Artificial Intelligence and Youth: Cognitive, Educational, and Behavioral Impacts

Abstract

Background: Artificial Intelligence (AI) and Generative AI (GenAI) are increasingly integrated into educational and professional settings, offering personalized learning, productivity gains, and enhanced engagement. However, excessive reliance may compromise critical thinking, autonomous problem-solving, and emotional regulation among youth (i.e., adolescents and young adults) and early-career professionals. Aim: This review examines the cognitive, educational, and behavioral impacts of AI and GenAI use in youth, highlighting implications for their responsible integration in learning and professional development. Methods: A narrative review was conducted, synthesizing empirical studies, psychometric instruments, and international policy frameworks addressing AI engagement. Emphasis was placed on cognitive, behavioral, educational, and ethical dimensions across youth and early-career professionals. Results: AI enhances learning efficiency, creativity, and professional decision-making but may also foster cognitive offloading, dependency, and addiction-like behaviors. Instruments such as the Conversational AI Dependence Scale (CAIDS) and the Problematic ChatGPT Use Scale (PCGUS) help identify maladaptive patterns. Effective strategies include structured pedagogy, human oversight, reflective practice, AI literacy, and ethical guidance. Paradoxically, higher AI competence and trust may increase reliance, underscoring the need for guided and balanced engagement. Conclusions: Responsible AI integration requires multidimensional approaches combining instructional scaffolding, metacognitive strategies, supervision, and governance to preserve autonomy, professional judgment, and cognitive development in youth.

1. Introduction

1.1. The Global Diffusion of AI

Artificial Intelligence (AI) has emerged as a transformative technology reshaping society at multiple levels. Across industries—from healthcare and finance to education and creative arts—AI applications are becoming pervasive, influencing decision-making, workflow automation, and the creation of new services. Generative AI tools, in particular, have accelerated content production and personalized digital interactions, contributing to rapid uptake in both professional and everyday contexts [1]. Recent research shows that global AI diffusion continues to grow, with generative AI tools being adopted faster than many earlier general-purpose technologies, highlighting the unique speed of this transformation [1].

However, this diffusion is uneven across regions and populations, largely shaped by infrastructure, digital skills, economic capacity, and policy environments [2]. Global reports indicate that while about one in six people worldwide now use a generative AI product, adoption rates are significantly higher in digitally advanced regions compared with under-connected areas, revealing persistent digital divides in access to AI technologies [3]. Research on determinants of generative AI adoption also shows that IT infrastructure, economic stability, and workforce readiness are critical drivers of uptake across countries [4], while barriers such as regulatory uncertainty and skills shortages slow AI integration in less developed regions [2,3].

Demographic patterns further nuance this picture. Survey data reveal that younger and more digitally literate cohorts tend to engage with AI tools more frequently than older ones, reflecting broader trends of technology adoption that align with digital fluency and comfort with emerging interfaces [5]. In educational contexts, large studies report that AI adoption among students is substantial and appears to rise with educational level: one survey found generative AI usage increasing from ~15% in younger adolescents to over 50% in older adolescents, indicating that older students engage with AI tools for schoolwork far more frequently than younger peers [6]. Taken together, these observations indicate that AI’s global penetration is both deep and complex—an interplay of technological capability, infrastructure, policy, and human behavior.

As AI tools become embedded in everyday life, used for tasks ranging from research support to creative generation, individuals increasingly depend on them for problem-solving, information retrieval, and productivity tasks. This backdrop sets the stage for examining how these technologies interact with the youth population, whose developmental, educational, and social environments may render their relationship with AI qualitatively distinct from that of older users.

1.2. Youth in AI Diffusion

Young people occupy a distinctive position within the rapid diffusion of artificial intelligence. Adolescents and young adults have grown up in environments characterized by continuous digital immersion, early exposure to algorithm-mediated tools, and high engagement with interactive technologies. This digital fluency shapes how they approach innovations like AI and contributes to relatively high adoption rates among youth compared with older age groups [7].

Empirical research using passive sensing data shows that generative AI use among youth is already significant: in a large US cohort of 6488 children and adolescents aged 4–17, nearly one-third (31.9%) had used generative AI apps on their devices, with usage highest among teens aged 15–17 years (50.4%) and substantial use also found among younger preteens [7]. These patterns highlight that AI engagement among youth is both widespread and age-graded, increasing across adolescent developmental stages.

Beyond sheer prevalence, psychological and contextual factors influence how young people interact with AI. A study of Italian high school students found that performance expectations, social influences, and habitual internet behaviors were significant predictors of AI use for schoolwork, with habitual problematic internet use showing a spillover effect into more sustained AI engagement [8]. This suggests that youth are not only adopting AI tools because they are accessible, but because specific motivational and behavioral traits shape their usage patterns.

Survey research also indicates that youth engage with generative AI not only for schoolwork but in informal everyday contexts. National surveys show that many teenagers and young adults report using AI tools for brainstorming, homework support, and other tasks, and that familiarity and comfort with these tools often exceed that of parents or teachers [9,10]. For example, around half of adolescents aged 14–22 report having used generative AI tools at least once, and a notable subset engage with these technologies regularly [9,10].

In addition, research exploring youth perspectives on generative AI in health-related decision contexts reveals that many young people are already forming distinct expectations and concerns about how AI might influence future roles, reflecting not only patterns of use but emerging attitudinal positions toward AI in domains beyond education [11].

Taken together, these findings indicate that young people’s engagement with AI is shaped by a convergence of developmental, cognitive, behavioral, and social influences, distinguishing their adoption patterns from those of older users.

1.3. Opportunities and Challenges for Youth

The integration of artificial intelligence (AI) into educational and social contexts presents a complex landscape of opportunities and challenges for young users. On the positive side, generative AI tools have demonstrated considerable potential to enhance learning, creativity, and engagement across age groups. Research shows that AI applications can support the development of critical thinking, problem-solving, and digital literacy skills, particularly when used as co-creative partners in writing, project-based tasks, and personalized learning experiences [12,13]. Moreover, AI-enabled educational interventions have been associated with increased motivation, higher engagement in classroom activities, and improved retrieval and application of knowledge, suggesting that these technologies can be leveraged to foster cognitive growth in students [14,15].

However, alongside these opportunities, the rapid and widespread adoption of AI also introduces several challenges, especially for adolescents and young adults who are still developing cognitive, emotional, and self-regulatory capacities. Evidence indicates that frequent reliance on AI for problem-solving and content generation may alter learning behaviors, potentially reducing independent thinking, creativity, and confidence in one’s own abilities [12,13]. In addition, the ethical, social, and cognitive implications of AI use are still under investigation; concerns include the potential for digital burnout, over-reliance on algorithmic outputs, and increased vulnerability to misinformation or biased recommendations [14]. Patterns observed in analogous technological behaviors, such as smartphone overuse or social media engagement, suggest that youth may be particularly susceptible to habitual or compulsive interaction with AI tools, highlighting the need for awareness, guidance, and structured use [15].

Taken together, these findings underscore a dual reality: AI offers powerful tools for learning and skill development, but unmoderated or pervasive use may contribute to emerging forms of problematic engagement among youth. This tension establishes a clear rationale for research exploring how young people interact with AI, which factors promote healthy versus maladaptive use, and how interventions might optimize benefits while mitigating risks [12,13,14,15].

1.4. Key Questions and Aim

Given the rapid adoption of AI technologies among youth in educational or early professional contexts—and the emerging evidence of both opportunities and potential risks—several questions naturally arise. How do young people engage with AI tools in their daily learning, creative, and professional practices? Which cognitive, emotional, social, or environmental factors contribute to healthy versus problematic use? To what extent might over-reliance on AI impact independent thinking, creativity, and self-confidence, and how does this compare with other digital behaviors such as smartphone overuse or excessive social media engagement? Furthermore, what strategies, educational frameworks, or policy measures can support balanced and responsible AI integration, fostering skill development while minimizing potential harms?

