Characterizing User Needs for GenAI Incorporation in Educational Games

Abstract

This work explores user needs for educational games and gamification that incorporates Generative Artificial Intelligence (GenAI). As GenAI is increasingly incorporated in educational settings, we must consider both the wide-spanning literature on gamification and games that have been shown to benefit learning, and characterize the needs and desires of relevant stakeholders in developing educational games that incorporate GenAI generally, and specifically for higher education. A mixed-methods questionnaire inquired 345 undergraduate students about their perceptions, use patterns, needs, and desires related to GenAI, educational and non-educational games, and text-based games. GenAI tools are widely used for educational purposes already, but mostly as a supplementary source. Despite the wide use, participants expressed being concerned with accuracy, transparency, and quality. Participants also expressed a desire for an educational game/tool to have scaffolded interactions and to help with learning material in math, science, and language arts. Taken together the findings provide a road map and specific recommendations for developing an educational game incorporating GenAI. The roadmap includes instructional design (i.e., the gamified tools’ content and type(s) of instruction and interaction) through information regarding preferred platforms, game genres, gamified properties (e.g., characters, challenges), and lastly, clear information about concerns students have related to trust and equity that will need to be addressed in an educational game incorporating GenAI.

1. Introduction

The field of education has recently experienced a technological shift with the rapid advancement of Generative AI (GenAI) (Alasadi & Baiz, 2023; Baidoo-Anu & Ansah, 2023; Uanachain & Aouad, 2025). GenAI tools offer adaptative, generative and personalized capabilities that can be operationalized at scale, introducing new opportunities for transforming teaching and learning. One educational domain where GenAI shows promise is game-based learning and gamification. GenAI offers properties like dynamic environments (Kiili et al., 2018), adaptive problem solving (Annetta, 2010), collaboration and teamwork (Squire, 2013). These properties align with long-standing evidence that educational games can improve students’ knowledge retention, foster problem-solving skills, and support the transfer of learning (Clark et al., 2016; Wouters et al., 2013). A compelling application of GenAI in educational games lies in its ability to support dynamic, student-driven narratives. For example, rather than requiring developers to manually construct exhaustive decision trees for narrative progression, GenAI allows for dynamic interaction and can generate context-sensitive storylines in response to student inputs. This shift not only reduces design complexity and development time but also allows for greater narrative flexibility.

In this context, text-based games offer an accessible entry point for incorporating GenAI into educational game design. Commonly used GenAI tools rely on text-based interactions and can be leveraged for educational games and gamification (e.g., Gemini, Google, 2025; ChatGPT, OpenAI, 2023). Thus, educators, researchers, and educational game developers can build on established principles from educational games and gamification to design effective, GenAI-supported tools for learning. However, the combination of GenAI and educational games is relatively novel, and its success partly depends on how well it aligns with users’ experiences, expectations, and needs. Given the growing use of GenAI in learning contexts, it is essential to center learners’ voices in the design of such tools (Shehata et al., 2024).

The current study conducts a user needs analysis using a survey to generate insights for educators and game designers developing GenAI-supported educational games. The following sections review related work on educational games and gamification, text-based games, the potential of GenAI incorporation in these domains, and user needs analysis practices. We then present insights from a survey of undergraduate students (n = 345) and discuss the implications of their responses for the design of learner-centered educational games that integrate GenAI.

2. Related Literature

2.1. Generative AI and Education

The rapid advancements in GenAI, driven by large language models (LLMs), have initiated a paradigm shift in education (Alasadi & Baiz, 2023; Baidoo-Anu & Ansah, 2023; Uanachain & Aouad, 2025). Institutions and schools are widely integrating GenAI tools, creating AI literacy modules, and issuing guidelines to support responsible and effective classroom use of GenAI (Ng et al., 2025). As such, a growing body of research has reported that GenAI-supported tools are increasingly implemented in both teaching and learning processes. For example, educators are leveraging GenAI to automate lesson planning (Moundridou et al., 2024) and generate assessment questions (Kaldaras et al., 2024). On the learner side, GenAI enables personalized learning experiences (Udeh, 2025), interactive virtual tutoring (Belawati & Prasetyo, 2025), and real-time formative feedback (Uanachain & Aouad, 2025). Such integration of GenAI into educational settings holds the potential to reshape traditional models of instruction and learning. By offering adaptive and personalized support, GenAI can meet diverse learner needs, enhance student engagement, and enable self-directed learning that goes beyond one-size-fits-all approaches (Ayeni et al., 2024). However, achieving these potential benefits requires careful consideration of issues related to user needs, equity, accessibility, and ethical usage to ensure that these technologies serve all learners fairly and responsibly (Nguyen, 2025). In this study, we focus on the practical application in educational games, a domain where GenAI’s adaptive and generative capabilities can be operationalized at scale.

2.2. Educational Games, Gamification, and GenAI

Educational games are games that are explicitly designed for the purpose of learning (Gee, 2003; Plass et al., 2015). These games integrate content knowledge acquisition, skill-building, and pedagogical frameworks into interactive gameplay experiences aimed at both education and entertainment, a segment of a wider category called “serious games” (Michael & Chen, 2005). Educational games typically blend narrative, simulation, and systems-thinking mechanics, offering situated learning contexts in which learners apply knowledge to solve meaningful challenges (Gee, 2003). Research in this domain demonstrates that educational games can enhance knowledge retention, transfer of learning, and problem-solving skills across a wide range of cognitive domains (Clark et al., 2016; Wouters et al., 2013).

Science-based games enable learners to test hypotheses in simulated environments, while mathematics games can foster conceptual understanding through adaptive problem solving (Annetta, 2010). Importantly, these games facilitate active learning by requiring learners to engage with content dynamically and experientially, rather than passively receiving information in written or auditory form (Kiili et al., 2018). Beyond cognitive outcomes, studies highlight the affective and social dimensions of educational games. Multiplayer and collaborative environments foster teamwork, communication, and interpersonal skills (Squire, 2013), while story-driven educational games can nurture empathy and ethical reasoning—skills aligned with 21st-century learning goals (Schrader & McCreery, 2008). Yet, challenges persist with educational game adoption, including concerns around scalability, costs, curriculum integration, and varied learner acceptance (de Freitas, 2006; H. N. Eukel et al., 2017). These challenges can be alleviated through the inclusion of user (i.e., student) input being incorporated in the design process. Involving learners contributes to educational game design that aligns with user needs and preferences, making games more engaging, accessible, and as a result, more widely adopted.

