GenAI accelerates surface localization but not deeper culturally responsive teaching in UAE higher education

Introduction

The launch of ChatGPT in late 2022 triggered widespread and rapid shifts in higher education practices. Generative artificial intelligence (GenAI), powered by large language models, can generate text, images, audio, and other forms of content, providing creative opportunities for teaching, learning, and assessment (Istrate, 2026; Xia et al., 2026; Zawacki-Richter et al., 2024). Yet GenAI is not a culturally neutral resource. Because these systems are trained largely on globally dominant bodies of knowledge, they can reproduce Western-centric assumptions, cultural stereotypes, and normative framings (Agarwal et al., 2025; Watson et al., 2024) while presenting them as authoritative or universally applicable (Grab, 2025; Tao et al., 2024). Their speed and accessibility may therefore accelerate the circulation of culturally misaligned content and further privilege dominant knowledge systems unless faculty actively evaluate, revise, and contextualize their outputs.

These concerns are especially significant in the United Arab Emirates (UAE), where universities operate within a highly internationalized social and institutional environment. The country's population is predominantly expatriate (i.e., 88%; UAE Ministry of Foreign Affairs, 2026), and its universities bring together culturally diverse students and internationally recruited faculty, many of whom have been educated within Western higher education traditions (Aydogan et al., 2025; Eppard et al., 2021; Shahin et al., 2025a). In this context, GenAI may make it easier to adapt imported teaching materials and curricula by editing names, examples, language, or tone, yet such surface localization does not necessarily challenge the underlying assumptions or epistemic hierarchies embedded in those materials (Shahin et al., 2025b). Culturally responsive and decolonially informed integration therefore requires faculty to examine whose knowledge is represented, which perspectives are treated as authoritative, and whether GenAI-supported teaching meaningfully engages local cultural realities and ways of knowing rather than simply repackaging Western content (Satia and Ma, 2026; Vanthuyne et al., 2026). Against this tension between the accelerating adoption of GenAI in higher education and faculty members’ efforts to create locally tailored pedagogy for diverse student populations, this study examines how faculty evaluate, adapt, and localize GenAI outputs to support culturally responsive teaching (CRT) in UAE higher education.

GenAI in higher education

GenAI is increasingly embedded in higher education teaching, learning, research, and administrative practices. For faculty, GenAI tools can support course planning, material development, assessment design, feedback, and communication with students (Istrate, 2026). Their capacity to generate and revise content rapidly may improve efficiency and expand the range of resources available to educators (Wu, 2024). However, faculty adoption is not uniform; it is shaped by institutional context, disciplinary expectations, available resources, perceived usefulness, and organizational support (Arowosegbe et al., 2024; Luo et al., 2026). GenAI integration is therefore understood as a contextual pedagogical process rather than a worldwide technological development (Grab, 2025).

In this contextual adoption process, a central challenge concerns how faculty assess and use GenAI-generated outputs. Although GenAI can support idea generation and the exploration and synthesis of different perspectives, it does not assume responsibility for the accuracy, quality, or appropriateness of the final content (Liu et al., 2024). Responsibility remains with faculty, who must monitor, verify, and critically examine outputs throughout the human–GenAI collaboration process (Henadirage et al., 2026; Qiu et al., 2025). This requires evaluative judgement: the capacity to determine whether generated content is credible, fit for purpose, and pedagogically appropriate (Bearman et al., 2024). Such judgement extends beyond factual accuracy, because people's expectations of and interactions with AI are shaped by culturally specific understandings of agency, relationships, and technology that mainstream AI systems do not consistently reflect (Ge et al., 2024).

Widely used large language models tend to align more closely with values associated with English-speaking and Protestant European contexts (Tao et al., 2024). As a result, the outputs they generate may privilege particular cultural assumptions while presenting them as neutral or universally applicable (Watson et al., 2024). Evidence of this tendency is emerging across different contexts. For example, Naous et al. (2024) found that language models trained on widely used Arabic-language data sources, including Arabic Wikipedia, generated scenarios featuring Arab characters that disproportionately centered on poverty, with descriptors such as “poor” appearing consistently across models. Similar patterns were identified by Agarwal et al. (2025), who found that AI-generated writing suggestions shifted non-Western users toward Western writing styles and reduced indigenous forms of expression. Such findings suggest that GenAI outputs may unintentionally reinforce dominant cultural perspectives unless they are critically examined and adapted for local educational contexts.

This places additional responsibility on faculty to evaluate not only the factual accuracy of AI-generated content but also its cultural assumptions, representations, and pedagogical suitability. Although concerns regarding cultural appropriateness in education long predate GenAI, the CRT framework offers an established pedagogical approach for addressing them (Gay, 2002; Grab, 2025). CRT therefore provides a useful framework for examining how faculty can engage with GenAI while remaining attentive to students’ cultural contexts, local norms, and pedagogical needs.