The aim of this review is to provide an exploratory synthesis of the current literature on AI engagement among youth, with a particular focus on problematic or potentially maladaptive patterns of use. By drawing together findings from educational, psychological, and digital behavior research, this work seeks to identify key thematic areas, highlight gaps in current understanding, and outline directions for future investigation. Rather than generating new empirical data, the review maps the evolving landscape of AI use in youth, providing a conceptual foundation to inform educators, policymakers, and researchers concerned with the responsible integration of AI technologies.

2. Design and Organization of the Narrative Study

2.1. Methodology of the Narrative Review

This narrative review aims to provide a comprehensive overview of problematic or excessive engagement with Artificial Intelligence (AI) among youth (i.e., adolescents and young adults). Unlike systematic reviews, the narrative approach allows the integration of quantitative, qualitative, and conceptual studies, as well as systematic and narrative reviews, theoretical frameworks, and policy analyses, offering a nuanced understanding of cognitive, emotional, behavioral, and social implications of AI engagement. This approach enables synthesis across educational, recreational, and informal contexts, highlighting opportunities and challenges, integrating foundational work with recent studies, and mapping emerging trends.

To capture the relevant literature, searches were conducted in PubMed, Scopus, and Web of Science. The narrative review followed a standard checklist for narrative reviews (ANDJ).

Search queries included, but were not limited to, terms related to excessive or problematic engagement with Artificial Intelligence, such as AI overuse, AI dependency, generative AI dependency, habitual or compulsive AI use, reliance on AI, and AI addiction, as well as closely related expressions referring to cognitive, behavioral, or emotional consequences of AI use. The composite key was:

(“AI overuse”[Title/Abstract] OR “GenAI overuse”[Title/Abstract] OR “Generative AI overuse”[Title/Abstract] OR “AI dependency”[Title/Abstract] OR “GenAI dependency”[Title/Abstract] OR “Generative AI dependency”[Title/Abstract] OR “compulsive AI use”[Title/Abstract] OR “compulsive GenAI use”[Title/Abstract] OR “compulsive Generative AI use”[Title/Abstract] OR “reliance on AI”[Title/Abstract] OR “reliance on GenAI”[Title/Abstract] OR “reliance on Generative AI”[Title/Abstract] OR “addiction to AI”[Title/Abstract] OR “addiction to GenAI”[Title/Abstract] OR “addiction to Generative AI”[Title/Abstract] OR “AI addiction”[Title/Abstract] OR “GenAI addiction”[Title/Abstract] OR “Generative AI addiction”[Title/Abstract] OR “Artificial intelligence addiction”[Title/Abstract])

Snowballing of references from retrieved studies was also performed to ensure comprehensive coverage of the literature. During this process, additional studies not exclusively focused on AI were considered when they provided relevant theoretical or empirical frameworks (e.g., behavioral addiction models, technology dependency, FoMO, or digital engagement patterns). These works were included to establish conceptual foundations for interpreting AI-related behaviors and were categorized accordingly in the thematic synthesis.

Eligible studies included research on youth in educational, social, recreational, or informal learning contexts. Studies focusing on educators and healthcare professionals were considered only when findings provided transferable insights into behavioral patterns, professional dependency, or cognitive and ethical implications relevant to youth. Both empirical research (quantitative surveys, experimental and longitudinal studies, qualitative interviews), conceptual work (narrative and systematic reviews, theoretical frameworks), and policy/professional analyses were included to provide a multi-faceted perspective. Studies in related populations or slightly different contexts were included if they offered insights applicable to AI-related behavioral, cognitive, emotional, or social phenomena.

Exclusion criteria were limited to non-peer-reviewed material and studies focused exclusively on industrial, purely technical, or professional AI applications lacking behavioral, educational, or developmental relevance.

This approach allowed the identification of studies addressing patterns of AI reliance and their potential impacts on learning outcomes, professional practice, decision-making, and psychological well-being.

Table 1. Inclusion and Exclusion Criteria.

Criteria	Inclusion	Exclusion
Population	Youth in educational, social, recreational, or informal learning contexts; educators or healthcare professionals if transferable insights	Adults outside the target demographic; purely professional/industrial contexts
AI Context	AI engagement in learning, social, recreational, or professional contexts with behavioral relevance	Purely technical, industrial, or professional AI use without behavioral or developmental relevance
Study Type	Quantitative (surveys, experiments, longitudinal), qualitative (interviews, focus groups), narrative reviews, systematic reviews, theoretical frameworks, policy/professional analyses *	Non-peer-reviewed studies
Language	English	Non-English
Publication Date	No restrictions	–
Outcomes	Cognitive, emotional, behavioral, social, or professional impacts of AI engagement	Studies not addressing AI engagement or outcomes relevant to youth
Criteria	Inclusion	Exclusion

* In addition to empirical studies, selected regulatory documents and authoritative manuals were included to provide contextual, normative, and conceptual grounding. These sources are used to corroborate findings and strengthen interpretations throughout the Section 8.

reports the inclusion and exclusion criteria.

A table of included studies with a brief summary of focus/domains is provided in the Supplementary Material, facilitating a broad, integrative understanding of the primary focus of the study (Table S1).

2.2. Themes and Structure

From a comprehensive analysis of the current literature on AI and Generative AI (GenAI) use among youth [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96], several recurring themes have emerged. Each study was assigned a dominant theme, representing its primary focus, and, where applicable, a secondary theme capturing an additional dimension addressed by the research. This dual categorization allows for a nuanced understanding of how AI impacts cognitive, educational, professional, and behavioral outcomes.

Table S1 details the selected studies, providing justification for the assigned focus areas along with a brief description of each study. To enhance interpretability, the table employs a color-coded scheme based on the dominant theme, enabling quick visual recognition of thematic clusters. Secondary themes are included to highlight additional, yet relevant, perspectives addressed by the studies, without overcrowding the classification. This structured presentation supports the Section 8, corroborating findings and reinforcing interpretations across the broader literature.

While AI offers substantial opportunities for learning, creativity, and skill development, frequent or unmoderated engagement may also introduce cognitive, behavioral, and ethical challenges. Across studies, it becomes clear that the impact of AI is multi-dimensional, affecting students’ academic performance, professional practices, psychological well-being, and the development of 21st-century skills.

Based on this synthesis, the literature can be organized into the following key thematic areas, each addressing a distinct aspect of AI use and potential overreliance. The literature highlights five major themes, developed in Section 3, Section 4, Section 5, Section 6 and Section 7, which can be outlined as follows:

• Cognitive, educational and academic impacts

AI and GenAI provide personalized learning opportunities and can enhance content production, engagement, and skill development [16,17,21]. Among university students and EFL learners, perceived AI dependency has been linked to higher levels of Fear of Missing Out (FoMO), negatively affecting reading comprehension and vocabulary acquisition [16]. Overreliance on GenAI may inflate students’ self-efficacy inaccurately, potentially undermining academic achievement, although supportive teacher behaviors can mitigate these effects [17].

• Professional Consequences

For educators and healthcare professionals, excessive AI use risks eroding critical thinking, professional autonomy, and intrinsic motivation [18,20]. AI overuse can also suppress creativity, reduce peer collaboration, and introduce ethical concerns related to plagiarism and the uncritical use of AI-generated content [18,20]. These findings underscore the importance of professional development, AI literacy, and structured integration strategies to ensure responsible and balanced adoption.

• Psychological and Behavioral Dimensions

Emerging evidence points to behavioral addiction-like patterns associated with GenAI, sometimes described as a potential new syndrome of AI dependency [22]. Personality traits such as neuroticism, self-critical perfectionism, and impulsivity can exacerbate dependency through needs frustration and negative academic emotions [25]. Additionally, reliance on Large Language Models (LLMs) may create the Illusion of Explanatory Depth (IoED), where students overestimate their understanding of complex topics and fail to develop independent problem-solving skills [26].

