Co-creating custom GPTs: an autoethnographic study of undergraduate students as partners in generative AI innovation

Synthesis: Lo et al. (2026) report an autoethnographic study at Monash University in which undergraduate science students co-created and evaluated custom GPT chatbots (cGPTs) within a students-as-partners (SaP) framework. Across five subjects (physiology, physical chemistry, organic chemistry, scientific inquiry, polymer science), student teams in a Summer Scholarship program and a placement unit partnered with academics to design, test, and refine unit-specific chatbots aligned with pedagogical goals. Inductive analysis of summer scholarship autoethnographic logs revealed the iterative, unpredictable, and creative nature of working with Generative AI. Abductive analysis of placement reflections, guided by SaP principles, surfaced themes of Learner Agency, reciprocity, reflexivity, identity shift, authenticity, adaptability, and a technology-pedagogy nexus. The authors argue that co-creation with GenAI can extend the SaP paradigm to include GenAI tools, advocating student voice as central to responsible AI innovation and offering a model for inclusive, critical, and transformative engagement with emerging technology in Higher Education.

Partnership as a Lens on GenAI

Much early discourse around Generative AI in Higher Education positioned students either as potential cheaters or as passive victims of the technology, often omitting student Learner Agency entirely. In parallel, students-as-partners (SaP) pedagogy has repositioned students as co-creators of knowledge, curriculum, and educational innovation. This study sits at the intersection of these trends, asking what happens when undergraduates and academics partner not merely to use AI but to build it. The authors adopt a dual theoretical lens — SaP for the collaborative process and autoethnography for capturing the lived experience and meaning-making of the students involved — to examine both the mechanics of partnership and its impact on participants.

Co-Creating Unit-Specific Chatbots

Students partnered with academics to build custom GPT chatbots specialized for the content and learning outcomes of specific units, working through an iterative cycle of design, tailoring, testing, and optimization. This arrangement extends SaP beyond curriculum co-creation into the co-development of AI tools, a domain where student co-creation without formal computer science training remains nascent. The authors draw on participatory design traditions and on legitimate peripheral participation to frame how students' learner expertise was combined with academics' disciplinary and pedagogical knowledge to produce trustworthy, context-sensitive tools.

Themes From Student Reflections

• Learner Agency and ownership — students reported genuine influence over the design and pedagogical direction of the tools, consistent with core SaP values of respect, reciprocity, and shared responsibility.
• Reciprocity and reflexivity — autoethnographic reflection surfaced metacognitive awareness, a hallmark benefit of partnership approaches.
• Identity shift — participants moved from consumers of educational technology to co-producers and shapers of it, an experience that demystified AI and built critical understanding of AI.
• Authenticity and adaptability — students balanced the creative, unpredictable nature of working with GenAI against the need for reliable, unit-appropriate outputs.
• Technology-pedagogy nexus — co-creation foregrounded how tool design decisions are inseparable from pedagogical intent, echoing the broader turn toward using GenAI as a catalyst for innovation rather than a threat to be managed.

Extending Students-as-Partners Into the AI Era

The central argument is that co-creation with GenAI can extend the SaP paradigm to include GenAI tools themselves, moving students from passive use toward genuine partnership in shaping educational technology. The authors reflect on the ethical and pedagogical implications of GenAI and advocate for student voice as central to responsible innovation. By embedding partnership pedagogy into GenAI co-creation, the study offers a model for more inclusive, critical, and transformative engagement with emerging technology — one that positions student co-design as both a learning activity and a AI Governance mechanism in the responsible integration of AI across university settings.

What this means for practice

• Educators. Partner students in building unit-specific AI tools rather than only teaching about them, since this project moved undergraduates from consumers of educational technology to co-producers who understood the technology critically.
• Educators. Provide scaffolding for Generative AI co-creation — critique triggers, clear strategies for iterative testing, checkpoints for reflection, and explicit discussion of coping with failure or unexpected outcomes.
• Designers. Keep technical and pedagogical training inseparable: have student partners frame prompt and design choices around learning purposes, as the Socratic cGPTs did by responding to queries with questions rather than answers.
• Designers. Budget for the iterative, unpredictable nature of the work, including repeated cycles of design, tailoring, testing, and optimization before a tool becomes unit-appropriate.
• Educators. Recruit with equity in mind — offer student-partnership roles through more than one route (scholarship and for-credit placement) so that co-creation does not reproduce the "usual suspects" of already confident and well-connected students.

Limitations

• The dataset comprised five undergraduate students working with five academics, a self-selected, high-performing cohort likely predisposed to reflective and independent learning, so the themes of agency, confidence, identity shift, and adaptability may partly reflect who chose to participate.
• Students served as both participants and reflectors, which the authors state may have shaped the reflections toward more constructive or affirming narratives and introduced a degree of positive bias.
• The work took place in English-speaking STEM units at a single institution (Monash University), where institutional culture, disciplinary norms, and language fluency may all have influenced engagement with GenAI and partnership practices.
• The authors position the study as a situated proof-of-concept for students-as-partners GenAI co-creation rather than evidence that such partnerships operate identically across broader, less self-selecting student populations.

Citation

Lo, S., Lea, T., Mavromoustakos, A., Joseph, B., Yii, M., Exintaris, B., Karunaratne, N., Pearson, D. L., Ritchie, T., Yuriev, E., & Pyun, J. (2026). Co-creating custom GPTs: An autoethnographic study of undergraduate students as partners in generative AI innovation. Frontiers in Education, 11, 1818781.