In recent years, remarkable progress has occurred in the development of large language models (LLMs), driving artificial intelligence (AI) into increasingly sophisticated realms of human-computer interaction. Leading examples include OpenAI’s ChatGPT and Google’s Gemini AI, both of which demonstrate impressive capabilities in natural language understanding, text generation, translation, and other language-driven applications (Rane, Choudhary, & Rane, 2024).

As AI technologies continue to advance and data usage expands, it becomes increasingly crucial to comprehend their broader societal implications, particularly within the fields of education and training. Educators and school leaders must possess at least a foundational understanding of AI and data practices to engage with these technologies positively, critically, and ethically, thereby leveraging their full potential (European Commission, 2022, p. 4). When applied appropriately, AI has the potential to enhance teaching, learning, and assessment, improve educational outcomes, and increase institutional efficiency.

However, the misuse or poor design of AI systems may lead to negative or even harmful consequences. It is therefore essential for educators to evaluate whether the AI tools they employ are reliable, fair, safe, and trustworthy, and ensure that educational data is managed securely, respects privacy, and serves the public good. The concept of “Ethical AI” underscores the importance of developing and deploying AI systems that adhere to ethical standards, guiding principles, and fundamental human values (European Commission, 2022, p. 11).

Gender biases in artificial intelligence models concern the academic community, especially in the field of education. The integration of AI into the educational process is a directive of the educational policies of the European Commission. Thus, more research is needed on making those AI models unbiased and ethical as far as gender issues are concerned, so that the transition can be smooth, taking the available resources in education into account.

The purpose of this research is the comparative study of the results provided by two artificial intelligence models compared with the results obtained by male and female students who were trained in gender issues in education, at a higher educational institution in the country.

The method for the texts written by the students is content analysis and was used to identify gender patterns in texts. For the texts produced by the AI ​​models a script in Python was written to do context and statistical analysis with the help of python dictionaries.

Through a feminist approach, this paper seeks to understand whether Artificial Intelligence perpetuates existing stereotypes or not, when compared with the texts produced by students. The findings showed that Gemini was the AI model that gave better results than the ChatGPT and the students, in terms of the breadth of the gender dimension in the texts they were trained to create. The students who attended the course on gender issues showed a small understanding of their perceptions about gender equality. The only notable observation is that the students included some LGBTQI issues in their answers.

Two fundamental issues arise from the foregoing research that warrant further exploration:
a) The introduction of AI in education presents a critical issue concerning how efficient systems can be trained to avoid reproducing stereotypical gender representations and
b) The education and continuous professional development of teachers who will be expected to integrate AI into gender-related teaching practices is vital to ensure that gender biases are not perpetuated through modern educational tools and resources.

The research was conducted within a higher education department where the researcher taught the course “Education and Gender” during the 2025–26 academic year. Thirty male and female students attended lectures on theories and concepts related to gender relations and participated in group activities and experiential workshops designed to foster a deeper understanding of gender stereotypes. At the end of the semester, students were provided with a set of materials to which they were asked to respond. The same exercises were then administered to ChatGPT and Gemini with the aim of comparing (a) human responses with those generated by AI systems, and (b) the responses of the two AI systems with each other. While similar studies have been conducted abroad, this research constitutes the first study within higher education to compare the responses of students, who received formal instruction on gender issues and subsequently engaged in creative writing and sentence completion tasks, with those generated by AI systems trained on the same prompts and asked to complete the same exercises. A feminist content analysis of the students’ stories and sentence completions revealed a weak correlation between the material taught during the academic semester and their final responses. This indicates that, despite exposure to feminist theory and gender studies, their answers remained markedly stereotypical. From the AI perspective of the research, the same set of prompts given to students was entered into ChatGPT and Gemini, supplemented by the instructions: “Make the people in the stories for A and B more human. Give the character a name, a profession, or any distinctive trait” and “Respond as a person aged 18–30”, all written in Greek. Without these additional prompts, the models tended to respond in a definitional or abstract manner, lacking the narrative richness necessary for meaningful interpretation. Accordingly, two separate chat sessions, one per AI model, were trained to respond as if they were “different students.” Subsequently, a Python script was developed to perform contextual and statistical analysis using Python dictionaries, associating stereotypical terms with social groups and calculating their frequency of occurrence within each category.