Recent research has increasingly emphasized adaptive educational games, which can personalize learning through data-driven adjustments to difficulty, feedback, and narrative progression (Plass et al., 2015). The incorporation of GenAI extends these capabilities by enabling real-time personalization grounded in learner’s performance, preferences, and behavioral patterns (Borah et al., 2024; Moon et al., 2025; Oliveira et al., 2023). As a result, educational games can adapt interactions to support students’ strengths, preferences, and knowledge gaps. Personalization can be achieved through adaptive game world building, narrative branching, and interactive dialog systems (French et al., 2023; J. S. Park et al., 2023; Ratican et al., 2023; Xu et al., 2023). It can also include adjusting question difficulty and providing dynamic, more human-like feedback to support the learning process (Anjum et al., 2024). Moreover, GenAI enables feedback loops and performance-based responses without developers hand-authoring exhaustive decision trees (Koivisto & Hamari, 2019). GenAI can also automate goal setting and progress visualization, which have been shown to add to learners’ sense of achievement and learning agency (Seaborn & Fels, 2015), without programmers having to develop dedicated mechanisms to meet those goals.

Gamification is another way to make learning immersive and engaging that has been widely researched and lends itself to GenAI applications. Gamification, the use of game mechanics such as points, badges, and quests in non-game contexts, has been shown to increase motivation (intrinsic and extrinsic, Toda et al., 2019) and engagement in learning (Deterding et al., 2011), and to foster autonomy, persistence, and deeper interaction with learning materials (Hamari et al., 2014). For the purposes of this study, we consider gamification not only as a motivational tool but also as a structural framework to support interactive, learner-centered engagement in text-based games. That said, GenAI-powered gamification comes with important caveats. Critics warn against superficial implementations of gamification, noting that over-reliance on extrinsic rewards may undermine long-term motivation (Hanus & Fox, 2015). To maximize educational impact, gamification should be principled in its alignment with pedagogical intentions, providing meaningful learning experiences rather than merely incentivizing participation (Domínguez et al., 2013; J. J. Lee & Hammer, 2011).

Considerable research has investigated games and gamified environments and indicates that both are useful in education. The exponential growth of GenAI is rapidly reducing barriers to entry into the development of new educational tools that are more engaging for learners. Thus, we frame GenAI integration as a continuation of the broader shift toward adaptive, personalized educational games and gamification.

2.3. Text-Based Games for GenAI Gamification

Text-based games offer a low barrier to incorporating GenAI into educational games due to their lower graphic design and programming requirements. These games use natural language as the primary interface for world modeling and interaction, with minimal or no visual elements. Prior work has developed and studied text-based games and authoring platforms both in educational settings and beyond (Côté et al., 2018; Hausknecht et al., 2020; Madotto et al., 2020; Shridhar et al., 2020). Historically, text-based games involved game designers crafting a finite set of narratives, responses, eventualities, and actions to which players could type in their response. If a player typed in words or phrases that were not programmed into the game, they would get an error or be asked to type a valid input. The advancement of GenAI allows for free-form inputs from users and generates relevant responses to such input (Karpouzis et al., 2024). Moreover, prompt engineering for GenAI tools allows educators and designers with minimal coding and visual design experience to easily build conversational and gamified learning experiences.

The ‘quick and easy’ affordance of GenAI makes it appealing for integrating gamification into text-based learning experiences. However, further characterization of user needs and testing of educational games and gamification incorporating GenAI is required for these newly available GenAI affordances (Oliveira et al., 2023). It is crucial to understand the needs, preferences, and stances of the intended users to ensure the game design is both human-centered and inclusive such that the system can leverage the strengths of diverse users. The following section describes the user needs analysis methodology and its utility in this context.

2.4. Understanding User Needs

Designing high quality, adoptable tools or products requires consideration of the relevant stakeholders’ needs and desires (e.g., Bardzell, 2010; de Freitas & Jarvis, 2006; Kashfi et al., 2017) alongside the stakeholders’ trust requirements (Goldshtein et al., 2022). In the case of educational games, stakeholders can include the students/users, instructors if the game is employed in a school setting, and the developers of the game (Bunt et al., 2024; Goldshtein et al., 2025). User needs analyses can lead to insights pertaining to user behaviors (Owen & Baker, 2020) like the type of game mechanics students prefer; cognitive preferences such as feedback mechanisms (S. S. Lee & Moore, 2024); and educational games’ technological requirements, such as platforms.

Designing an educational game should involve users in various stages of the design to ensure it meets to students’ needs (see Maxim & Arnedo-Moreno, 2025), beginning with conceptualization (Oliveira et al., 2023; E. J. Park, 2021; Ravyse et al., 2017; Suresh Babu & Dhakshina Moorthy, 2024), continuing after developing a prototype (H. Eukel & Morrell, 2021; Eun et al., 2022), and lastly, in assessing the efficacy and user experience of a completed game. Iterative work on the product can then proceed in this user-experience cycle, which can also be characterized as a learning engineering cycle when applied to educational contexts (Baker et al., 2022; Goodell & Kolodner, 2022; McNamara, 2024).

Identifying users’ needs and experiences increases use, improves usability, and builds trust (Chiou & Lee, 2023). For example, previous research on user needs in the domain of educational games has shown that student engagement and participation increase when games incorporate students’ preferred forms and modalities of feedback (Hyland & Hyland, 2019; Zhang, 2020).

Finally, because needs and abilities vary across populations, inclusive design is essential in games, especially for educational games that address specific user needs such as learning disabilities (Hassan, 2024; Lukava et al., 2022; Spiel & Gerling, 2021). These considerations are even more salient with GenAI, which enables highly personalizable experiences that can better adapt to individual learners (Khan et al., 2024; Moon et al., 2025).

The current study focuses on characterizing the needs and preferences of students, as they are a consistent stakeholder, whether an educational game is employed in a school setting or outside it. In exploring more specific contexts (e.g., the classroom), researchers must also consider other relevant stakeholders’ needs in their analysis (e.g., instructors, Jiang & Ziden, 2025; parents and therapists, Stefanidi et al., 2023).