Culturally responsive teaching as a theoretical framework

This study draws on the CRT as its guiding theoretical framework. CRT is an approach that uses students’ cultural knowledge, experiences, identities, and ways of communicating as meaningful resources for designing curriculum, instruction, classroom interaction, and assessment (Gay, 2002). CRT is especially relevant in GenAI-integrated higher education contexts such as the UAE. Faculty must translate disciplinary expectations and institutional standards into classroom practices that are meaningful and appropriate for students. This requires attention to English-medium instruction, culturally shaped interaction norms, and local expectations surrounding respectful communication (Aydogan et al., 2025; Gay, 2015). From this perspective, cultural responsiveness is applied through concrete pedagogical decisions, including how teaching materials are selected, cases are framed, feedback is communicated, participation is structured, and particular forms of knowledge are positioned as legitimate or relevant (Gay, 2010a).

Gay's CRT framework emphasizes that CRT requires more than general awareness of diversity (Gay, 2010a). It involves developing knowledge of learners’ cultural contexts and using that knowledge to design relevant curricula, communicate across cultures, build supportive learning communities, and create culturally congruent instructional practices (Gay, 2010b). Within this framework, three interrelated dimensions are particularly relevant to faculty work with GenAI: (a) cultural competence involving understanding and valuing learners’ cultural contexts; (b) critical consciousness, representing the interrogation of biases and assumptions that reproduce inequity; and (c) academic success, including the support for students’ academic flourishing through rigorous and culturally congruent teaching (Gay, 2010a).

These dimensions are especially important when GenAI becomes part of the pedagogical process. GenAI can support drafting, translation, localization, tone adjustment, and material development, but it cannot determine whether an output is culturally appropriate, misrepresentative, or likely to reproduce bias (Agarwal et al., 2025). CRT therefore helps distinguish surface localization, such as changing names, examples, or language, from deeper cultural responsiveness that also considers meaning, participation norms, representation, and epistemic authority (Gay, 2010a; Shahin et al., 2025b). Although GenAI may accelerate access to culturally situated materials, culturally responsive use ultimately depends on faculty members’ cultural knowledge, critical awareness, and interpretive judgement rather than on the technology alone (Grab, 2025; Liu et al., 2024).

The current study

Existing research has highlighted both the pedagogical potential of GenAI in higher education teaching, learning, and assessment (Istrate, 2026; Zawacki-Richter et al., 2024) and the cultural risks associated with its use, including bias and cultural misalignment in AI-generated outputs (Tao et al., 2024). Research has also emphasized that the educational value of GenAI depends on how faculty interpret, revise, and contextualize its outputs rather than on the technology itself (Arowosegbe et al., 2024; Bearman et al., 2024; Liu et al., 2024). However, relatively little is known about how faculty make these decisions in everyday teaching, particularly in highly internationalized and non-Western higher education contexts. Existing studies have examined faculty adoption, perceptions, and the broad potential of AI-supported CRT (Grab, 2025; Luo et al., 2026), but less attention has been given to how faculty judge whether GenAI-generated materials are culturally appropriate, how they respond to embedded biases and normative assumptions, and when they adapt, localize, filter, or reject particular outputs. This gap is especially important in the UAE, where globally circulating academic practices intersect with culturally diverse classrooms, English-medium instruction, and local expectations regarding communication, authority, social interaction, and respectful engagement (Aydogan et al., 2025; Shahin et al., 2025a).

Accordingly, the current study contributes to the literature in three ways. First, it shifts attention from a general GenAI adoption discussion and efficiency toward the cultural and pedagogical judgement involved in faculty use. Second, it provides contextually grounded evidence from UAE higher education, a setting that remains underrepresented in research on GenAI and CRT. Third, guided by Gay's CRT framework (Gay, 2002), it examines how faculty apply cultural responsiveness through decisions to evaluate, adapt, localize, filter, or reject GenAI outputs across teaching preparation, learning materials, student engagement, and assessment. In doing so, the study clarifies how CRT may be supported, constrained, or reshaped when GenAI becomes part of everyday pedagogical practice. The study addresses the following research question: How do faculty members use GenAI to support CRT in UAE higher education?

Method

A qualitative interpretivist approach was adopted for this study because our primary aim was to understand how faculty members make sense of, negotiate, and enact the use of GenAI for CRT within the culturally diverse higher education context of the UAE. Interpretivist approach is well suited to inquiries that foreground meaning making and the situated nature of practice, recognizing that participants’ accounts are shaped by the social, cultural, and institutional contexts in which they work (Denzin and Lincoln, 2011; Schwandt, 1994). Accordingly, we treated faculty members’ narratives as contextually grounded interpretations of their pedagogical decisions, how they assess cultural appropriateness, adapt GenAI outputs, and manage sensitivities in multicultural classrooms, rather than as neutral reports of “objective” behaviors (Denzin and Lincoln, 2011; Merriam and Tisdell, 2016). This stance aligns with a constructivist view of knowledge because the study examines how faculty make meaning of GenAI and culturally responsive teaching within their own institutional and cultural contexts. Rather than treating participants’ accounts as objective reports of practice, we understood the findings as co-produced through participants’ perspectives and the researchers’ interpretive engagement with the data (Lincoln and Guba, 1985).