• AI Literacy, Trust, and 21st-Century Skills

AI literacy and trust are crucial in shaping both the extent of dependency and its impact on essential skills [21,23]. Paradoxically, teachers and pre-service professionals with higher AI literacy may demonstrate increased AI dependency, potentially reducing self-confidence, problem-solving, critical thinking, and collaboration skills [21,23]. These observations suggest that AI integration should be carefully balanced to foster skill development while avoiding cognitive outsourcing.

• Strategies for Responsible AI Integration

Practical recommendations across studies emphasize the importance of structured guidance and educational governance. Iterative prompting techniques, clearly defined limits for AI use, and hybrid approaches combining human oversight with AI assistance can help mitigate dependency risks [17,19,26]. Additionally, fostering teacher involvement, promoting digital wellness practices, and raising ethical awareness are essential strategies to maximize AI benefits while minimizing potential cognitive, psychological, and ethical drawbacks [16,17,18,19,20,21,22,23,24,25,26].

3. Cognitive, Educational and Academic Impacts

Among adolescents and young adults (youth), AI and Generative AI (GenAI) are increasingly integrated into educational contexts, offering opportunities for personalized learning, interactive content creation, and the development of digital and cognitive skills. These tools can facilitate engagement and motivation, provide immediate feedback, and support individualized learning pathways. However, growing reliance on AI also introduces risks of cognitive dependency, particularly when students habitually turn to AI for tasks that require active problem-solving, comprehension, or critical thinking.

The study reported in [16] highlights that perceived AI dependency is strongly associated with higher levels of Fear of Missing Out (FoMO) among university students and EFL learners. Students experiencing FoMO may feel pressured to constantly use AI to keep up with peers or complete tasks quickly, which can inadvertently reduce deep engagement with learning materials. This behavioral pattern has been linked to measurable decreases in reading comprehension and vocabulary acquisition, suggesting that reliance on AI tools can interfere with foundational cognitive processes such as memory consolidation, inferential reasoning, and self-monitoring during learning.

Further evidence from studies conducted between 2023 and 2025 reinforces these findings. Research on the emotional and motivational dimensions of AI use [27,28,29,30,31] suggests that FoMO and digital anxiety contribute to fragmented attention, procrastination, and occasional learning burnout. These effects are particularly pronounced in learners facing high-stakes academic tasks or in EFL contexts, where mastery of language skills requires consistent practice and reflection. Overreliance on AI for writing, reading comprehension exercises, or vocabulary development can result in students bypassing essential cognitive effort, leading to shallower understanding and reduced long-term retention.

Measurement studies [38,39,40,41] provide validated scales to quantify AI dependency and its impact on learning behaviors. These studies show consistent patterns: students who engage with AI frequently for academic tasks often exhibit lower self-regulation, reduced resilience to cognitive challenges, and higher susceptibility to motivational disruptions. Additionally, cross-sectional analyses [43,45] reinforce that habitual AI reliance correlates with lower reading comprehension and vocabulary acquisition among EFL learners, highlighting the cognitive dimension of dependency.

Importantly, the literature also identifies positive mediating factors. Students with structured guidance, explicit strategies for AI use, and teacher support demonstrate better balance, leveraging AI for efficiency while maintaining independent problem-solving and comprehension skills. For example, guided iterative prompting or scaffolded AI-assisted exercises allow learners to engage critically with AI-generated content, mitigating the risk of inflated self-efficacy and cognitive outsourcing.

These findings underscore a nuanced reality: AI and GenAI offer substantial benefits for engagement, personalized learning, and skill development, yet unmoderated use introduces cognitive vulnerabilities. Effective interventions involve not only monitoring usage frequency but also cultivating metacognitive awareness, fostering teacher involvement, and embedding AI use within structured, pedagogically sound frameworks. This ensures that learners can reap the benefits of AI while preserving essential cognitive processes critical for long-term academic success.

In summary, current scholarship primarily focuses on balancing the cognitive benefits of AI with potential dependency risks, particularly regarding problem-solving, comprehension, and self-regulation. Mainstream research has evolved in recent years to include empirical, longitudinal, and validated measurement studies, enabling a clearer mapping of underlying cognitive and motivational mechanisms. Emerging studies also indicate that structured and guided AI use can not only mitigate overreliance but also promote metacognitive awareness, autonomy, and transferable skills, offering a complementary perspective to the predominant focus on risks.

Figure 1 reports a sketch of the cognitive and educational impact of AI.

4. Professional Consequences of AI Use

This section addresses the professional consequences of AI and Generative AI (GenAI) use, focusing on effects related to professional competence, decision-making, ethical responsibility, and instructional quality. While pedagogical outcomes are acknowledged, they are discussed only insofar as they intersect with professional practice and the development of domain expertise.

Educators and healthcare professionals increasingly interact with AI tools during both training and professional practice. These technologies can enhance efficiency, creativity, and access to knowledge; however, excessive reliance may produce unintended consequences for professional competence, critical judgment, and ethical responsibility.

Habitual overuse of AI for lesson planning, content generation, or assessment can erode critical thinking, reduce professional autonomy, and diminish intrinsic motivation. Such cognitive outsourcing may compromise professional judgment and decision-making, ultimately affecting teaching quality and learners’ experiences [18,48,49,52,55].

In healthcare, overreliance on AI can reduce peer collaboration, suppress creative problem-solving, and introduce ethical risks such as uncritical acceptance of AI-generated recommendations [20,47,57]. While AI encodes substantial domain knowledge, improper reliance can reduce critical oversight, introduce bias, and impair decision-making [47,57]. Studies on digital health adoption and governance frameworks highlight the importance of ethical training, oversight, and organizational structures to safeguard professional standards [58,60,66].

Organizational and systemic factors also influence AI integration. Institutional support and professional development significantly shape AI adoption and collaborative engagement [53,54]. AI can optimize workflows and predictive accuracy, but only if professionals maintain critical oversight [56,62,64]. Regulatory frameworks, including GDPR and EU AI laws, are crucial for ethical and responsible use [66].

Evidence emphasizes the importance of early professional formation. Formative training in AI literacy, ethical reasoning, and collaborative practice reduces the risk of overreliance and fosters adaptive expertise. Structured interventions—such as guided iterative AI exercises, peer collaboration frameworks, and reflective practice—help professionals balance efficiency gains with critical judgment, autonomy, and ethical responsibility [47,48,50,51,52].

In summary, AI offers substantial benefits for professional efficiency, creativity, and knowledge access, but without deliberate governance, it risks undermining critical thinking, autonomy, and ethical standards. Effective interventions require a multilevel strategy: individual training in AI literacy and metacognitive skills, structured guidance for AI use, and organizational policies supporting ethical and professional standards. This approach ensures AI enhances professional capabilities without replacing the cognitive, ethical, and reflective skills essential for responsible practice.

Figure 2 provides a visual summary of the emerging evidence in this specific domain.

5. Psychological and Behavioral Dimensions

Recent studies suggest that interactions with AI and generative AI (GenAI) tools are influenced not only by professional factors but also by psychological traits and behavioral patterns. Personality, cognitive style, and habitual engagement with AI can shape the likelihood of overreliance and, in extreme cases, behaviors resembling behavioral addiction.

AI addiction syndrome has been described, highlighting behavioral markers similar to classical addiction: compulsive engagement with AI, prioritization of AI-mediated tasks over autonomous problem-solving, and emotional distress when AI tools are unavailable [22]. This phenomenon aligns with the broader literature on behavioral addiction, drawing parallels with gaming, social media, and compulsive online behaviors [70,71,76]. Neurobiological perspectives suggest that AI engagement can activate reward-related pathways, reinforcing habitual use [68,69].

Additional studies expand on these psychological dimensions:

• Generative AI chatbots can foster psychological dependence, particularly when perceived as companions or mentors [76].
• Frequent engagement with AI may reduce self-regulatory capacities, impacting mental well-being [76].
• Diagnostic criteria for behavioral addictions provide a conceptual foundation applicable to compulsive engagement with digital technologies, including AI [68,69].
• Professional contexts, such as healthcare or educational assessment, can influence susceptibility to habitual AI reliance, linking occupational habits to psychological patterns [76].