2.5. Current Study

This study reports on a user needs analysis survey conducted with 345 participants to examine students’ preferences and stances related to games, educational games, text-based educational games, GenAI, and the combination of these. Our research questions are as follows:

• RQ1: Based on their experiences, what are students’ perceptions on GenAI in everyday life and educational contexts?
• RQ2: Based on their experiences, what are students’ perceptions of games and educational games?
• RQ3: What are students’ perceptions regarding the incorporation of GenAI in educational games?

Study findings will inform our own work on developing educational gamified tools incorporating GenAI but will also serve as guidelines for other game developers and educational researchers for incorporating GenAI in educational settings, specifically in gamified educational tools.

3. Materials and Methods

3.1. Participants

A total of 345 students consented to participate in the study. Participants’ ages ranged from 18 to over 26, with an average age of 19.35 years (SD = 1.73). One entry was removed from the age analysis due to missing age data. In terms of academic standing, most participants were freshmen (62.32%), followed by sophomores (22.03%), juniors (11.01%), seniors (4.06%), and graduate students (0.58%). The sample was predominantly female (57.10%), with 41.74% male and 1.16% preferring not to disclose their gender. Participants represented a variety of academic disciplines. The most common majors were Business (45.80%), STEM (34.20%), and Social Sciences (6.38%). A smaller proportion reported majors in Arts (3.48%), Humanities (1.16%), or other (8.99%) fields. Within the ‘other’ category, notable concentrations included Nursing, Graphic Design, and Criminal Justice.

3.2. Study Design

This study employed an exploratory mixed-methods survey design to explore user needs and preferences related to the use of GenAI and games in educational and non-educational contexts. The survey included both quantitative (e.g., Likert scale) and qualitative (e.g., open ended) items. The survey was designed to include items spanning several fields: education, educational games, gaming, and use of Generative AI. Prior to the beginning of the survey, all participants were presented with an informed consent statement outlining the purpose of the study, the voluntary nature of participation, and data confidentiality. Surveys were administered online via Qualtrics over a period of four weeks. Approval for the study was granted by the researchers’ Institutional Review Board (IRB). Participants were recruited through the psychology subject pool and received course credit for participation. The full questionnaire is provided in Appendix A.

3.3. Survey Structure

The survey began with demographic information, e.g., age, gender, educational level, and major. The rest of the survey contained 36 questions (a mix of open-ended and multiple choice questions, see Appendix A) and was structured into four sections addressing key themes relevant to GenAI and educational game design:

• Usage of GenAI in personal and educational contexts: Participants were asked about their usage patterns of GenAI tools such as ChatGPT, DALL-E, Perplexity, and Gemini. Participants were asked to report which tools they use, as well as the frequency and context (e.g., personal vs. educational) of their GenAI usage. Participants were further asked how such tools were employed in educational contexts, if applicable. Based on their responses, participants were prompted to elaborate on the tools used, reasons for non-use (if participants reported not using GenAI tools), and learning scenarios in which GenAI played a role (e.g., concept clarification, assignment help, writing support, etc.). In addition to reporting their usage of GenAI tools, participants also rated the tools’ perceived helpfulness in educational contexts. Additionally, respondents indicated the subjects for which they found GenAI most useful (e.g., science, math, language arts) and selected the importance of GenAI in their learning process (e.g., primary resource, supplementary tool, last resort). To explore some limitations of GenAI use in different contexts, a checklist item assessed common challenges associated with GenAI use for learning, such as difficulty understanding responses, misinformation, ethical concern, and lack of guidance. The responses to these questions provided essential insights to inform learner-centered design in educational game development.
• Familiarity with games, educational games, and text-based games: To understand participants’ gaming behaviors and preferences, this section focused on games in both recreational and educational contexts and included a set of questions focused on frequency, modality, and content of gameplay. Participants reported how often they play games, with options ranging from “every day” to “never”. Those who indicated any level of game play were asked to specify their preferred game modalities (e.g., console, computer, phone, etc.) and genres (e.g., strategy, RPGs, etc.), as well as to list examples of games they enjoy. An additional item assessed participants’ preferred game features (e.g., storytelling, problem-solving, multiplayer interaction, etc.). Furthermore, to assess familiarity with educational games, participants were asked whether they had played such games, and if so, to describe them. Follow-up questions probed the location and timing of playing educational games (e.g., at school, home, etc.), what topics the game addressed (e.g., science, math, coding), and the devices used to access them (e.g., PC, mobile, VR). Finally, participants were asked to indicate which STEM-related subjects they would prefer to learn in a future educational game. The survey also included questions aimed at capturing students’ prior experiences and preferences. Responses to these questions offer insights into how gameplay habits in recreational and educational contexts can inform the design of an engaging, learner-centered AI-supported narrative game.
• Perceptions of games within GenAI context: This section of the survey explored participants’ perceptions of the potential future applications of GenAI in learning and gaming contexts. Participants were asked whether they believed GenAI tools could provide new ways to support their learning beyond their current use of those tools, with an open-ended follow-up prompting ideas for additional use cases. They were also asked whether they believed GenAI should be used in games, and if so, to suggest types of games that might benefit from AI integration. Lastly, participants indicated their willingness to engage with GenAI features within gameplay, offering insight into their openness to AI-enhanced gaming experiences.
• Design Recommendations: The final section of the survey focused on eliciting participants’ design preferences for a GenAI-supported educational game. Open-ended questions asked participants to describe the features that would make a GenAI-supported game appealing to them and how they envisioned GenAI enhancing their overall gaming experience. Lastly, participants were encouraged to offer additional suggestions for features they would like to see in a STEM-focused educational game using GenAI.

3.4. Data Processing and Analysis

Survey responses were cleaned and later analyzed using descriptive statistics and thematic coding to derive insights to guide the design of an AI-driven narrative educational game. For multiple choice questions, we calculated descriptive statistics (i.e., means, standard deviations, distributions) to summarize the trends, while open-ended questions were examined through thematic analysis.

3.4.1. Data Cleaning

The raw dataset contained 401 survey responses. A multi-step data cleaning process was employed to ensure data quality and validity. Entries without participant consent and entries with less than 100% completion rate were excluded from the dataset. Duplicate responses were identified and removed by keeping only the first submission associated with each unique participant to capture their first interaction with the questionnaire. Following cleaning, we assessed response validity based on survey completion time. The average completion time was 22.37 min (SD = 93.15). An outlier threshold was defined as two standard deviations above the mean. Implementing the data cleaning steps resulted in a sample consisting of 345 responses.