Participants

The study employed purposive and snowballing sampling strategies to recruit participants at five major universities in the UAE (i.e., UAE University, Zayed University, Khalifa University, Abu Dhabi University, University of Sharjah). We selected these institutions because they enroll substantial numbers of Emirati students and employ internationally diverse faculty. To be eligible, faculty members were required to (a) hold a higher education degree (master's and/or doctoral), (b) actively utilize GenAI tools in their teaching practices, and (c) have had at least one year of teaching experience at a UAE institution. Invitation emails were sent to faculty members at the participating universities, inviting them to take part in an interview during the Fall 2025 semester. After each interview, participants were asked to suggest other potential participants meeting the inclusion criteria. Data were subsequently collected through interviews with 20 volunteer academics. As shown in Table 1, participants represented 10 nationalities, had high educational mobility, and were from a wide range of disciplines reflecting a wide range of cultural backgrounds and experiences.

Table 1. Participant demographics.

NA: data not available. In case of multiple degrees from different countries, the last degree obtained was recorded.

Data collection and analysis

The interview questions were developed in alignment with the study objectives and relevant literature on CRT and GenAI in higher education. Using a semi-structured format, the protocol included open-ended questions around (a) cultural considerations in teaching Emirati students, (b) how participants use GenAI to design culturally relevant materials and learning activities, and (c) how they evaluate the cultural appropriateness of AI-generated outputs. The protocol was first developed by aligning the study's research question with relevant literature and Gay's CRT framework, particularly cultural competence, critical consciousness, and academic success. The initial questions were then reviewed by two experts with higher education and GenAI research and teaching backgrounds, whose feedback informed revisions to wording, focus, and sequencing. Two pilot interviews were subsequently conducted, and insights from these interviews were used to refine the final version of the semi-structured interview protocol. Probing questions were included to encourage participants to elaborate on how they judged cultural fit, revised GenAI outputs, and responded to possible bias, stereotypes, or culturally sensitive content.

Following ethical approval (Approval number: ZU25_036_F) data were collected through 20 interviews conducted online via Zoom. The final sample size of 20 participants was considered appropriate for the study's qualitative interpretivist design because the interviews generated rich and recurring accounts across institutions, disciplines, and faculty backgrounds. All interviews were audio-recorded with consent and transcribed verbatim. We employed thematic analysis to identify patterns of meaning across the dataset (Braun and Clarke, 2006). In line with this interpretivist approach, the analytic aim was to extract meaning by segmenting transcripts, assigning codes, examining overlap across codes, and consolidating these into higher-order categories and themes (Creswell, 2013). Analytic memos also captured salient interactional cues (e.g., pauses, emphasis, and noticeable shifts in tone) when these provided context for interpreting participants’ intended meanings (Norris, 2013).

Analysis proceeded inductively through iterative cycles of coding and theme development rather than through predetermined categories. The second and third authors repeatedly read the transcripts for familiarization and generated initial codes across the full dataset. Coding was collaborative rather than a reliability-based independent coding process; the aim was to develop shared interpretations through discussion and repeated return to the data, rather than to calculate agreement. Codes were reviewed, compared, and clustered into higher-order categories using constant comparison to refine boundaries and relationships among codes (Merriam and Tisdell, 2016; Saldaña, 2016). When differences in interpretation emerged, the research team revisited the relevant excerpts and refined the codes, categories, or theme boundaries until the interpretation was supported by the data. The first author contributed an additional reflexive perspective by challenging emerging interpretations, requesting supporting transcript evidence, and examining theme coherence.

Member checking was not conducted because participants requested only one point of contact, citing concerns about job security and the sensitivity of discussing institutional practices. However, trustworthiness was supported through analytic memoing, peer debriefing, repeated team discussions, and returning to the transcripts throughout theme development. As shown in Table 2, the final analytic structure consisted of four themes with associated categories and codes. To make the coding process more transparent, Table 3 provides a sample thematic map showing how selected raw interview excerpts were developed into initial codes, categories, and final themes.

Table 2. Thematic framework and code tree GenAI and CRT.

Table 3. Sample thematic map and coding process.

