BlogResearch summary
How Does Generative AI Affect Critical Thinking?
When AI writes an email, finds information, or solves a problem for us, it saves effort. But does it also change how much we think?
A 2025 study by Hao-Ping (Hank) Lee and colleagues, titled The Impact of Generative AI on Critical Thinking, examined when people think critically while using generative AI, what makes them question its answers, and whether AI changes the mental effort involved.
What did the researchers study?
Previous research had explored AI's impact on critical thinking in education and professions such as nursing and clinical psychology. But less was known about everyday knowledge work, where people rely primarily on their expertise rather than physical labour.
The researchers surveyed 319 knowledge workers from different professions to answer two questions:
- When and how do people think critically while using GenAI?
- When and why does GenAI increase or decrease the effort required for critical thinking?
They used Bloom's framework, which describes six types of thinking: remembering, understanding, applying, analysing, combining, and evaluating information.
But how can critical thinking be measured when thinking itself cannot be directly observed?
The researchers focused on actions that indicate critical thinking, such as checking the tone of an AI-generated email, verifying computer code, or identifying possible bias in an analysis.
Before the survey, participants were shown such examples to help them understand what critical thinking meant. They then answered questions about their tasks, their confidence in AI and themselves, and their use of critical thinking.
The researchers analysed both numerical responses and written explanations to identify patterns and understand participants' experiences.
When do people think critically while using AI?
The study identified three main activities:
Defining the task: People clarified their goals and refined the instructions they gave AI.
Evaluating the response: They checked whether AI-generated results were accurate and met their requirements.
Using the response: They selected, modified, and incorporated AI-generated material into their work.
Three factors encouraged critical thinking: maintaining work quality, avoiding negative consequences, and developing professional skills.
Three barriers discouraged it. Some people did not recognise the need to question AI, particularly when they trusted it or considered a task unimportant. Others lacked the time or motivation. Some lacked the expertise needed to identify errors or improve AI-generated responses.
What role does confidence play?
One important finding concerned confidence.
People who trusted AI more were less likely to report thinking critically while using it. Conversely, people who were more confident in their own ability to complete a task tended to report more critical thinking.
The researchers also found that people who regularly reflected on their work without AI were more likely to continue doing so when using AI.
These findings describe associations between confidence and reported behaviour, rather than proving that one causes the other.
Does AI reduce the effort of critical thinking?
The second research question examined whether people felt critical thinking required more or less effort when using AI.
Overall, participants reported less effort with AI than without it.
But this happened for different reasons. Some felt supported by AI, some delegated parts of their thinking to it, and others reported doing less critical thinking altogether.
Participants also sometimes confused a reduction in overall workload with a reduction in critical thinking effort. Completing a task more easily does not necessarily mean that evaluating it requires less thought.
Greater confidence and trust in AI were also associated with lower perceived effort.
The researchers identified an important distinction: AI does not necessarily eliminate critical thinking. It can change where people direct their mental effort.
They observed three shifts:
From gathering information to verifying it: AI can quickly provide information, leaving people to check its accuracy.
From solving problems to integrating solutions: Rather than developing solutions entirely themselves, people adapt AI-generated answers to their needs.
From executing tasks to overseeing them: As AI performs more of the work, people focus on guiding it and evaluating the results. The researchers call this task stewardship.
These findings suggest that the reduction in perceived effort does not always mean people are thinking less. In some cases, they are performing different kinds of thinking.
How can AI better support critical thinking?
Based on their findings, the researchers suggest several approaches for designing AI tools.
Strengthen users' own abilities and confidence. AI systems could allow people to control how much assistance they receive, depending on their expertise and the difficulty of the task. They could also help users judge the reliability of AI-generated responses.
Help people recognise when critical thinking is needed. Some systems could proactively alert users when an AI response deserves closer examination. Others could let users request critical evaluation whenever they feel it is necessary.
Make critical thinking part of professional development. People may be more motivated to examine AI output when doing so helps them learn. Tools could support this by explaining responses, suggesting improvements, and guiding users through evaluating results.
Develop skills suited to AI-assisted work. Training could place greater emphasis on verifying information, adapting AI-generated content, and overseeing AI-assisted tasks.
The researchers also stress the importance of preserving foundational skills, including gathering information and solving problems independently, to avoid excessive reliance on AI.
What does the study tell us?
The study identifies how knowledge workers report using critical thinking with GenAI, what encourages or discourages them, and why AI can make thinking feel less demanding.
It also reveals a shift in the nature of critical thinking: from gathering information to checking it, from producing solutions to adapting them, and from performing tasks to overseeing AI-generated work.
These findings offer directions for designing AI tools that support critical thinking while helping people work efficiently.
The study examines participants' reported experiences, not measured changes in their abilities over time. It therefore does not establish whether AI makes people better or worse critical thinkers.
Instead, it helps explain how the practice of critical thinking is changing as AI becomes part of everyday work.
Source: 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. CHI 2025.