3.4.2. Thematic Analysis

A total of three raters rated all open-ended questions, with two raters collaboratively coding each question. Seven open-ended questions were coded using a conventional content analysis procedure (e.g., Hsieh & Shannon, 2005). Specifically, for each question, two raters read through all responses to each question and inductively generated a coding schema based on the topics identified in the response. This was followed by comparing the two coding schemas until an agreed upon coding schema was derived. The two raters then separately coded each question according to the agreed coding schema until an agreement rate of 75% was reached. At that point disagreements were resolved through discussion until 100% agreement was reached. When raters were unable to reach an agreement rate of 75%, coding schemas were changed to better reflect the data or adjusted to be more well-defined for both raters, before re-coding the data.

4. Results

This section reports students’ experiences with and attitudes toward GenAI across personal and educational contexts. Following the research questions, we summarize patterns of adoption, common use-cases, perceived usefulness, and concerns that shape how students engage with GenAI tools.

4.1. RQ1—Based on Their Experiences, What Are Students’ Perceptions on GenAI in Everyday Life and Educational Contexts?

Responses to the first section of the survey questionnaire (Usage of GenAI in personal and educational contexts) address this research question. Participants described their GenAI usage patterns in general, and for educational purposes. They also shared their stances on the helpfulness of GenAI tools in general and within educational contexts.

• Usage of GenAI in Personal and Educational Contexts

Participants’ usage of GenAI is described in terms of perceived value of GenAI tools, challenges and reasons for not using GenAI tools, and perceived future potential uses for such tools.

• Overall Use and Perceived Value of GenAI Tools

Participants reported widespread use of GenAI tools in both daily life and educational contexts. Over 70% indicated using GenAI tools several times a week or more (Figure 1), with ChatGPT being the most frequently mentioned, followed by tools such as Gemini, Copilot, Grammarly, and Snapchat. Educational usage was particularly prominent as well: over 95% of participants reported using GenAI for both school-related purposes (e.g., assignments) and for informal learning (e.g., self-directed study or information search), with more than 65% indicating use for learning occurred several times a week.

Figure 1. Reported frequency of GenAI use for personal and educational purposes.

When describing their daily GenAI use, participants most commonly cited support for learning academic concepts (42%), helping with assignments (35%), writing and grammar assistance (23%), and summarizing texts (5%) (Table 1). Other frequently mentioned uses included ideation (26%) and information seeking as a replacement or enhancement to traditional search engines (24%), while only 1.5% reported not using GenAI at all. It is worth noting that when asked about personal use of GenAI, student responses indicate that use for educational purposes (e.g., help with writing tasks, homework) overlaps with non-education related use (e.g., ideation, search for information).

Table 1. Participants’ reported general use of GenAI. Responses are summarized in themes and percentage counts.

Category	Description	%	Examples from Data
Learn, understand, explain academic concepts	School-related learning	42.3%	”Help with school or to simply answer questions I have”, “no, only study”
Help with homework/ assignments	School-related homework/assignment support	34.5%	“For homework question help and studying for exams”, “Studying and helping with homework”
Help with brainstorming, creative ideation, and organization	Personal or education-related tasks that have to do with ideation and organization	26.1%	“I usually use AI tools to keep me organized and track what I need to do in a particular week.”, “Most of the time I use it is to brainstorm topics for anything”, “Mostly getting ideas for essays or organizing my notes.”
Help summarize texts	Describes summarizing texts	5.2%	“I use them to summerize [sic] text”, “I mostly use ChatGPT to have summaries of articles that would take me too long to read…”
Enhance/replace search engine	General searches and question asking	23.8%	“i ask it questions i cant find the answers for”, “If google doesnt come up with an answer to a problem I ask chat GPT”, “I use these tools to find out more research if I do not know something. AI is now so commonly used for many platforms and websites that I find myself interacting with AI more than I intend to.”
Help with writing/grammar	Assistance with writing related tasks and assignments	23.2%	“To help me with finding articles to write essay”, “help in writing”, “Only for educational purposes like asking for grammar fixing or if I answered the essay prompt correctly”
Personal use	Interaction related with personal life and personal tasks	24.6%	“ChatGPT. I ask ChatGPT about its opinions on things happening in my life.”, “I will sometimes ask ChatGPT to rate shows as what they think I would like best.”, “I ask AL to help me and give me advice on relationships. I ask AL for good gifts for certain people and holidays. “
Professional tasks	Work, job applications, and professional settings not related to school	4.6%	“Work”, “I use them for generating text for resume’s […]”, “[…] writing resumes, preparing for interviews.”
Don’t use GenAI	Mentions not using GenAI	1.5%	“I don’t us generative ai in my daily life”, “
Other/Invalid/ Unclear	No response, irrelevant, or not codable	7.54%	“CHATGPT”, “apple”

In a multiple-response question specifically about educational use cases (Table 2), the most common uses included searching for and understanding information, brainstorming project ideas, and writing/editing text, with less frequent but notable mentions of language learning.

Table 2. Participants’ reported educational uses of GenAI tools. The table below shows the percentage of selecting each options (multiple selections allowed).

Category	%
Understanding new concepts or topics	85.7%
Searching for information	71.7%
Brainstorming ideas for learning	67.8%
Assisting with homework/assignments	55.9%
Practicing or reviewing material	52%
Researching academic topics	49.2%
Writing/editing text	42.86%
Language learning/translation assistance	15.5%
Other	3%
None. I do not use AI for learning	1.8%

The perceived value of GenAI for learning was overwhelmingly positive: 97% of participants found these tools helpful, and over 80% rated them as very or extremely helpful. Mathematics, language arts, and science emerged as the most common subject areas in which students applied GenAI for academic support.

• Challenges and reasons for not using

The most commonly reported challenge to using GenAI tools for learning was inaccurate or misleading information (78%). About a third reported ethical concerns (36%) and difficulty interpreting AI responses (35%). Among participants who reported not using GenAI tools for learning, the most common reasons were lack of interest/need or perceived usefulness, concerns about AI unreliability, and a preference for other resources or one’s own work. A small share (0.29%) cited environmental harm.