Authors’ reflexivity statements

As a research team, we recognize that our positionalities and professional experiences shaped how we interpreted participants’ accounts and how we constructed patterns of meaning in the dataset. We approached analysis reflexively attending not only to what participants explicitly stated but also to the underlying assumptions, cultural logics, and tensions that organized faculty talk about GenAI and CRT. The team's disciplinary backgrounds in multiculturalism, international higher education, and language teaching in diverse classrooms sensitized us to issues of cultural responsiveness and power in classroom communication, while also posing a risk of over-privileging pedagogical interpretations over technical ones. Because the first two authors work within a majority-Emirati public university, we were mindful that participants might frame their responses in institutionally and politically acceptable terms; we therefore treated hesitations, indirect speech, and ambiguity as meaningful data rather than as gaps. As regular users of GenAI for teaching and academic work, we were attentive to the ways our own comfort with these tools could shape what we interpreted as effective or culturally responsive use. Additionally, interviews were conducted in English, and we remained attentive to how English-as-a-lingua-franca may have shaped participants’ phrasing, especially when discussing culturally sensitive issues.

The first author is a Western-educated faculty member with approximately 10 years of experience in UAE higher education, which supported contextual sensitivity to local institutional norms and classroom dynamics but also risked taking familiar practices for granted. The second author is also Western-educated and has been based in the UAE for approximately 3 years; this position offered both a comparative lens and an awareness of the interpretive work required when navigating local cultural expectations as a relatively newer member of the system. The third author is trained in another Middle Eastern country and resides outside the UAE, contributing analytic distance that helped surface implicit assumptions in our readings and prompted us to justify interpretations with evidence across transcripts. All authors use English as a second language, and we remained reflexive about how this shared linguistic positioning could influence both interview interaction (e.g., word choice, politeness strategies, and indirectness) and analytic interpretation (e.g., how we understood nuance, emphasis, or culturally loaded terms). Throughout the project, we used memoing, peer debriefing, and repeated return to the transcripts to interrogate our assumptions and preserve complexity, including ambivalence and critique, as analytically meaningful.

Results

We organized participants’ collective accounts into four overarching themes: (a) Culturally responsive pedagogical strategies using GenAI, (b) Faculty sensemaking of GenAI for cultural responsiveness, (c) Cultural and relational benefits of GenAI-supported teaching, and (d) Institutional, ethical, and cultural risks/barriers. These themes helped us explore and understand how faculty members in the UAE make sense of GenAI and how they use it to support culturally responsive pedagogy in this diverse learning environment. In the next sections, we present each theme and its subcategories, drawing on participants’ own words using pseudonyms assigned by the authors. Direct quotations are provided to preserve participants’ intended meanings and to represent the range of perspectives reflected in the dataset, consistent with the study's interpretivist approach.

Theme 1: culturally responsive pedagogical strategies using GenAI

Most participants agreed that using GenAI strengthened their pedagogical approach by supporting content localization, language and communication scaffolding, and the design of culturally relevant engagement-focused learning experiences. GenAI tools provided speed and easy access to teaching materials for faculty who are often unfamiliar with the local students’ cultural characteristics. Khalid said “it [GenAI] is saving time, giving good and quick ideas, generate a lot of fantastic outcomes.” A salient theme was that participants described their teaching experiences and materials as deeply embedded in Western knowledge and examples, which they viewed as a major challenge for pedagogical adaptation; GenAI tools came into play as a pragmatic resource for reworking, localizing, and reframing these materials to better fit students’ cultural contexts.

Localizing content happened in multiple ways. Creating case scenarios with UAE's cultural dynamics was often emphasized by the faculty members. These participants valued GenAI's support in finding information about local companies or government offices providing students a connection between global and local perspectives. Katerina recalled “in one of my courses about leadership, [using GenAI] I might go with an example of a local company, let's say, an Emirati company, which is very well known for the students in the country, rather than using global examples.” Importantly, participants offered mixed assessments regarding the extent to which localization amounted to superficial substitutions (e.g., changing names or minor contextual details) versus more substantive integration of culturally relevant elements into the narratives and scenarios embedded in teaching materials. As Dana put it: “changing the company name from Amazon to Noon or Kareem is very superficial.”

A dominant narrative centered on the language and communication scaffolding role of GenAI. Participants valued the ability to manipulate language and its content based on learners’ needs. They welcomed GenAI for translating teaching materials into what Dana described as a “relatable and acceptable” medium for local students. Adaptation through translation ranged from cross-language translation to adjusting culturally specific nuances. Andreas aptly described this: “I'm in a business school, and so very often I get a case meant for the US, or the UK, or Australia. It's meant for native English speakers in a particular environment. I'll very often take that, use Copilot or ChatGPT to clean it from idioms and phrases non-English speakers cannot understand… It is like English-to-English translation.”

Across interviews, a clear pattern emerged that faculty used GenAI to polish the tone of their communication with students. These included assignment descriptions, announcements, emails, or even in-class talks. Sean navigated cultural differences in communication styles using GenAI as a mediator: “Sometimes I will paste an email and ask it to make it more polite, because you know, culturally here, you need to be careful how you phrase things.”