Collectively, these studies indicate a feedback loop: personality traits and habitual behaviors increase AI reliance, which in turn reinforces cognitive distortions, such as the Illusion of Explanatory Depth (IoED), and compulsive use. This may undermine autonomous problem-solving, critical thinking, and emotional regulation.

Proposed interventions include AI literacy programs, reflective practice, metacognitive training, and self-regulation strategies. Recognizing personal vulnerability factors—such as compulsivity, reward sensitivity, and need for closure—can help users maintain balanced AI engagement while preserving cognitive autonomy [22,26,67,68,69,70,71,72,73,74,75,76].

In summary, research on the psychological and behavioral dimensions of AI use focuses on individual vulnerability factors, habitual engagement patterns, and cognitive distortions. Scholarship has evolved from descriptive analyses of AI overuse toward empirical investigations linking personality traits, stressors, and behavioral outcomes to dependency risk. Emerging perspectives highlight the interplay between professional contexts, educational settings, and neurobiological mechanisms, emphasizing the need for interventions combining AI literacy, reflective practice, and metacognitive strategies to preserve cognitive autonomy and well-being.

Figure 3 presents a schematic summary of the emerging evidence.

6. AI Literacy, Trust, and 21st-Century Skills

The recent literature increasingly emphasizes the role of AI literacy and trust as key mediators in shaping users’ interactions with artificial intelligence systems, particularly in educational and professional contexts. Contrary to early assumptions that higher levels of digital competence would inherently protect users from problematic or excessive AI reliance, emerging evidence suggests a more nuanced and paradoxical relationship.

Analysis of the interplay between AI literacy, trust in AI systems, and 21st-century skills—such as critical thinking, problem-solving, and self-regulated learning—indicates that while AI literacy enhances users’ ability to effectively engage with AI tools, it may simultaneously increase trust to a degree that reduces critical oversight [21]. This phenomenon gives rise to a “dependency paradox,” whereby more knowledgeable and skilled users become increasingly inclined to delegate cognitive tasks to AI systems, potentially undermining independent reasoning and reflective judgment. Rather than fostering autonomy, excessive trust—when not balanced by metacognitive awareness—can erode the very competencies that AI literacy is intended to support.

Trust and skill-related factors also mediate AI dependency across different user profiles [23]. Evidence shows that trust acts as a critical intermediary between technical competence and usage intensity: users with higher perceived competence tend to develop stronger trust in AI outputs, which predicts more frequent and less critically mediated use. Skills alone do not guarantee responsible engagement; the absence of explicit strategies for monitoring, questioning, and contextualizing AI-generated outputs can amplify reliance, even among highly trained users.

Effective engagement with AI and generative AI (GenAI) among learners and professionals depends not only on technical proficiency but also on critical literacy, epistemic vigilance, and reflective practices. When AI is perceived primarily as a reliable cognitive partner rather than as an assistive tool, users may progressively offload higher-order thinking tasks. This improves efficiency but risks superficial understanding and reduced reflective engagement [77,78,79].

Educators’ competencies play a pivotal role in shaping these outcomes. Teachers’ AI-related digital skills are closely linked to the development of students’ 21st-century competencies, but these skills must extend beyond operational knowledge to include ethical awareness, critical evaluation, and pedagogical intentionality [80,82]. Embedding AI within collaborative, reflective learning environments can further mitigate dependency risks, as students engage in shared problem-solving and critical discussion rather than relying passively on AI-generated outputs [83].

At a systemic level, AI literacy and trust are essential for sustainable educational transformation. Unchecked enthusiasm for generative AI, particularly in high-performing or efficiency-focused contexts, may obscure potential cognitive downsides and foster uncritical dependence [84,85,86]. Collectively, these findings highlight that AI literacy must integrate technical skill, critical thinking, and reflective use practices to safeguard learning outcomes, professional judgment, and the development of transferable 21st-century competencies. In conclusion, scholarship in this domain emphasizes that AI literacy and trust are critical mediators of user engagement, influencing the balance between efficiency and independent reasoning. Research has progressed from equating technical competence with protective effects to recognizing the “dependency paradox,” whereby skilled users may over-trust AI outputs. Emerging work stresses the integration of critical evaluation, epistemic vigilance, and reflective practices as essential complements to technical proficiency, ensuring that AI literacy supports rather than undermines 21st-century competencies, including critical thinking, problem-solving, and self-regulated learning. Figure 4 reports a sketch of the emerging evidence in this specific domain.

7. Strategies for Responsible AI Integration

As evidence accumulates on the cognitive, behavioral, and educational risks associated with excessive AI reliance, recent studies have increasingly shifted attention from descriptive analyses of dependency toward the identification of concrete strategies for responsible AI integration. Rather than advocating restriction or prohibition, the emerging consensus emphasizes structured, guided, and reflexive use of AI systems across educational and professional settings.

Empirical evidence supports the central role of instructional design and teacher-mediated interventions in mitigating the negative effects of AI dependency [17]. Unstructured or autonomous use of generative AI tools is associated with inflated academic self-efficacy and weaker performance outcomes. In contrast, structured integration—characterized by guided feedback, scaffolding, and explicit task framing—helps preserve independent cognitive engagement. Educators function as critical regulators of AI use: when teachers actively contextualize AI outputs, encourage verification, and require reflection on AI-assisted work, students are less likely to outsource core cognitive processes. These findings suggest that responsible AI integration is primarily a pedagogical challenge, grounded in instructional practices that balance efficiency with cognitive autonomy.

Extending this perspective beyond the classroom, governance-oriented frameworks for responsible AI adoption emphasize iterative human–AI interaction combined with institutional oversight mechanisms [26]. Rather than treating AI as a static tool, continuous evaluation cycles allow AI outputs to be routinely reviewed, corrected, and contextualized by human experts. This approach prevents automation bias and over-trust, ensuring AI remains embedded within decision-making processes that retain clear human accountability. Governance structures—such as usage guidelines, transparency requirements, and role-specific boundaries for AI deployment—align AI integration with professional standards and ethical norms.

Recent research also highlights the relevance of addressing psychological and behavioral dimensions of AI use. Studies on mental health, academic stress, and technology-related behavioral patterns indicate that AI dependency can interact with pre-existing vulnerabilities, amplifying risks of cognitive offloading and reduced well-being [87,88,89,90,91]. For example, academic and family stressors can exacerbate students’ reliance on AI tools, while AI chatbots—though accessible and supportive—require careful integration to prevent unintended dependency in mental health interventions. Validated measures for monitoring AI–digital life balance provide practical tools to support responsible engagement.

These findings inform actionable strategies for responsible AI integration:

• Scaffolded use: AI should be employed purposefully, with clear expectations for when and how it supports tasks, promoting balanced engagement and preventing over-reliance.
• Human oversight: Central to all AI-assisted activities, oversight ensures that AI complements rather than replaces critical decision-making, preserving accountability and ethical responsibility.
• Reflective practices: Users should justify AI-assisted decisions and compare AI-generated outputs with human-generated alternatives, fostering critical evaluation and metacognitive awareness.
• Assignment of responsibility: Clear delineation of roles prevents diffusion of accountability, ensuring users remain engaged and ethically responsible.

These strategies are applicable across educational and professional domains. In learning environments, they support the development of durable skills such as critical thinking, self-regulation, and metacognitive awareness. In professional contexts, they help preserve judgment, ethical responsibility, and collaborative decision-making. Framing AI as an augmentative tool, these strategies emphasize deliberate design, supervision, and governance to enhance cognitive autonomy, professional competence, and well-being.

Overall, evidence from [17,26,87,88,89,90,91,92,93,94,95,96] provides a foundation for translating concerns about AI dependency into actionable strategies, highlighting structured pedagogy, iterative human–AI interaction, psychological monitoring, and governance as key levers for responsible integration.