• Perceived future potential

Many participants reported that GenAI features can be leveraged to aid their learning in ways that extend beyond their current use (70%), some participants (24.6%) were unsure, and 4.4% did not feel that GenAI can be useful for learning. When asked to describe what future uses GenAI may have in education, most participants brought up uses like understanding concepts (31%) and generating practice questions and quizzes (14%). Some participants mentioned GenAI’s ability to assist with their personal learning work such as homework and assignments (2.7%) and help with their writing (7%). More specifically to GenAI, participants mentioned that GenAI can contribute to a more personalized (7%) and engaging (4%) learning experience. A small number of participants (4%) brought up additional responses like generally incorporating GenAI within schools (e.g., ‘I think it would be helpful if ai became more acceptable to use for school, i feel like it is an extremely useful tool and many professors already support using it’ and ‘One engineered for school’), and 10% of participants did not know or did not respond to the question.

4.2. RQ2—Based on Their Experiences, What Are Students’ Perceptions of Games and Educational Games?

The two following questionnaire sections in the survey serve to address our second research question. In these sections, participants reported their experience with games, educational games, and text-based games and shared their perceptions about incorporating GenAI into those types of games. We describe the responses to each subsection separately below.

4.2.1. Familiarity with Games, Educational Games and Text-Based Games

Participant’s familiarity with games, educational games, and text-based games was assessed through their exposure, context of use, and preferred devices. The participants also stated needs and priorities for both general and educational gameplay.

Games (General)

Participants exhibited diverse gaming habits, ranging from daily play (31.9%) to rare or no play (22.9%). While 44% reported playing games at least weekly, a substantial portion engaged occasionally, providing a broad range of prior gaming experience among respondents. The most popular modalities were mobile devices, gaming consoles, and computers (Figure 2). The most popular game genres reported were action/adventure and puzzle. Participants reported rewards and progression as a feature they most enjoy in games, but also mentioned problem-solving, social/multiplayer interaction, and customization as desirable features.

Figure 2. Proportion of game modalities played by participants. Response options refer to technology-related modalities (cell phone, consoles, personal computers), and in-person modalities (party games, table top games, live action role playing games). In addition, participants could refer to modalities not mentioned by choosing ‘other’ (note. Participants could choose multiple options).

Educational Games

We also asked participants about their experience with educational games. Nearly two-thirds of participants (64.35%) reported having played educational games.

• Contexts

When asked about the stages of education at which they played educational games, participants most commonly reported playing educational games in high school (57.6%), followed by middle school (53.6%) and elementary school (50.7%); fewer reported playing them in college/university (36.6%). With respect to the location, participants reported playing educational games mostly at school (50.7%), followed by at home or at a friend’s house (34.8%). The commonly cited educational game topics were math problem solving (27.8%) and language arts (26.7%), followed by problem-solving skills (20.9%); additional topics included science concept understanding (15.7%), history/social studies (14.5%), creativity (14.2%), and coding/computer skills (9.9%) (these percentages reflect a multi-select item).

• Devices and user needs/priorities

Unlike the responses for games (Figure 2), participants typically used PCs/computers (56.5%) to play educational games. Some mentioned using mobile phones (29.7%) and tablets (8.0%), and very few cited consoles (1.5%) or VR headsets (1.1%) (Figure 3). When asked which STEM subject would be most useful for a future educational game, participants prioritized science (53.9%), followed by mathematics (24.1%) and technology (14.5%); relatively few selected literature learning (8.1%), health science (5.2%), or environmental science (2.6%).

Figure 3. Proportions of modalities of educational games played. Response options refer to technology-related modalities (mobile phone, consoles, personal computers, tablets, VR headsets). In addition, participants could refer to modalities not mentioned after choosing ‘other’. (note. Participants could choose multiple options).

Text-Based Games

Participants were also asked about their exposure, preferred features, and challenges associated with text-based games.

• Exposure

When asked specifically about their experience playing text-based games, over 60% of the participants reported either never playing text-based games or being unsure, while 38% reported some level of familiarity (n = 131). Among participants who reported familiarity with text-based games, most named titles that align closely with the genre (e.g., A Dark Room, 20 questions), while 22% were unable to name any specific text-based games. Some participants listed games that are not strictly text-based but include text elements within primarily visual gameplay, such as Legend of Zelda: Breath of the Wild, Pokémon, and Mario Party. These responses suggest that the term “text-based game” may be poorly defined or misunderstood by those less familiar with the genre. The remaining questions about text-based games were answered by this subgroup of 131 participants.

• Preferred Existing Features

Participants were asked about their favorite features of text-based games. Engaging storytelling (74%) and flexibility in choices and outcomes (65%) were voted as most liked features in text-based games. These features were followed by imaginative gameplay (53%), character development (42%), and problem solving (34%). A small percentage of participants (9%) mentioned they do not enjoy text-based games.

• Desired Features

Participants were further asked about features they would like to see in new text-based games being developed. Responses (Table 3) mentioned the desire to have solid and engaging storylines (26.2%), visuals (3.2%), characters (6.2%), and rewards (1.7%). Participants also mentioned wanting the game to be in their preferred genre (15.2%) and referred to specific preferred types of tasks and challenges (10.7%) and game durations (9.7%). Additionally, participants mentioned these elements generally (e.g., “interactive story”) but also envisioned ways in which those elements might be improved using GenAI (e.g., “I would like to see one where you make up your own scenario and generative AI makes up a long adventure based off that prompt.”). This GenAI-specific improvement specifically included mention of personalization and interactive affordances (19.2%).

Table 3. Categories and examples of responses to the open-ended question ‘What features would you like to see in a text-based game that is being developed (e.g., type of story, duration, type of task, etc.)?’. The percentage for each response is out of a sample of 345. Some responses received multiple codes, when relevant.