Participants also positioned GenAI-enabled multimedia and gamification as pedagogical tools for culturally aligned engagement. Moayad described how interactive games helped him maintain students’ attention and classroom momentum: “Because they get bored very quickly, you know?… So in order to keep the class interactive and ongoing, I usually put some gaming… using HTML5,” Yet participants also cautioned that cultural responsiveness is not automatic; visuals and examples must be reviewed for audience fit, including tone, appropriateness, and relevance. Sean captured this ongoing need for judgement: “If you ask it to create visuals or examples, it can do it quickly, but you still need to check if it fits the audience.”

Theme 2: faculty sensemaking of GenAI for cultural responsiveness

Across interviews, participants made sense of GenAI less as an answer machine and more as a pedagogical tool that must be guided, edited, and ethically constrained by the educator. Their accounts clustered around (a) clarifying the role of AI versus the role of the educator, (b) developing practical decision rules to make outputs fit culturally, and (c) articulating adoption stances that ranged from professional growth orientations to resistance and heightened awareness of bias and stereotyping.

A dominant narrative positioned GenAI as extending capacity rather than replacing relational pedagogy and human judgement. As Rana emphasized, “..it's a tool. It doesn't replace a human being… We created that. The most important thing, it's a tool that helps us improve [higher education] further.” Similarly, Lina framed GenAI as a way of scaling responsiveness while retaining the human role: “It's not about replacing the human role. We need to always have that human role, but it's all about extending our capacity to be responsive to student needs.” Faculty repeatedly located cultural responsiveness in the educator's relational work, facilitation, mentoring, and emotional attunement, rather than in the technology itself. Theresa described this stance as facilitative teaching: “I'm there as a facilitator, I'm there to guide them, but I'm not there to force knowledge into them, okay?” Rana made the relationship explicit: “I'm not a teacher. I'm not a teacher to you. I'm a mentor to you, you know?” Andreas further anchored cultural responsiveness in affective human qualities GenAI cannot replicate: “generative AI is not human. It's just gonna take whatever the average is, right? I feel like this empathy, these feelings that you do not see in AI, you know, this is what I care about.”

Participants described “fit” as something they produce through inputs and human-in-the-loop revision. Yulia was direct that cultural responsiveness depends on intentional prompting: “the weakness is generally just a lack of specificity in its… generalities, unless you prompt it in a different way. …, AI will never give you a culture-responsive content” This logic was echoed as a rule of thumb about inputs: Lina noted, “I'll put the case study and then I'll ask, okay, can we just change this into something that is less Westernized,” while Sean summarized it aptly: “the output is dependent on the input. Garbage in, garbage out kind of things.” Faculty also described routine editing as part of cultural care, especially around sensitive phrasing. Lina explained, “I check them. I read it, I check it, some of the scenarios were not really possible or cannot be used in the classroom… and I just erased the scenarios” Rami similarly described his role as explicitly adaptive: “So, most of the roles that I do here would just be filtering and editing and adapting the outputs that AI will give me to resonate within the Emirati culture.” Alongside verification and triangulation, participants warned against taking outputs for granted. Sean used a vivid metaphor for over-trusting outputs: “you have to know that you are driving into a river, no matter what the device says… it's still at a stage where it occasionally directs you to drive your car straight into the river, and you've got to know enough.”

Participants’ orientations ranged from openness and professional growth to strong resistance grounded in perceived risk. For some, GenAI was tied to staying current: Theresa reflected, “Well, it's… it's the buzzword. Everyone's talking about AI nowadays… if I failed to update my knowledge, or failed to evolve with what's going on in the world, then it's not going to be helping me…” and Rana similarly noted, “I like always to be up-to-date, because I believe this is part of professionally developing myself as well.”

Theme 3: cultural and relational benefits of GenAI-supported teaching

Across interviews, participants described benefits of GenAI that extended beyond efficiency, including inclusive access, stronger engagement and motivation, and greater cross-cultural connection and belonging. These outcomes were linked to concrete practices, subtitles, translation, interactive activities, and rapid contextualization, rather than assumed as automatic.

Participants noted that GenAI-supported materials could reduce participation barriers for students with different communication needs. Michael highlighted real-time accessibility supports: “with the new technologies now, we use online screens, BenQ for students who cannot catch up with me, or who have problems.” Omar also referenced hearing-related support: “in hearing problems, this [GenAI-based subtitles] would be a one way of inclusion”. Translation and multilingual delivery were similarly framed as inclusion strategies. Yannis noted that students “can, ask the translation and can read it… in many different languages,” while Arif emphasized real-time language support in English-medium settings.