In conclusion, studies in this area focus on actionable frameworks for embedding AI responsibly within educational and professional contexts. Mainstream research has shifted from describing AI dependency toward identifying structured, guided, and reflective interventions, emphasizing scaffolded use, human oversight, and accountability mechanisms. Recent evidence underscores the importance of integrating pedagogical design, governance structures, and psychological monitoring to ensure that AI enhances cognitive autonomy, ethical reasoning, and professional competence, rather than substituting for core skills. Figure 5 shows a sketch of the emerging evidence in this specific domain.

8. Discussion

This discussion is organized into seven interrelated sections, each addressing a distinct aspect of AI and generative AI (GenAI) use among youth, including adolescents, young adults, and professionals. The structure reflects both the multidimensional nature of AI engagement and the narrative approach adopted in this review. By dividing the discussion in this way, we aim to guide the reader from conceptual synthesis to practical implications, highlighting opportunities, risks, and policy-relevant insights.

• The Added Value of the Narrative Review

The first section justifies the narrative and exploratory approach, which enables the integration of heterogeneous evidence from diverse populations, methodologies, and constructs. Unlike systematic reviews, which prioritize comprehensiveness and replication, the narrative framework allows identification of recurring domains, patterns, and early signals of risk—such as cognitive offloading, overreliance, and addiction-like behaviors—while situating them within broader educational, psychological, and ethical contexts.

• Cognitive, Behavioral, and Educational Impacts of AI and GenAI Use

The second section synthesizes empirical findings on the opportunities and challenges of AI use, highlighting its potential to enhance learning efficiency, creativity, and professional productivity, while also fostering dependency, inflated self-efficacy, and behavioral or cognitive maladaptation. The section identifies recurring domains of impact, emphasizing the interplay between individual traits, learning environments, and AI affordances.

• Artificial Intelligence Policy and Governance

This section reviews international frameworks and ethical guidelines that shape responsible AI deployment, including UNESCO recommendations and European Commission directives. It emphasizes principles such as structured pedagogical integration, AI literacy, monitoring, and inclusive practice, and underscores the need for developmentally appropriate and context-sensitive implementation.

• Monitoring AI Engagement and Emerging Tools for Assessment

The fourth section presents psychometric instruments specifically developed to measure AI engagement, overuse, or dependency, such as CAIDS and PCGUS. It situates these tools within established methods for behavioral and digital technology assessment, drawing parallels with smartphone use scales, and highlights their role in early detection and structured interventions.

• Critical Considerations and Knowledge Gaps in AI and GenAI Use

This section highlights critical knowledge gaps and emerging challenges related to the use of AI and generative AI among youth and professionals. Evidence shows that disparities in AI engagement are shaped by layered digital divides, including unequal access to infrastructure and differences in digital and AI literacy, which can influence both opportunities for learning and risks of dependency. At the same time, AI dependency represents a phenomenon distinct from general technology overuse, with generative tools potentially fostering cognitive offloading and overreliance, particularly when AI literacy and trust are not balanced by critical reflection. Finally, while international policy frameworks provide important guidance for responsible AI use, their effectiveness depends on context-sensitive adaptation to local educational systems, cultural environments, and institutional capacities.

• Emerging Recommendations for Responsible AI Use

The sixth section synthesizes evidence-based and policy-informed strategies for promoting responsible AI engagement. Recommendations focus on structured pedagogical support, human oversight, AI literacy, ethical awareness, and cognitive-behavioral safeguards. This section connects research findings to actionable guidance for educators, policymakers, and learners, reinforcing the multidimensional nature of responsible AI use.

• Limitations of the Study

Finally, the seventh section acknowledges the limitations of this narrative review, including its non-systematic design, partial coverage of national or local policies, and the emerging nature of empirical evidence. It also emphasizes the challenges posed by the rapid evolution of AI technologies, which may affect the relevance and applicability of findings over time.

Together, these six sections provide a structured framework to understand AI and GenAI engagement across cognitive, educational, behavioral, ethical, and policy domains. This organization allows the discussion to move from conceptual synthesis to practical recommendations, highlighting both current evidence and areas for future research and policy development.

8.1. The Added Value of the Narrative Review

The findings synthesized across this narrative review highlight the value of adopting a narrative and exploratory approach to the study of AI and Generative AI (GenAI) dependency. The rapid diffusion of AI technologies has produced a fragmented literature, characterized by heterogeneous constructs (e.g., overuse, reliance, dependency, addiction-like engagement), diverse populations, and variable methodological designs. In this context, a narrative review allows integration of evidence that would be difficult to reconcile within a strictly systematic framework [97,98,99].

This approach proved particularly useful in identifying recurring domains across the included studies, allowing patterns to emerge across otherwise heterogeneous designs and populations.

First, cognitive, educational, and academic impacts were consistently reported, with evidence suggesting that AI and GenAI can enhance learning efficiency and content production, while at the same time fostering overreliance, cognitive offloading, and inflated self-efficacy that may ultimately undermine deep learning and academic achievement [16,17,21]. These effects were shown to be context-dependent and modulated by mediating factors such as self-efficacy, teacher support, and educational governance [17,19].

Second, the review highlights professional and pedagogical consequences of AI dependency, particularly among educators and healthcare professionals. Qualitative and mixed-methods studies indicate that excessive or uncritical AI use may erode pedagogical autonomy, critical reflection, and intrinsic motivation, while also raising ethical concerns related to authorship, responsibility, and professional judgment [18,20,23]. Importantly, insights from these professional contexts offer transferable implications for youth, as they reflect downstream effects of long-term reliance on AI-mediated decision-making.

Third, psychological and behavioral dimensions emerged as a distinct and increasingly relevant domain. Several studies describe AI dependency in terms that parallel behavioral addiction frameworks, linking excessive GenAI engagement to negative academic emotions, Fear of Missing Out (FoMO), needs frustration, and vulnerability traits such as impulsivity or perfectionism [16,22,25]. These findings suggest that AI dependency may extend beyond instrumental overuse, intersecting with emotional regulation and well-being.

Finally, the narrative synthesis underscores the complex role of AI literacy, trust, and 21st-century skills. While higher AI literacy and trust are often framed as protective factors, evidence indicates a paradox whereby greater familiarity with AI tools can coincide with increased dependency and reduced confidence in independent problem-solving, critical thinking, and collaboration skills [21,24]. This reinforces the view that AI dependency is not simply a function of low competence, but rather emerges from a dynamic interaction between individual characteristics, educational practices, professional norms, and the affordances of increasingly capable AI systems.

Taken together, these interconnected domains illustrate that AI dependency cannot be understood as a unidimensional or purely behavioral phenomenon. Instead, it reflects a dynamic and context-sensitive process, shaped by personal attributes, learning environments, institutional cultures, and design features of AI technologies themselves—an insight that is best captured through a narrative and exploratory synthesis. By prioritizing exploration over quantification, the review captures early signals of risk—such as cognitive offloading, illusion of explanatory depth, and addiction-like patterns—while simultaneously acknowledging the opportunities associated with AI-supported learning, creativity, and professional efficiency. This balance is particularly important given the developmental stage of the field, where definitions, thresholds, and normative benchmarks for AI use are still under debate.

Overall, the narrative and exploratory perspective adds value by situating empirical findings within broader educational, psychological, and ethical discussions, offering a coherent conceptual map that can inform guideline development, future empirical research, and evidence-informed recommendations for responsible AI engagement.

8.2. Cognitive, Behavioral, and Educational Impacts of AI and GenAI Use: Opportunities and Gaps/Challenges

Across the reviewed literature, opportunities and challenges associated with AI and GenAI use emerge as interdependent. AI offers benefits such as efficiency, personalization, and cognitive support, yet these are counterbalanced by risks of cognitive offloading, behavioral dependency, and reduced critical engagement, particularly in educational and early professional contexts.

As summarized in

Table 2. Opportunities and gaps/challenges emerging on the overview of the studies.