Category	Description	%	Examples from Data
Storyline	Responses related with storyline	26.2%	”An interesting story”, A good story…”, “A good story is the backbone of a text-based game”, “Plot twists”, “Stories that are easy to follow and engaging”
Genre/ Writing Style	Specific genre, games	15.2%	“Action/adventure”, “horror”, “roleplay”, “More mystery based”, “ …fantasy story”, “drama stories or mystery…”, “world building stories”, “…mystery or romance storyline…”, “I enjoy D&D so I think anything with a good story like that…”, “Related to this generation maybe the gym to anything with nature”, “tasks like GTA text tasks”
Character	Responses related with characters	6.2%	“Female protagonist that isn’t just boobs”, “NPCs”, “Realistic story and characters”, “Character development/progression”, “…good characters”, “…relatable characters”, “Character customization”
Rewards	Receiving rewards	1.7%	“Collection of awards”, “Reward system”, “incentives”, “short side quests to help earn badges…”, “Tasks with rewards”
Problem Solving/ Challenges	Tasks/challenges/problems	10.7%	“Puzzles, riddles, humor”, “Varying tasks to make it more interesting”, “…unique and non repetitive types of tasks”, “harder problem solving…”, “…engaging puzzles”, “problem solving”, “Fun tasks”, …different tasks the player can do…”
Personalization/Interactive	Multiple choices, adaptive narratives/ characters, multiple narratives/endings	19.2%	“A story that I can change the results of based on my choices”, “…games that envelop the user in an interactive world”, “More branching outcomes”, “A story that I can manipulate based on my answer choices…”, “Interactive story”, “A choice in your own storyline”, “… a game that changes as you play”
Visualization	Visual elements	3.2%	“Good graphics”, “make video”, “good plot and graphics”, “…visuals (could just be 2d cardboard cutout style depending on the game, not just a black screen or something like that)”, “…graphically pleasing”
Length	Duration of the game, length of text	9.7%	“Short duration”, “Long duration of the game…”, “short, straight-forward”, “Long duration and immersive”, “A long engaging stories where you have main and side tasks you can do along the way”, “Quick, does not take to [sic] long, incenctitives”, “short text”, “not too much reading…”
NA/Unclear/ Invalid/ Don’t want	No response, vague, not interested in text-based games	34.4%	“Not sure”, “Not very interested in text-based games”, “idk”, “I wouldn’t play that type of game”, “I don’t have any ideas”
Other	Responses that did not fit in with other codes	7%	“Less paying”, “Accessibility and variety”, “it has co op elements”, “Creativity”, “Realistic responses”, “some sort of mental stimulation”, “what the game is”, “no need to pay money for the good options”,

• Challenges and uncertainty

When asked about challenges regarding text-based games, a large group of responses (34.4%) included unclear or blank responses, lack of suggestions, or explicit disinterest in text-based games. The ‘other’ category captured unique suggestions from individual participants, such as making the text-based game free to use, focusing more on accessibility, or adding cooperative/multiplayer modes. See Table 3 for examples of all response categories.

4.3. RQ3—What Are Student Thoughts, Desires, and Perceptions Regarding the Incorporation of GenAI in Educational Games?

The last section of the questionnaire was used to answer RQ3. In this section participants could express their thoughts, desires and concerns related to the idea of incorporating GenAI in educational games. This section more explicitly requested participants to think of GenAI features that would be appealing to them, enhance their gaming experience, and add on any additional comments, suggestions, or ideas they may have not been asked about.

4.3.1. Perceptions of Games Within GenAI Context

When asked whether participants think that GenAI can be used in gaming, 41.3% of participants responded ‘Yes’, 51.2% responded ‘Maybe’, and the remaining 7.5% of participants responded ‘No’. While most participants reported either a desire to use GenAI tools in a game (30%) or potentially considering that option (52%), 18% of participants reported not wanting to do that.

4.3.2. Design Recommendations

Participants were asked four open-ended questions probing for their design recommendations. When asked to describe features they would like in a GenAI-supported game (Table 4), many participants brought up improvements in two domains: First, gamified features such as characters (17%), multiplayer (5%), enhanced storytelling (17%), specific game genres (25%), and the desire to form games similar to other games they like (3%). Second, participants mentioned features pertaining to personalized user experiences that they believe GenAI could enhance, such as general personalization (12%), learning assistant (7.5%), and help in solving problems (3%).

Table 4. Categories and examples of responses to the open-ended question ‘Please describe the features that would make a GenAI-supported game appealing to you. (e.g., game genre, type of story, type of task, type of input from players, or type of response)?’. The percentage for each response is out of a sample of 345.

Category	Description	%	Examples from Data
Game genre	Mention of specific type of game genre	25%	”Action game genre”, “Action/Adventure”, “Mystery”, “An open-world RPG or simulation game”
Storytelling features	Conversational, creative, engaging, immersive. Students mentioning features related to story/narrative	17%	“The AI would be able to modify storylines to the players liking”, “It might be able to provide more relatable stories”
Characters	Mention of character features (e.g., ‘female protagonist’)	7%	“Characters with similar personalities and interests. So players can fully use their imagination”, “Realistic characters students can relate with”, “female protagonist”
Game design features	Aesthetics, colorful, levels, music/sound, rewards	10%	“I have no idea what GenAI would be useful for in games except for making the graphics and mechanics better”, “Easy to use, interesting, extremely detailed but aesthetically pleasing”, “Immersive gameplay”
Multi-player	Ability to play with multiple Players within the game	5%	“Multiple players”, “Being able to play with other people”
Game references	Mentioning specific game titles	3%	“Recently I was playing Disco Elysium which is a text based game that challenges ideals and perceptions, it was deeply engaging but uncomfortable.”, “Similar to among us”, “Open world story game somewhat like Legend of Zelda Breath of the Wild”
Personalization	Mentioning personalization or personalized experiences	12%	“Generated text specific to the player”, “If ai was to be used in a game, it would be for open responses to situations the viewer does mainly in dialogue that could be very customizable for games.”
Encourage problem solving	Mention of problem solving or help with problem solving	3%	“problem solving”, “Something challenging to test the mind of players”, “Very interactive and uses problem-solving”
Helps with learning	Mentioning learning help, or specific subjects	8%	“Sometimes for hard levels help to find a solution”, “A math game that could educate and teach would help me”
Choice driven features	Wanting options to choose from	7%	“Multiple choice questions with interactive feature”
Invalid/Unclear/No response	No response, or response that is not clear	23%	“IDK”, “Not sure”, “game genre”
Other	Responses that are clear but do not fit within the other categories	14%	“It wouldn’t”, “I feel like using AI to create a game takes away the creativity of video game makers.”

