approach.

Zhang et al. (2026) compared two approaches to reviewing GenAI-related writing concerns. The approach that emphasized explanation, student response, and opportunities for correction was associated with greater writing-process transparency, stronger perceptions of procedural fairness, and lower anxiety about false accusations than an approach centered more heavily on detector thresholds.

The study did not evaluate VisibleAI, and it should not be interpreted as product validation.

But it supports an important principle:

Evidence should start a conversation, not automatically make the decision.

Technology can provide context.

The instructor still decides what that context means.
‍

Preparing Students for Work Beyond the Classroom

Students are likely to encounter AI throughout their future careers.

They are also likely to encounter situations where they need to:

• Evaluate someone else's work
• Give constructive feedback
• Compare alternatives
• Defend a recommendation
• Identify weak evidence
• Recognize incomplete information
• Decide whether an AI-generated response can be trusted
• Take responsibility for a final decision

That means AI literacy cannot stop at knowing how to write a good prompt.

Students need to know how to use AI.

They also need to know when to question it.

They need to know how to generate possibilities.

They also need to know how to choose among them.

AI can contribute information.

The student still needs to own the judgment.
‍

Authentic Work. Real Critical Thinking.

AI will continue to make content easier to generate.

That makes the student's ability to judge, explain, evaluate, verify, and improve even more important.

VisibleAI gives instructors visibility into how students create.

Kritik360 gives students practice evaluating what others create.

One makes the process of creation more visible.

The other makes the process of judgment part of learning.

Together, they shift assessment away from focusing only on the final answer and toward something much more valuable:

evidence that students are still doing the thinking the assignment was designed to develop.
‍

Explore Kritik360 + VisibleAI

Create assessment experiences that combine authorship transparency, responsible AI use, peer feedback, and critical thinking.

CTA: Explore Kritik360 + VisibleAI

‍

‍