Domain	Opportunities	Challenges and Gaps	Key References
Cognitive, Educational, and Academic	Personalized learning; immediate feedback; support for content creation and language learning; efficiency in routine academic tasks	Cognitive offloading; inflated self-efficacy; reduced deep learning, reading comprehension, and vocabulary acquisition; FoMO-driven usage	[16,17,18,19]
Professional and Pedagogical	Enhanced productivity; support for lesson planning and clinical decision-making; access to up-to-date knowledge	Erosion of critical thinking and professional autonomy; automation bias; ethical risks; reduced peer collaboration	[18,20,21]
Psychological and Behavioral	Emotional support; reduced cognitive load; perceived companionship in some contexts	Compulsive use patterns; addiction-like behaviors; illusion of explanatory depth; emotional distress when AI is unavailable	[22,26]
AI Literacy, Trust, and 21st-Century Skills	Improved technical competence; facilitation of problem-solving and creativity; support for life-long learning	Dependency paradox: increased literacy may foster greater trust and reliance, potentially reducing critical oversight and contributing to skill atrophy	[21,23,24]
Governance and Responsible Integration	Structured AI use; hybrid human–AI decision-making; pedagogical and ethical frameworks	Lack of standardized guidelines; limited long-term evidence; uneven institutional policies	[17,26,66]

, the narrative synthesis identifies recurring domains including cognitive and academic impacts, professional and pedagogical consequences, psychological and behavioral dimensions, and the role of AI literacy and trust in shaping 21st-century skills. Personalized learning and immediate feedback can enhance engagement, but excessive reliance may foster inflated self-efficacy, Fear of Missing Out (FoMO), and superficial understanding [16,17].

In professional settings, AI can enhance productivity and support decision-making, but habitual or uncritical use may erode autonomy, critical thinking, and ethical judgment [18,20,21]. Psychological research indicates that AI engagement may resemble addiction-like patterns, reinforced by cognitive biases such as the Illusion of Explanatory Depth and reward-driven habitual use [22,26]. AI literacy and trust show a paradoxical effect: higher competence can increase reliance, potentially undermining independent reasoning and reflective judgment [21,23,24].

Critical gaps remain in governance, standardized guidelines, and longitudinal evidence. Responsible AI integration appears most effective when supported by structured pedagogy, human oversight, reflective practices, and institutional governance mechanisms [17,26,66]. Taken together, these findings illustrate how opportunities and risks co-develop and highlight areas for future research, policy, and interventions aimed at balancing innovation with cognitive autonomy and well-being.

8.3. Artificial Intelligence Policy and Governance: Global Perspectives

As the adoption of AI, particularly generative AI (GenAI), accelerates in educational and professional contexts, international initiatives and policy frameworks have emerged to guide responsible use, mitigate potential risks, and support the development of digital competencies among the youth. These policies emphasize the need to balance technological innovation with ethical, cognitive, and developmental considerations, ensuring that AI functions as a tool for learning and creativity without fostering dependency or uncritical reliance. The UNESCO Guidance for Generative AI in Education and Research highlights equitable access, ethical literacy, and critical thinking skills for the youth, advocating instructional designs that integrate AI support while promoting independent cognitive engagement [100]. Similarly, the Recommendation on the Ethics of Artificial Intelligence provides a normative framework for policymakers and educational institutions, outlining principles to protect rights, inclusivity, privacy, and the well-being of learners in AI-mediated environments [101]. At a regional level, the European Commission has issued ethical guidelines for educators on AI and data use in teaching and learning, emphasizing structured integration, transparency, and reflective practices as central to safeguarding educational quality and promoting responsible technology adoption [102].

Together, these frameworks converge on several key principles relevant to the youth (

Table 3. Key International AI Policy Frameworks Relevant to Responsible Integration.

Initiative/Framework	Objective	Target Population	Relevance to Responsible AI Use	Reference
UNESCO Guidance for Generative AI in Education and Research	Promote equitable access, ethical literacy, and critical thinking; integrate AI while fostering independent engagement	Students, educators, researchers	Supports scaffolded AI use, encourages reflection, prevents cognitive dependency	[100]
Recommendation on the Ethics of Artificial Intelligence (UNESCO)	Provide normative principles for ethical AI deployment; protect rights, inclusivity, and learner well-being	Policymakers, educational institutions, learners	Ensures ethical and inclusive AI deployment; guides policy and curricular integration	[101]
European Commission Ethical Guidelines for AI in Education	Guide educators in responsible AI and data use; emphasize transparency and structured integration	Educators, school administrators, students	Promotes structured pedagogical integration, reflective practices, and safeguards educational quality	[102]

):

• Ethical and inclusive deployment—AI systems should be accessible to diverse learners, minimizing inequities and supporting social and emotional development [100,101].
• Structured pedagogical integration—AI should augment rather than replace core learning processes, emphasizing scaffolding, feedback, and reflective exercises [100,102].
• Digital and AI literacy—Users should develop critical evaluation skills, epistemic vigilance, and an understanding of the cognitive implications of AI engagement [100,101].
• Monitoring and governance—Institutions are encouraged to establish clear policies, usage guidelines, and accountability mechanisms to prevent unmoderated reliance and potential cognitive or behavioral harms [100,101,102].

While these global recommendations are valuable, implementation remains uneven, and empirical evidence on their impact among the youth is still limited. There is a pressing need for longitudinal studies, culturally sensitive interventions, and validated metrics to assess both AI literacy and dependency, as well as their interaction with cognitive, educational, and psychological outcomes. Notably, regulatory and pedagogical frameworks align with emerging research on AI dependency, FoMO, cognitive offloading, and behavioral patterns, providing actionable guidance to mitigate risks while maximizing learning opportunities [16,17,18,19,20,21,22,23,24,25,26,100,101,102].

8.4. Monitoring AI Engagement and Emerging Tools for Assessment

As AI and generative AI (GenAI) tools become increasingly embedded in educational and social contexts, concerns about excessive or maladaptive use among the youth are growing. While most research has focused on describing patterns of engagement, recent studies emphasize the need for structured assessment instruments to identify behaviors indicative of overreliance, cognitive offloading, or emotional dependency.

Initial efforts have produced psychometric tools specifically designed for AI engagement. The Conversational AI Dependence Scale (CAIDS) [103] provides a 20-item measure capturing compulsive use, withdrawal symptoms, mood modification, and functional impairments in college students. Similarly, the AI Addiction Scale for Researchers [104] operationalizes compulsive and maladaptive AI behaviors in professional contexts, demonstrating that structured assessment is feasible even in cognitively demanding domains.

Additional instruments, such as the Problematic ChatGPT Use Scale (PCGUS) [105], focus on unidimensional measurement, assessing maladaptive engagement with generative AI and its associations with psychological distress and self-control. These tools illustrate how AI-specific behaviors can be quantified, monitored, and linked to cognitive and emotional outcomes.

These emerging instruments reflect methodological strategies long established in digital and behavioral technology research. For instance, scales measuring problematic use of digital tools consistently target core dimensions such as salience, mood modification, loss of control, and functional impairment [105]. Well-validated measures of smartphone use, including the Smartphone Addiction Scale–Short Version (SAS-SV) [106,107,108], have been pivotal in identifying risk patterns, correlates, and developmental trajectories of technology-related dependency among the youth. AI-focused assessment tools are beginning to adopt a similar approach, capturing early signs of behavioral and cognitive overreliance.

Integrating these instruments into local educational or longitudinal monitoring programs allows for early detection of maladaptive patterns, provides feedback for learners and educators, and informs interventions that aim to balance cognitive benefits with responsible AI engagement. Importantly, such approaches enable research into the developmental dynamics of AI use and the interaction between individual traits, contextual demands, and AI design features.

Recent evidence indicates a parallel with smartphone dependency: just as psychometric scales have enabled the identification and tracking of problematic mobile phone behaviors, emerging AI-specific instruments allow researchers to quantify early signs of AI dependency in the youth. Both domains highlight the importance of structured assessment, reflective practices, and interventions that strengthen self-regulation, metacognitive awareness, and adaptive engagement, ensuring that AI functions as a tool for cognitive and emotional development rather than a source of maladaptive dependency.