When asked about ways in which GenAI could enhance their gaming experience (Table 5), participants said they believe GenAI will generally improve games (e.g., ‘…add something extra…’). Participants also mentioned improved visuals (24%) and improvement in the narrative, story, and genre elements (21%), and potentially making the game more fun and engaging (4%). Some participants mentioned enhancing the personalization experience and support in learning (2%). A relatively small percentage of participants (7%) mentioned they did not want GenAI involved in games, fearing that it would have a negative impact on the game quality (e.g., narrative, visuals). While some did not provide specific reasoning, others focused on aspects where GenAI was good for enhancing stories but not visuals, and others the other way around (e.g., ‘I don’t feel like AI could create compelling stories’, or ‘AI shouldn’t touch storytelling, visual art, or programming in any way’).

Table 5. Categories and examples of responses to the open-ended question ‘How would GenAI enhance your gaming experience? (e.g., AI affecting story, AI affecting visual elements or type of input from players.)’. The percentage for each response is out of a sample of 345.

Category	Description	%	Examples from Data
General improvement/ positive	Positive stance with no elaboration	14%	”It is a simple way to add something extra to a users experience “, “I would be less confused”
Enhanced story/ narrative/genre	Improvement in narrative/story	21%	“AI affecting story”, “they could make the stories shorter so it is less reading when you want to play a game”, “good stories that are always different”
More characters/ NPCs	Enhancement related with characters	6%	“making the game experience feel more involved and the NPCs less like NPCs when they respond in more unique ways”, “NPC characters utilizing AI to respond to my vocal output”, “Provide better NPCs”
Improved visuals/ graphics	Enhancement in visuals	24%	“The visuals”, “AI Visual elements”, “It might make the visuals actually very pleasing”
Personalized learning	Suggested strategies, hints, learning support, learning personalization	2%	“It could accelerate the learning aspect”, “Informing me on why I was right or wrong based on my choices, not just telling me I;m wrong. Correcting my choices with explanations basically.”
Game personalization	More options, interactivity, responsivity, more realistic interaction, game support	5%	“AI response to player input”, “Create a more unique environment based on our choices”
Invalid/unclear/ don’t know	No response, or response that is not clear	13%	“Better scores”, Not sure”
Don’t want GenAI involved in game	Negative impact, lack of desire to have GenAI as part of games	7%	“It would make it more engaging in a way, but also lackluster if the story is just made from AI and doesn’t have the creative intent or input from developers.”, “It wouldn’t for me personally”, “Idk if it would”
Other	Responses that are clear but do not fit within the other categories	2%	“pirate themed”, “Uniqueness”
Fun	Fun, engaging, entertaining	4%	“It could have the ability to change someones mood in a positive way”, “I think it would help get me entertained”

Lastly, participants were asked ‘Do you have any additional suggestions or comments about features you would like to see in a STEM educational game that uses GenAI (e.g., abilities, type of game, type of character, type of device it is designed for)’? to provide them with an opportunity to bring up any thoughts they were not previously asked about. A total of 37% of the participants responded to this question. Of those participants, 20% indicated they would be interested in availability in multiple modalities (e.g., phone, computer); 30% mentioned they would like the GenAI-powered educational game to be similar to existing games and to have additional gamified features (30%) like more characters, better character design or more choices in the game narrative (e.g., ‘maybe a character that adapts using ai responses instead of general npcs with the same coded lines and interactions.’, ‘choice-based storyline’, or ‘customization, different choices, realistic visuals’); and 9% indicated an interest in personalization in learning.

5. Discussion

This study explored students’ experiences and perspectives on GenAI usage as well as their familiarity with games and educational games, with a focus on potential incorporation of GenAI. The findings are framed through the research questions and suggest implications for educators and educational game designers who wish to leverage GenAI in education/classroom.

5.1. RQ1—Based on Their Experiences, What Are Students’ Perceptions on GenAI in Everyday Life and Educational Contexts?

Survey results reveal that GenAI tools are being widely adopted by students, particularly for education-related uses. A majority of the participants reported weekly use of GenAI, with over 95% using it specifically for educational purposes. Interestingly, when asked about personal use, many of the examples brought up were also related to education. However, over half of the respondents characterized their use of GenAI as a supplemental resource rather than a primary one, suggesting that students leverage these tools primarily as supports to traditional instruction rather than replacements. This aligns with growing expert consensus that GenAI should be used to enhance, not replace, human guided teaching and learning (Kloos et al., 2024). Framing GenAI as an instructional support layer allows for more flexible, responsible, and gradual integration strategies and highlights the need for pedagogical design.

The results provide finer granularity on the types of tasks for which students find GenAI most valuable. Both open-ended responses and checklist selection responses converge on use cases such as concept understanding, brainstorming, search, and writing-related tasks such as drafting and editing. These findings align with previous studies which found GenAI tools are widely adopted in the educational context (Ng et al., 2025). Educators can leverage GenAI for developing cognitive support functions such as ideation, explanation, and content organization that could enhance student learning. Despite the high rates of perceived usefulness of GenAI (e.g., 97% of the students reported it as helpful), a majority of the students cited concerns about accuracy (78%), ethical use (36%), and lack of explainability (35%). These findings carry important instructional design implications for educators: GenAI holds promise as a learning support tool. However, educators must integrate it thoughtfully to mitigate risks of misinformation, over-reliance, and uncritical use. To support responsible integration, educators can (1) incorporate AI literacy instruction by teaching students how to evaluate AI-generated content for accuracy, bias, and reliability; (2) establish responsible use policies, such as requiring students to disclose and reflect on their use of GenAI tools in assignments (Guest et al., 2025); and (3) emphasize transparency and verification through classroom tasks that include explainability prompts (i.e., asking students to justify or critique GenAI outputs) or designing rubrics that reward thoughtful revision and AI-assisted drafting. Together, these strategies promote the effective and ethical use of GenAI in educational settings, ensuring it enhances rather than replaces human-guided learning.

Finally, student reports of current use along with suggestions for future uses of GenAI indicate a broad interest in expanding its educational applications. Subject-specific responses showed the greatest current utility for students in math, language arts, and science, with frequent use reported for assignments and writing. Importantly, the strong interest students expressed in future GenAI integration emphasizes the urgency of developing equitable and functional access, particularly for students who may be less familiar with these tools or lack adequate support.

