Generative AI in higher education assignments: Evolution or evaluation?

In autumn 2024, I explored the early impact of generative AI on student assignments in higher education. (eSignalsPRO Generative AI in higher education assignments: Revolution or risk?). This sparked discussion on the balance between innovation and academic integrity. Now, after a full academic year of experimentation, reflection, and refinement, it is time to revisit the topic. What has changed? What have students learned? And how should educators respond?

From curiosity to critical thinking

In the autumn 2024, students approached AI tools like ChatGPT with curiosity and optimism. Many used them to generate initial drafts, brainstorm ideas, or simulate academic tone. However, issues quickly emerged: hallucinated sources, vague references, and a lack of academic rigor. As one student noted: It would have been easier to write from scratch.

By spring 2025, the tone has shifted. Students have become more discerning, using AI as a starting point rather than a shortcut. They experiment with a broader range of tools—Elicit AI, Microsoft Copilot, Perplexity, and ChatGPT+ — and have begun to compare their outputs critically.

This shift reflects a broader trend in higher education: the growing emphasis on AI literacy and critical thinking as essential academic competencies (Hazari 2024; Walter 2024). While AI tools can support argument construction and information analysis, they may also reduce students’ motivation for self-reflection if used uncritically (Zawacki-Richter et al. 2019).

Comparing the tools: accuracy vs. accessibility

Students found that Microsoft Copilot provided the most academically credible sources, often linking to peer-reviewed journals and university repositories.  Elicit AI was praised for its relevance and depth, though many of its sources were behind paywalls or inaccessible. ChatGPT, while helpful for ideation, continued to struggle with citation accuracy.

Elicit AI generated accurate references, but the permission to the provided documents was denied. We had to use Google Scholar to find accessible versions. – Group reflection

This evolution reflects a growing awareness of AI’s limitations. Students no longer accept AI outputs at face value. Instead, they verify, cross-reference, and supplement with their own research—skills that are essential in academic and professional contexts (Luckin et al. 2016; Holmes, Bialik & Fadel 2019).

Interestingly, AI tools also helped students identify emerging themes in hospitality and tourism research. Concepts like guest empowerment, co-creation, and emotional engagement surfaced repeatedly. These insights guided students toward more original and forward-thinking analyses.

This aligns with recent literature emphasizing the shift from transactional service models to experience orchestration in hospitality (Pine & Gilmore 2019; Sigala et al. 2012). Moreover, the ability of AI to surface underexplored topics demonstrates its potential as a creative partner in academic inquiry (Walter 2024).

AI as a partner, not a proxy

The spring reflections reveal a more mature, nuanced understanding of AI’s role in academic work. Students now see AI not as a replacement for research, but as a partner in the process—useful for brainstorming, outlining, and identifying gaps, but not for final citations or conclusions.

This shift mirrors broader trends in higher education, where AI literacy is becoming as important as digital literacy. As educators, we must support this transition by teaching students how to critically evaluate AI outputs, verify sources, and integrate human judgment into their workflows (Holmes, Bialik & Fadel 2016; Hazari 2024).

Ethics, equity, and the future of AI in education

As AI tools become more embedded in academic life, ethical concerns are also rising. Students and educators alike are grappling with questions of academic integrity, bias, and accessibility. A recent international study emphasised the need for culturally responsive policies and ethical guidelines to ensure that AI enhances rather than undermines educational equity (Zawacki-Richter et al. 2019).

This is especially important in multicultural learning environments like Haaga-Helia, where students bring diverse expectations and experiences to their academic work. Responsible AI use must be taught not only as a technical skill but as a civic and ethical responsibility (Walter 2024).

From risk to responsibility

The journey from autumn to spring shows clear progress. Students have moved from passive users to active evaluators of AI. They have learned that while AI can accelerate the research process, it cannot replace the rigor, ethics, and creativity that define academic work.

As we continue to integrate AI into higher education, the objectivwe should not be to eliminate risk—but to cultivate responsible, reflective, and research-savvy learners.

Picture: Haaga-Helia