8.5. Critical Considerations and Knowledge Gaps in AI and GenAI Use

This section addresses key knowledge gaps and critical considerations that have emerged from the literature on AI and GenAI use among adolescents, young adults, and professionals. By highlighting these areas, we aim to guide future research, policy, and educational interventions.

8.5.1. Intersectional Disparities and Digital Divides

Understanding disparities in AI engagement among youth requires distinguishing between multiple layers of the digital divide, each with distinct implications for access, learning outcomes, and potential dependency. Two key dimensions are particularly relevant:

• Infrastructure Divide—disparities in physical access to internet connectivity and digital devices, which can limit the ability of youth to meaningfully engage with AI tools and platforms [109,110].
• Digital and AI Literacy Divide—disparities in the competencies required to navigate, interpret, and critically engage with digital content and AI outputs, shaping not only the quality of engagement but also the risk of overreliance or superficial use [109,110].

• Infrastructure Divide

The infrastructure divide captures unequal access to basic digital connectivity and devices, which remains a foundational barrier to AI use. Despite global growth in internet penetration, inequalities persist at global, national, and intra-national levels. As of recent estimates, 2.6 billion people—roughly one-third of the world’s population—remained offline, with most of the unconnected concentrated in low- and middle-income countries [111]. Internet penetration in lower-income regions often lags far behind that of high-income countries, and rural populations have substantially lower connectivity than urban counterparts, compounding educational and social exclusion. These gaps are not solely geographic: gender, age, and socioeconomic status also predict connectivity disparities, with women, low-income households, and rural communities having significantly lower access to reliable broadband [111].

A substantial body of research demonstrates that inequities in access to ICT infrastructure hamper educational participation and outcomes [112]. In online education settings, inequalities in digital readiness—shaped by socioeconomic background—strongly influence learning effectiveness, reinforcing pre-existing social stratification even when schools provide similar ICT resources.

From an AI perspective, infrastructure deficits critically constrain the ability of youth to use GenAI tools meaningfully. Without reliable internet or suitable devices, learners cannot access AI-enhanced learning platforms, adaptive tutoring systems, or real-time feedback tools—limiting both opportunity and exposure to emerging technologies. This means that structural exclusion at the infrastructure level risks deepening existing educational and cognitive divides rather than ameliorating them.

• Digital and AI Literacy Divide

Even when basic connectivity exists, digital and AI literacy—the set of skills needed to navigate, interpret, and critically evaluate digital content and AI outputs—remains unevenly distributed. The digital literacy divide goes beyond device ownership or internet access, encompassing competencies such as information evaluation, problem-solving with digital tools, understanding algorithmic limitations, and recognizing bias in AI systems. Research on digital inequality emphasizes that disparities in digital skills often correlate strongly with socioeconomic background, with students from more advantaged households displaying higher digital competencies and greater capacity to leverage technology for learning [113].

This literacy gap persists in educational contexts: students with similar access to devices and connectivity may nonetheless diverge widely in how effectively they use digital technologies. Those with stronger digital skills are more likely to engage in information-seeking and creative use of online tools, whereas others may be limited to basic or entertainment-oriented usage patterns, undermining educational benefit and critical understanding.

The emergence of AI and GenAI introduces an additional layer of complexity to digital literacies. Beyond basic ICT skills, learners must develop AI-specific literacies—including the ability to interpret model outputs, understand uncertainty, and apply AI tools responsibly in context. In the absence of structured educational programs addressing these competencies, populations with digital access may still struggle with effective or critical AI use, making them vulnerable to overreliance, superficial engagement, or uncritical trust in AI outputs.

• Intersectionality and Combined Effects

Infrastructure and literacy divide rarely operate in isolation. Individuals and communities experiencing multiple disadvantages (e.g., low-income, rural residence, limited formal education, or gendered barriers) face compounded barriers to equitable AI engagement. These intersectional disparities mean that policies focusing solely on infrastructure expansion may be insufficient unless paired with targeted literacy and pedagogical interventions.

Furthermore, the interplay between access and competencies influences how youth engage with AI tools: some groups may be excluded entirely due to lack of connectivity, whereas others may be connected yet lack the skills to use AI critically, increasing the risk of dependency without understanding.

Addressing these multilayered divides requires integrated strategies that combine investment in broadband and device access with digital and AI literacy education, tailored to the diverse needs of learners. Such approaches can both expand opportunity and reduce the risk that existing inequalities are reproduced or exacerbated in the age of AI.

8.5.2. Differentiating AI Dependency from General Technology Dependency

While parallels exist between AI dependency and general technology or digital device dependency, AI introduces unique cognitive, behavioral, and contextual dynamics. Unlike general technology use, which may involve habitual smartphone, social media, or gaming behaviors, AI dependency often centers on interactive, generative, or decision-support tools, including chatbots, tutoring systems, and AI-driven learning platforms [16,17]. These tools engage users in real-time problem-solving and creative tasks, fostering both learning opportunities and potential patterns of overreliance.

• Variability Across Tool Types and Contexts

Patterns of dependency vary notably by AI tool type and context of use. Generative AI applications, such as ChatGPT, are often perceived as essential for study and work tasks, which can drive frequent usage and feelings of being unable to manage without them. Empirical surveys indicate that over half of respondents (54.4%) reported using ChatGPT for academic purposes, with more than two-thirds of these using it at least weekly. Among active GenAI users, 27.3% reported using such tools “very often” and 36.4% “often,” indicating high engagement in learning contexts [16,28,114]. Conversely, AI tools used for professional decision support or personalized feedback may produce more strategic dependency patterns, where engagement is functional rather than compulsive [20].

• Psychometric Assessment of AI-Specific Dependency

Emerging psychometric instruments allow targeted assessment of AI-specific engagement. The Conversational AI Dependence Scale (CAIDS) and the Problematic ChatGPT Use Scale (PCGUS) capture distinct dimensions of AI dependency, including uncontrollability, withdrawal-like symptoms, mood modification, and negative functional impacts [103,105]. These tools provide more precise measurement than general technology-use scales (e.g., Smartphone Addiction Scale) [106], allowing researchers to differentiate AI dependency from broader digital overuse.

• Behavioral and Cognitive Implications

Evidence suggests that AI-specific dependency may affect both academic engagement and cognitive processing. High-frequency use of generative AI for writing or problem-solving can correlate with reduced effortful learning or diminished metacognitive reflection, particularly when students rely uncritically on AI outputs [17,19,49]. Structured educational interventions that promote AI literacy can mitigate dependency risk while enhancing skill acquisition and responsible usage [21,24].

• Future Directions

To refine understanding of AI-specific dependency, future studies should:

• Expand validation of AI-focused scales across diverse educational contexts and cultures.
• Investigate longitudinal impacts of AI engagement on learning outcomes and mental well-being.
• Explore how AI literacy, trust, and perceived usefulness interact with dependency patterns [16,21].

Overall, AI dependency represents a distinct phenomenon from general technology overuse, shaped by tool affordances, context of use, and individual cognitive and behavioral factors. Differentiating these forms of dependency is crucial for designing interventions that maximize AI’s educational benefits while minimizing risks of overreliance.

8.5.3. The AI Literacy–Dependency Paradox

Emerging evidence points to a paradoxical relationship between AI literacy, trust in AI, and user dependency: while greater competence with AI tools often facilitates adoption and effective use, it can also coincide with increased reliance on these tools in ways that may undermine independent problem-solving and critical engagement. AI literacy encompasses not only technical understanding of generative AI systems but also the abilities to evaluate outputs, recognize limitations, and strategically integrate AI support into complex learning tasks [114]. Such competencies foster informed adoption but do not inherently protect against habitual dependence.

Quantitative studies highlight this complexity. Measures of AI literacy and perceived trust have been associated with reliance on generative AI tools, and higher dependency profiles correlate with reductions in autonomous engagement and self-regulated problem-solving when not accompanied by metacognitive supports [21]. Increased competence with generative AI may correlate with reduced effort in independently initiating and verifying tasks unless explicit instructional is placed on evaluative reasoning, autonomous cognition, and structured adoption frameworks [115,116,117,118,119]. Responsible educational integration requires more than technical proficiency with AI systems. Effective strategies include:

• Metacognitive practices that encourage reflection on the why and how of AI outputs.
• Critical evaluation skills to distinguish reliable from spurious content.
• Pedagogical designs that use AI to support independent reasoning rather than bypass it [114].