5.2. RQ2—Based on Their Experiences, What Are Students’ Perceptions of Games and Educational Games?

Survey responses suggest that students are actively engaged with games, and particularly games played on mobile devices, PCs, and consoles. While general gaming trends skew toward mobile platforms, educational games are more commonly accessed via computers. Current topics for educational games are concentrated in math and language arts, but students expressed a desire to see future educational games expand into science. The desire for a shift in topics presents a clear design implication for game designers: educational games designers should optimize for computer use but also consider cross-platform compatibility for different environments. Developing versions that maintain consistent learning objectives and interactions across multiple devices (e.g., tablet, phone, or laptop) may enhance accessibility and sustained use. In addition, an expansion of topics covered by educational games to include sciences is warranted.

Interestingly, despite high general gaming familiarity, students showed relatively low familiarity with text-based games, with over 60% reporting lack of familiarity or uncertainty. Many students appeared to confuse text-based games with text-heavy visual games, such as Role Playing Games (RPGs). Nevertheless, among those who had experience with text-based games, features such as engaging storytelling (74%) and flexible, player-driven choices (65%) were highly valued. These suggest that while there might be a learning curve, students do see potential in narrative-driven experiences. However, it is worth noting that even students who are not familiar with text-based games are still well versed in text-based environments, as most student interactions with GenAI platforms are done in text-based environments. Together, these findings point to several considerations for game designers. To support user onboarding, developers should incorporate tutorials or example playthroughs that clarify the format and mechanics of text-based games. Enhancing text-based gameplay with optional visuals, diagrams, or text-to-speech features can lower the barrier to entry and accommodate diverse learning preferences (Cezarotto et al., 2022; Hassan, 2024). Additionally, developers should leverage existing GenAI platforms as a foundation for building narrative-rich, interactive learning games that align with students’ familiarity with text-based interfaces.

5.3. RQ3—What Are Students’ Thoughts, Desires, and Perceptions Regarding the Incorporation of GenAI in Educational Games?

Student attitudes reflect openness to the use of GenAI in games. Students were either supportive (41%) or open to the possibility (51%) of GenAI features in educational games, while 18% expressed hesitation or specific concerns (e.g., preferring human-designed game plots and visuals). This is partly due to expressed concerns around trust and lack of perceived utility (Đerić et al., 2025; Luo, 2025). Both educators and educational game designers should emphasize transparency and student agency in the design and implementation of GenAI-powered features. To achieve this, several actions are recommended: (1) developers should make features such as adaptive feedback, personalized hints, or scaffolding as opt-in supports rather than default settings, giving students control over how and when GenAI is used; (2) educators should clearly communicate how student data is used and how GenAI contributes to learning, as this transparency is the key to building trust; and (3) educators should focus on creating evaluation strategies that go beyond usability to include measures of learning outcomes, trust, and user acceptance metrics, ensuring that GenAI integration meaningfully supports pedagogical goals. Lastly, as students reported relatively low rates of educational game use in college, there is an opportunity for educators to reintroduce game-based learning at the post-secondary level through GenAI-powered systems that are relevant, personalized, and aligned with students’ evolving digital learning practices.

Student feedback via open-ended questions shows strong interest in using GenAI as a curated narrative engine within educational games. Participants mentioned GenAI’s potential to enhance storytelling (17%) and generate stories in various genres (25%). This aligns with earlier findings from students familiar with text-based games who preferred similar features as well. These insights suggest an opportunity to leverage GenAI for dialog variation and multiple narrative outcomes, along with creating diverse narrative genres, while operating under guardrails to preserve coherence and pedagogical integrity. As such, game designers should have narrative generation be guided by intentional guardrails that balance creative flexibility with educational structure. By carefully constraining and shaping the expressive outputs of GenAI, designers can deliver immersive, responsive narratives that remain aligned with educational goals.

Personalization also emerged as a critical area of value, in addition to narrative features. Students expressed interest in learning assistance (7.5%), game mechanics (e.g., characters, rewards, etc.), and adaptive experiences tailored to their needs (12%). These results highlight key design implications for game designers attempting to create effective GenAI-supported learning games: personalization features should be implemented in ways that actively support learning while maintaining user control. To do this, systems should provide clear, optional support such as tailored hints, feedback explanations or scaffolding prompts. These features should be visible and actionable, allowing students to engage with them as needed rather than by default. Such design choices not only enhance usability but also promote metacognitive awareness and learner autonomy. Implementing these design suggestions will also, in turn, contribute to fostering trust and transparency, which are essential for encouraging the adoption and sustained use of GenAI supported educational games. While many participants viewed GenAI’s role in educational games positively, particularly in enhancing game mechanics like characters, rewards, and personalization, others expressed skepticism about the quality of GenAI-generated content, especially in narrative coherence and visual design, indicating room for improvement in these features. These mixed perceptions may stem from differences in user experience; a post hoc observation suggests that more critical respondents could be frequent gamers or more literate GenAI users with higher expectations for creativity, accuracy, and system performance.

Finally, although not explicitly prompted in research questions, a notable pattern emerged around students requesting step-by-step support in their learning processes. Several open-ended responses described a desire for GenAI to function in a Socratic tutor-like role (Favero et al., 2024; Zare & Mukundan, 2015), guiding learners through problems interactively rather than simply supplying answers. This aligns with broader educational interest in inquiry-based learning approaches (Deák et al., 2021) and highlights the fact that designers of GenAI-supported educational games should integrate socratic prompting and conversational scaffolding to encourage active reasoning and deeper conceptual understanding. Incorporating such guidance mechanisms can transform GenAI from a content provider into a facilitator of reflective and inquiry-driven learning.

6. Conclusions

This paper offers insight into how students perceive, engage with, and envision GenAI tools in educational contexts, particularly within game-based learning environments. Survey results (n = 345) indicate widespread and frequent use of GenAI, especially for educational purposes, though students primarily view these tools as supplemental supports rather than replacements for traditional instruction. Additionally, engagement with educational games and text-based games varied, highlighting the need for clearer definitions, improved onboarding, trust-building, and alignment with students’ evolving digital habits and, in some cases, concerns about GenAI. Overall, the findings point to several key design implications for both educators and educational game designers. While this study centers around the student perspective, implications from the findings are applicable to educators and game designers. Future research can expand on these findings by incorporating perspectives from educators and designers to extend our understanding across diverse learning contexts. Additionally, usability studies on educational games and gamified tools using GenAI will expand on our understanding of student engagement with and preferences for GenAI incorporation. Exploring these complementary angles will help further validate and scale the design implications outlined in this work.