Participants framed cultural responsiveness as relational design, structuring classroom interaction so students can participate safely and feel a sense of belonging across differences. Yulia described culture work as managing who works with who in mixed-gender settings: “it's thinking about who's working with who, especially now that we’ve got mixed-gender classes. So who's working with who,” alongside “being aware of… culturally sensitive topics.” Similarly, Lina emphasized adjusting examples and group processes to reduce discomfort and support inclusion: “now we changed into the mixed gender classes. So here is where… we had to change a bit,” and “for me… it's about gender and culture and how you can tailor the class in a way that respects… the fact that they’re not really comfortable with each other,” adding that in grouping and workplace-prep discussions “you always make sure that they feel comfortable.” Lina also linked this relational work to resisting stereotyped “culture” shortcuts in GenAI outputs, when prompts over-localize, “you immediately get the camel, the desert and Ramadan defaults, and that this is not the reality.”

Theme 4: institutional, ethical, and cultural risks/barriers

Across interviews, participants emphasized that GenAI use in CRT is constrained by institutional governance and resources, academic integrity pressures, and cultural/ethical risks that require careful judgement rather than automated adoption.

Participants described uneven institutional guidance, ranging from permissive policies to discouraging stances, shaping how confidently faculty could integrate GenAI. Andreas noted, “our policy is that the faculty members just have to be very clear about what their AI policies are, but it's a pretty inclusive policy,” while Fadi contrasted this with a more restrictive stance: “So our policy so far in computing, we discourage students, from using, actually, generative AI.” Participants also pointed to the need for capacity-building. Yannis described institutional efforts: “We have professional development program… part of it about AI, a certificate about how to incorporate AI. “Yet Theresa observed that cultural dimensions were often missing: “in the training. there wasn't much talk about culture, honestly.” Resource constraints further limited uptake; Yannis raised cost and access concerns: “my university does not, pay for our subscription… Who's gonna pay for it?”

A dominant concern was how GenAI complicates authorship, effort, and detection. Rana described the dilemma of “too perfect” submissions: “You know that you have a level D or F students, and then his assignment is perfect. We are all suffering from this.” Others highlighted the limits of detection tools and the ambiguity of what “counts” as AI-assisted work; Saad captured this tension: “we are promoting it; students are using it… and sometimes AI is also not detected.”

Participants emphasized that culturally responsive practice requires boundaries around sensitive content, active editing, and skepticism toward outputs. Yannis stated plainly, “We avoid to discuss politics, we avoid to discuss religion, we avoid to discuss sensitive issues,” while Rana described human-in-the-loop screening as a safeguard: “if there are certain phrases that I feel like it's culturally sensitive… I change it.” Several participants warned that GenAI can amplify stereotypes; Khalid noted “a kind of bias… describing women… that all of them wear abaya… and Emirati, all of them rich,” and Dana cautioned that it can be “fostering more stereotypes.” Privacy and data sensitivity were also salient; Rami drew a clear boundary: “I cannot give Real-life data with them.” Finally, trust and quality problems were framed as an ever-present risk; Rana warned, “Whatever you say, he will agree with you,” and Andreas advised, “double-check everything… never do something with generative AI that then you submit without really… re-reading.”

Discussion

The aim of this study was to examine how faculty in UAE higher education make sense of and practically use GenAI to support CRT in culturally and linguistically diverse classrooms. The findings contribute to emerging scholarship on GenAI adoption by showing that instructors largely positioned these tools as capacity-extenders for localization, communication, and instructional accessibility. They also reveal the cultural and institutional conditions that determine whether GenAI-supported localization develops into meaningful pedagogical responsiveness or remains at the level of surface adaptation. Viewed through Gay's CRT framework, participants demonstrated clear efforts toward cultural competence and support for academic success, while evidence of critical consciousness was more uneven (Gay, 2010a). These findings suggest that while GenAI may facilitate certain dimensions of CRT, it cannot independently provide the cultural knowledge or critical awareness on which CRT depends.

A core contribution of the study was showing that faculty framed GenAI as a practical accelerator for CRT in the UAE, especially when their teaching materials were “embedded in Western knowledge and examples” that felt distant from local cultural references and institutional realities. Participants described using GenAI to generate locally situated examples, cases, scenarios, and activities that would otherwise require substantial time to develop. Such localization could involve adjusting names, settings, communication styles, family relationships, workplace expectations, or social situations so that materials were more recognizable to Emirati and wider Gulf-region students. These practices reflect Gay's (2010a) concept of cultural competence because they require faculty to understand students’ cultural contexts and use that knowledge to make instruction more relevant. They also respond to a setting in which internationally recruited faculty often work with curricula and resources developed within Western higher education traditions (Shahin et al., 2025a). A latent risk, however, is that some faculty may treat GenAI as a substitute for, rather than support for, the continuing development of their intercultural competence. This could produce surface relevance while deferring deeper cultural learning. This concern is reflected in previous research, which warns that unmediated GenAI may reinforce stereotypes and marginalize cultural perspectives (Naidu and Sevnarayan, 2025). Likewise, Naous et al. (2024) further found that even language models trained on widely used Arabic-language sources may produce culturally biased outputs.