While AI literacy and trust are essential for effective engagement with generative technologies, they may also increase the risk of dependency if not anchored in reflection, verification practices, and curricula that foreground autonomy and cognitive agency.

8.5.4. Adapting Global Policies to Local Contexts

International policy frameworks provide normative guidance for responsible AI integration, but their implementation requires adaptation to local educational, cultural, linguistic, and institutional contexts. Global guidelines emphasize universal ethical principles—transparency, fairness, accountability, and learner protection—but practical translation depends on local infrastructures, pedagogical traditions, regulatory systems, and socio-cultural attitudes toward technology.

Top-down digital policy frameworks often encounter challenges across heterogeneous educational systems [120,121]. Countries with centralized education systems may adopt uniform AI governance policies more rapidly, whereas decentralized systems require adaptation at regional or institutional levels. Cultural and linguistic factors further influence AI integration, as generative models are predominantly trained on datasets reflecting dominant languages, potentially limiting accessibility for underrepresented communities [100,122]. Institutional capacity, including teacher training, governance structures, and digital infrastructure, strongly affects AI adoption and the mitigation of cognitive offloading and uncritical reliance [121,123].

Effective implementation of international AI governance frameworks requires multi-level policy translation:

• Curricular adaptation to integrate AI literacy and critical evaluation skills into subject-specific learning activities.
• Teacher professional development to translate policy principles into practice.
• Institutional governance mechanisms, such as responsible AI guidelines, assessment integrity policies, and monitoring systems.
• Cultural and linguistic localization to reflect the needs of diverse learner populations.

Bridging the gap between global principles and local realities requires context-sensitive strategies, participatory policymaking, and ongoing empirical evaluation, ensuring sustainable, equitable, and pedagogically meaningful AI integration.

8.6. Emerging Recommendations for Responsible AI Use

As evidence grows on the cognitive, behavioral, and emotional risks associated with excessive AI use, both empirical studies and international policy frameworks highlight practical strategies to promote responsible engagement among the youths. These approaches aim to maximize the educational and cognitive benefits of AI while minimizing potential dependency, cognitive offloading, or emotional maladaptation.

Key principles emerging from the literature include:

• Structured Pedagogical Integration—AI should support, not replace, core learning processes. Scaffolding, guided feedback, and reflective exercises help preserve independent problem-solving skills and critical thinking ([16,17,21]).
• Human Oversight and Monitoring—Students’ AI engagement should be systematically observed. Psychometric instruments such as the Conversational AI Dependence Scale (CAIDS) and the Problematic ChatGPT Use Scale (PCGUS) allow early identification of maladaptive patterns, including compulsive engagement, mood modification, and cognitive offloading ([103,104,105]).
• Development of AI Literacy and Epistemic Vigilance—Learners should acquire critical evaluation skills, recognize AI’s cognitive limitations, and engage in reflective practices to balance trust with autonomy ([21,23,24,100,101]).
• Ethical Awareness and Inclusive Practices—Instructional design and AI deployment should uphold ethical principles, ensure accessibility, and foster social-emotional well-being ([100,101,102]).
• Integration with Behavioral and Cognitive Safeguards—Supporting metacognitive strategies, self-regulation, and emotional resilience can help prevent overreliance or compulsive use ([103,104,105]).

Together, these recommendations (Figure 6) underscore that responsible AI use is inherently multidimensional, requiring the integration of cognitive, behavioral, educational, and ethical considerations. Emerging research also draws a parallel with smartphone use in adolescents, where structured monitoring, literacy interventions, and reflective practices have proven effective in mitigating addictive behaviors. This analogy reinforces the importance of early, localized, and developmentally appropriate strategies to ensure that AI remains a tool for growth rather than a source of cognitive or emotional disruption ([106,107,108]).

8.7. Limitations of the Study

This narrative review has several inherent limitations. First, as a narrative rather than a systematic review, the selection of studies and policy documents was not exhaustive, which may introduce selection bias and limit the replicability of the findings. Second, while international policy frameworks such as UNESCO guidelines and EU recommendations were considered, the review did not comprehensively cover national or local AI policies, which are often context-specific and may offer additional insights into regulatory approaches and educational interventions. Third, empirical evidence on AI dependency, especially among adolescents and young adults, remains limited and emerging, with many studies relying on small or region-specific samples.

Finally, the rapid evolution of generative AI technologies means that both usage patterns and policy responses are continuously changing, potentially affecting the relevance and applicability of findings over time. Despite these limitations, the review provides a timely overview of current evidence, assessment tools, and emerging recommendations, highlighting directions for future research and policy development.

9. Conclusions

This narrative about young adults presents both significant opportunities and notable risks within educational contexts. Evidence suggests that AI can enhance learning efficiency, provide personalized feedback, and support content creation, fostering engagement and cognitive growth. At the same time, overreliance, cognitive offloading, and compulsive engagement may undermine deep learning, critical thinking, and academic autonomy.

The review emphasizes that AI dependency is a multidimensional phenomenon shaped by individual characteristics, pedagogical practices, institutional frameworks, and the design of AI technologies. Responsible AI integration in education requires structured scaffolding, reflective exercises, oversight mechanisms, and promotion of AI literacy and epistemic vigilance. International policy frameworks, such as UNESCO guidelines and European Commission recommendations, provide essential guidance but must be adapted locally to ensure inclusivity, developmental appropriateness, and cognitive well-being.

By synthesizing current evidence and emerging recommendations, this review offers a coherent conceptual framework for guiding responsible AI use in educational settings, supporting both cognitive growth and digital competence while mitigating potential harms.

10. Future Directions

Despite emerging evidence, the study of AI and generative AI use in educational contexts remains in its early stages. Several priorities can guide future research, policy, and practice to ensure responsible integration while maximizing cognitive and developmental benefits.

Longitudinal and Multisite Studies. There is a need for larger, longitudinal studies across diverse cultural and educational contexts to examine how AI engagement and dependency evolve over time. Such designs can clarify developmental trajectories and identify protective or risk factors ([16,17,21]).

Integration of Policy and Educational Evaluation. While international frameworks provide essential guidance—such as UNESCO’s Guidance for Generative AI in Education and Research ([100]), the Recommendation on the Ethics of Artificial Intelligence ([101]), and the European Commission’s Ethical Guidelines on AI in Teaching and Learning ([102])—future work should evaluate the impact of national and local AI policies on learning outcomes, digital literacy, ethical awareness, and dependency patterns.

Standardization of Assessment Tools. Reliable psychometric instruments are crucial for monitoring maladaptive engagement, cognitive offloading, and emotional consequences. The Conversational AI Dependence Scale [103] and the Problematic ChatGPT Use Scale [105] provide validated frameworks, but further testing across populations and educational settings is needed.

Intervention Design and Effectiveness. Structured pedagogical approaches, reflective exercises, digital literacy training, and self-regulation strategies can be combined to promote responsible AI use. Early evidence from digital media and smartphone studies provides models for such interventions [106,107,108].

Ethical, Social, and Cognitive Impacts. Future research should explore how AI literacy, trust, and autonomy interact with cognitive development, emotional regulation, collaboration, and social outcomes, particularly among adolescents and young adults [18,19,21,27,33].

Adaptive and Inclusive AI Design. Collaboration between educators, policymakers, and AI developers is essential to create tools that are equitable, accessible, and supportive of independent reasoning while minimizing dependency ([51,53,85]).

Taken together, these directions outline an evidence-informed roadmap to balance innovation with cognitive, emotional, and ethical well-being in educational contexts, providing actionable guidance as AI technologies evolve rapidly.

Figure 7 synthetizes a sketch of the future directions.