The study interviewed fifteen PhD candidates enrolled in public universities across Morocco, drawing participants from English studies, economics and finance, and philosophy. All had prior experience with AI tools and were interviewed through semi-structured sessions conducted over WhatsApp and Zoom, lasting between fifteen and twenty minutes. The researchers used purposive sampling to select participants who could articulate their experiences in English, reducing translation bias, and then applied manual thematic analysis following the framework established by Braun and Clarke in 2006. Three dominant themes emerged from the coding process, and together they sketch a portrait of technological adoption that is neither utopian nor catastrophic, but deeply ambivalent.

The first theme the researchers identified is what they call AI as an academic co-pilot, a framing that captures how students now treat chatbots and research assistants as collaborators rather than tools. Participants reported using systems including ChatGPT, Gemini, DeepSeek, and RDiscovery for lesson planning, brainstorming, simplifying dense research concepts, and generating practice worksheets. One participant, who simultaneously works as a PhD student and an English as a Foreign Language instructor, described using AI to elicit warm-up activities and to break down complex topics before discussing them with a faculty supervisor. The psychological effect is notable: students reported feeling less pressure when exploring preliminary ideas with a chatbot, precisely because the machine cannot judge them the way a human mentor might. This creates a low-stakes environment for intellectual experimentation that many participants found genuinely valuable.

But the same efficiency that makes AI appealing also displaces effort in ways students themselves recognize as problematic. The researchers documented what they term time and effort displacement, where repetitive technical tasks are delegated to algorithms, freeing cognitive resources for critical analysis and creative work. One participant described this as allowing greater focus on higher-order thinking, while another called AI a lifesaver for identifying key researchers in a field and summarizing content. Moroccan doctoral students face particular structural pressures here, as limited institutional access to global literary resources makes autonomous learning more difficult than it might be elsewhere. AI tools effectively fill that infrastructure gap, streamlining resource acquisition in ways that feel indispensable. Yet the researchers note that this convenience comes with hidden cognitive costs, as the literature increasingly links heavy AI reliance to poor memory retention and procrastination.

The second theme is more psychologically intricate and, arguably, more troubling. The researchers call it the paradox of AI-mediated confidence and self-reliance, and it describes a double-edged dynamic where external validation from algorithms simultaneously boosts and undermines students’ belief in their own abilities. Many participants reported feeling insecure about their grammar and syntax, particularly as English becomes the dominant language of research publication. AI proofreading tools offer a solution that raises their assurance in finished drafts. One participant explained that AI helps avoid grammatical mistakes and improves writing, which somehow raises confidence. Another admitted not feeling confident about grammar because mistakes often go undetected without AI assistance. The pattern is consistent: students increasingly depend on algorithmic quality control to feel comfortable submitting work at all.

The researchers found that this external reinforcement works as a cognitive dependency loop. Participants described seeking validation from AI tools whenever sharing work for feedback, and not feeling fully confident submitting anything without consulting the software first. One participant reflected that AI makes them feel slightly less independent, especially when used for quick solutions. This continuous feedback cycle undermines the sense of academic self-efficacy that postgraduate training is designed to cultivate. The researchers connect this to broader findings in the literature, including work by Alhur and colleagues in 2025 documenting a paradox in higher education where technological validation inadvertently leads to cognitive dependency and a decline in perceived learner competence. Khan and Suhluli, also writing in 2025, similarly explain that while digital assistants offer immediate task scaffolding, directionless deployment carries hidden cognitive costs including task fatigue and intensified technology-related anxieties.

The third theme addresses ethical boundaries and the threat to originality. Participants demonstrated strong self-awareness about the risks of cognitive delegation, recognizing that the ease of handing tasks to AI can compromise the authenticity of their work. One participant noted that overuse can affect originality, but that they personally work to keep their own voice and perspective central. Another admitted it can be tempting to delegate an entire task to AI and copy the final result without much thought to the process. The researchers found that ethical boundaries become increasingly difficult to police as students grow accustomed to AI covering small tasks, gradually wanting more. The line between valid scaffolding and unreflective delegation of core intellectual work becomes harder to draw over time.

Participants also split on whether AI ultimately inhibits or enhances learning. Some reported feeling that AI tools reduce their learning and academic achievement, while others argued the opposite, claiming that critical use supports and enhances intellectual growth. The researchers found that some participants experienced guilt when using AI in academic settings, viewing it as a threat to independent learning, while others saw it as a legitimate resource when used responsibly. Perhaps most striking is what the researchers call the inevitability of dependency. One participant self-assessed as being at minimum sixty to seventy percent dependent on AI tools. Another reflected that having tasted what AI can offer, it is really difficult to go back to how research used to be. The convenience of immediate assistance and direct access to information has reshaped students’ expectations of the research process itself, making fully independent practices feel almost archaic.

The researchers argue that these findings carry significant implications for Moroccan higher education, which is currently transitioning toward digitized learning. They note that the pressures of institutional expectations and research publication demands trigger deep AI dependence among postgraduate students, and that technology adoption is not merely a technical adjustment but a psychological and cultural transition. The absence of intentional boundaries, they warn, risks transforming AI tools from educational assistants into cognitive substitutes that alter how future scholars perceive their academic roles. The study recommends specialized professional development workshops for faculty, comprehensive institutional frameworks defining permissible boundaries for AI use in research, and heavy training programs to improve digital literacy among teaching staff. The researchers also call for open communication between advisors and students about AI as a research aid, with training to maintain risk awareness, distinguish algorithmic bias, and identify fabricated data.

The study acknowledges limitations, including a relatively small sample size that restricts generalization across the broader postgraduate population, and data situated entirely within a single-country context. Future research, the authors suggest, should apply mixed-methods or longitudinal designs across multiple countries to track the evolution of AI dependency over time. What emerges from this research is a picture of technological adoption that defies simple narratives of progress or decline. Moroccan doctoral students are not passive victims of AI, nor are they fully in control of its influence on their minds. They are navigating a genuinely new cognitive landscape where the boundaries between assistance and dependency, between scaffolding and substitution, are still being drawn in real time. The researchers urge policymakers to cooperate with higher education institutions to establish guidelines, training, and resources specialized in digital skill development and ethical AI use, framing this as an investment in both faculty and students that will determine whether AI ultimately amplifies or erodes the intellectual autonomy of the next generation of scholars.

Subject of Research: Education

Article Title: The impact of AI on students’ achievement, confidence, and critical thinking in education in Morocco

Article References: Mellouk, F., & Ouelfatmi, M. (2026). The impact of AI on students’ achievement, confidence, and critical thinking in education in Morocco. Discover Education, 5(1), Article 1191. https://doi.org/10.1007/s44217-026-02332-w

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02332-w

Keywords: artificial intelligence, higher education, Morocco, critical thinking, student confidence, academic integrity, ChatGPT, qualitative research, doctoral students, AI dependency, educational technology, self-efficacy