Despite their enthusiasm for GenAI, faculty repeatedly identified mentoring, facilitation, emotional attunement, and judgement as human dimensions that the technology could not replicate. GenAI was described as extending faculty capacity rather than replacing the relational work through which students feel seen, supported, and guided. This finding is coherent with the relational character of UAE higher education described by Eppard et al. (2021), Moore-Jones (2015), and Shahin et al. (2025a). For example, Eppard et al. (2021) found that Emirati students valued collaborative learning and peer support, while Shahin et al. (2025a) showed that Emirati undergraduate research teams functioned as relational spaces where mentorship and shared responsibility shaped Emirati students’ academic growth. This also reflects CRT's emphasis on cultural caring and community-building, whereby responsiveness is enacted through the ways instructors structure interaction, communicate respect, and sustain a learning climate that supports engagement without violating local norms surrounding face, authority, and participation (Gay, 2015). The finding also supports literature arguing that responsibility for pedagogical meaning-making remains with the human instructor, even when GenAI assists with drafting and ideation (Liu et al., 2024). Henadirage et al. (2026) similarly emphasize the need for active human oversight, while Bearman et al. (2024) highlight the role of evaluative judgement in determining whether outputs are fit for purpose.

Participants described benefits extending beyond efficiency, including subtitles, translation, interactive activities, and rapid contextualization, which they linked to inclusive access, engagement, and classroom belonging when paired with purposeful instructional design. These benefits were associated with concrete practices, including subtitles, multilingual explanations, simplified wording, interactive activities, and structured collaboration, rather than being treated as automatic effects of the technology. Such practices are particularly relevant in UAE higher education, where English is commonly used as the language of instruction while many students primarily communicate in Arabic outside the classroom. GenAI could help faculty provide alternative explanations and bridge unfamiliar academic language, but instructors still had to ensure that simplification did not reduce disciplinary meaning or position students’ linguistic backgrounds as deficits. This supports the argument that AI can scaffold instruction around learners’ linguistic and cultural realities when educators intentionally design for accessibility and participation (Zawacki-Richter et al., 2024). From the perspective of Gay's CRT framework, these practices can be understood as creating conditions that support academic success by improving access, comprehension, participation, and belonging (Gay, 2010a). They also suggest that AI-supported linguistic scaffolding may become part of culturally congruent instruction in multilingual higher education settings.

Participants’ emphasis on localizing examples, polishing tone, translating language, and achieving cultural fit can be interpreted primarily as cultural competence at the level of materials and communication (Gay, 2010a). Yet their accounts were comparatively thin on questions of power and knowledge politics, including who is represented, whose knowledge counts, and how curricula may reproduce hierarchy through what is taught and what remains unspoken (Shahjahan et al., 2022). Culturally responsive practice therefore risks being framed mainly as adaptation—making Western materials more locally recognizable or acceptable—rather than transformation that re-centers local epistemologies and interrogates dominant frames. As Gay (2010b) emphasizes, genuinely responsive teaching depends on educators’ culturally informed interpretation of students’ actions, needs, and strengths, not merely on the presence of familiar names, examples, or settings.

The findings therefore suggest a refinement to Gay's CRT framework when applied to GenAI-integrated teaching. Cultural competence must involve not only knowing students and designing culturally congruent instruction, but also recognizing how GenAI mediates which cultural representations, values, and knowledge sources become readily available to faculty. AI-generated writing suggestions may similarly shift non-Western users toward Western forms of expression and reduce cultural distinctiveness (Agarwal et al., 2025). In Arab contexts, Naous et al. (2024) found that language models reproduced stereotypical associations involving poverty and traditionalism. GenAI may therefore accelerate a pattern of “cultural substitution without epistemic transformation” (Shahin et al., 2025b, p. 1) by making it easier to insert UAE names, places, and culturally acceptable language while leaving Western assumptions and authority structures intact. CRT in GenAI-integrated teaching thus requires cultural competence to include awareness of how the technology filters, prioritizes, and reproduces knowledge. Without this awareness, localization may improve surface relevance while leaving questions of positionality, representation, and epistemic authority largely unchanged as noted by the findings of the current study.

Faculty also emphasized that GenAI-supported CRT was bounded by institutional conditions, including uneven policies, limited training, and disparities in access to tools. Academic integrity pressures, privacy concerns, and boundaries surrounding sensitive topics further shaped what instructors felt able to do in practice. These findings align with previous research showing that faculty adoption is influenced by institutional context, available resources, and organizational support (Luo et al., 2026), while the responsible pedagogical integration of GenAI requires clear governance and guidance (Istrate, 2026; Zawacki-Richter et al., 2024). More broadly, GenAI adoption is occurring rapidly, often ahead of coherent governance, pedagogical guidance, and culturally grounded professional development (Arowosegbe et al., 2025; O’Dea, 2024; Wu, 2024). These constraints may be amplified in the UAE because faculty must navigate local expectations concerning appropriate classroom discourse, respectful communication, sensitive topics, and institutional risk (Aydogan et al., 2025). This aligns with Singh et al. (2021) and Williams et al. (2025), who similarly emphasize the importance of culturally situated norms and boundaries within UAE educational contexts.