The research was conducted within a university department of secondary education teacher training. The findings indicate that professional development on gender-related issues is essential for educators. The study compares the dataset (a) between human and AI-generated outputs, and (b) between ChatGPT and Gemini models. In the narratives created by male and female students about a “professional,” most responses reflected traditional and gender-stereotypical occupations. When asked to write stories about a person’s ambitions, students predominantly depicted men as individuals who strive, aim, and aspire to succeed, often as freelancers or scientists. In contrast, female characters were portrayed as “dreaming” of success, with goals oriented toward helping others or protecting the environment. In the sentence-completion exercises, both male and female students provided rather conventional responses. For instance, female students were described as “trying hard, studying, not missing classes,” whereas male students were characterized as “missing classes, failing exams, playing videogames to the detriment of their studies.” Descriptions of mothers included “taking care, cooking, organizing, working hard, offering moral support, showing love,” while fathers were depicted as “working all day, providing material goods, loving and supporting silently.” The representations of grandparents were also heavily stereotyped: grandmothers were said to “grumble, stay at home, cook for their grandchildren, and help their daughters with housework,” while grandfathers “go to cafés, sit and watch TV and give money to their grandchildren.” Regarding the AI-generated responses, ChatGPT tended to produce a range of topics containing stereotypical elements, whereas Gemini offered more diverse and non-stereotypical content characterized by a positive tone across all sentence themes. Continuous training on gender awareness should be provided throughout teachers’ professional careers, as deeply internalized gender stereotypes, manifested in both thought and behavior, are often difficult to recognize and even harder to change.

Armutat, S., Wattenberg, M. & Mauritz, N. (2024). Artificial Intelligence: Gender-Specific Differences in Perception, Understanding, and Training Interest. Proceedings of the 7th International Conference on Gender Research, (pp. 36-43). Academic Conferences International Ltd. https://doi.org/10.34190/icgr.7.1.2163 Consuegra-Ayala, J. P., Martínez-Murillo, I., Lloret, E., Moreda, P. & Palomar, M. (2024). A multifaceted approach to detect gender biases in Natural Language Generation. Knowledge-Based Systems, 303, 112367 https://doi.org/10.1016/j.knosys.2024.11236 Spillner, L. (2024). Unexpected Gender Stereotypes in AI-generated Stories: Hairdressers Are Female, but so Are Doctors. In: R. Campos, A. Jorge, A. Jatowt, S. Bhatia, M. Litvak (eds.): Proceedings of the Text2Story’24 Workshop, Glasgow (Scotland), 24-March-2024. https://ceur-ws.org/Vol-3671/paper10.pdf Thomson, E., Chapman, R., Cooper G. & Cooper, Z. (2026). Exploring bias in artificial intelligence: stereotypes and gendered narratives in digital imagery of early childhood educators, Gender and Education, 38, (1), 53-72. DOI: 10.1080/09540253.2025.2546048 Valerio-Ureña, G., Sevilla-Campoverde, G., Ortúzar, S. & Lazcano, C. (2024). Gender Stereotypes in the Creation of Educational Cases with ChatGPT. International Symposium on Multimedia (ISM) IEEE. DOI: 10.1109/ISM63611.2024.00059 Voutyrakou, D. A., & Skordoulis, C. (2025). Using AI Tools to Enhance Educational Robotics to Bridge the Gender Gap in STEM. Education Sciences, 15(6), 711. https://doi.org/10.3390/educsci15060711 Yim, Iris Heung Yue (2024). Artificial intelligence literacy in primary education: An arts-based approach to overcoming age and gender barriers. Computers and Education: Artificial Intelligence, 7, 100319. https://doi.org/10.1016/j.caeai.2024.100321