Limitations and future research

While this study makes a timely contribution to understanding GenAI use for cultural relevance in UAE's rapidly evolving higher education landscape, the findings should be interpreted considering several limitations. First, the sample is likely affected by selection bias. Although this qualitative work did not aim for statistical generalizability, our selection criteria might have skewed the dataset toward faculty who already use GenAI or who are more invested in CRT. Faculty with less confidence or competence in either GenAI or culturally responsive pedagogy may hold different views that are not adequately represented here.

Second, the disciplinary distribution of the sample also limits transferability, as participants were more strongly represented in education, humanities, and social science-related fields than in STEM, health sciences, or technical disciplines, where GenAI use may take different forms. In addition, participants’ years of experience in the UAE varied substantially, which may have shaped their familiarity with local cultural norms, institutional expectations, and confidence in evaluating culturally appropriate GenAI outputs.

Third, the study relies on faculty self-reports, which are inherently vulnerable to social desirability effects, particularly within UAE higher education, where faculty mobility is high and job security can feel uncertain. Additionally, because the first two authors are faculty members at a public university in the UAE, our institutional proximity may have shaped the interview context and participants’ responses. Participants may have framed their practices in ways they perceived as professionally acceptable within local policy, cultural norms, or institutional expectations. Another limitation of the study was that participants’ accounts offered limited evidence of measurable learning outcomes (e.g., gains in critical thinking), changes in belonging, or student voice that would confirm whether these practices are experienced by learners as genuinely culturally responsive.

Future research should, therefore, move beyond faculty perceptions alone by incorporating student perspectives and, where possible, learning evidence. Mixed or multi-perspective designs could examine not only how faculty use GenAI, but also whether students experience these GenAI-supported pedagogical choices as culturally respectful, engaging, or alternatively, patronizing or surveillant. More broadly, future studies could attend to both ends of the process: the instructor's GenAI-supported design decisions and students’ lived experience of those decisions in the classroom.

Implications for theory

The findings extend CRT theoretical framework by showing that cultural responsiveness becomes a technologically mediated form of pedagogical judgement when GenAI is incorporated into teaching. More specifically, as shown in Figure 1, the findings suggest two adaptations to Gay's CRT framework for GenAI-mediated higher education. First, cultural competence should be expanded to include cultural-technological judgement: the capacity to evaluate how GenAI selects, prioritizes, and represents cultural knowledge rather than merely whether its outputs appear locally familiar. Faculty use of GenAI supported cultural competence and academic success through localized examples, language assistance, accessible communication, and adapted teaching materials. However, the mixed evidence of critical consciousness indicates that these benefits do not automatically produce deeper cultural responsiveness. GenAI may therefore accelerate certain dimensions of CRT while leaving questions of representation, bias, epistemic authority, and power insufficiently examined. Within an adapted framework, critical consciousness should encompass scrutiny of both the teaching material and the technological systems, training data, and cultural defaults through which that material is generated.

Second, in practice, cultural-technological judgement means evaluating GenAI outputs not only for accuracy, but also for cultural fit, representation, bias, and whose knowledge is included or excluded. This judgement differs from general GenAI evaluation because it focuses on whether AI-generated materials support deeper cultural responsiveness rather than only surface localization. Faculty development can support this capacity through training that combines GenAI literacy, local cultural knowledge, CRT, and guided practice in adapting, filtering, or rejecting outputs.

Third, Gay (2002) originally emphasized that CRT requires educators not only to draw on students’ cultural knowledge and experiences but also to question dominant assumptions, inequities, and patterns of representation. The present findings do not revise this principle; rather, they show that it becomes more visible and consequential when teaching materials are generated through AI. The distinction between surface localization and epistemically responsive integration helps clarify this implication: surface localization changes names, settings, language, or tone, whereas epistemically responsive integration examines whose knowledge is legitimized, which perspectives are absent, and whether dominant assumptions remain embedded beneath locally familiar content. This process cannot be theorized solely as an individual faculty competence, because participants’ practices were also shaped by policies, access to tools, privacy concerns, academic integrity pressures, institutional risk, and uncertainty surrounding culturally sensitive issues. Consistent with Satia and Ma (2026) and Vanthuyne et al. (2026), culturally responsive GenAI use should therefore be understood as a sociotechnical and institutionally mediated practice requiring collaborative teaching cultures, institutional recognition, and structural support. In the UAE and Gulf contexts, a decolonial-informed application of CRT must also examine how English-medium education, imported curricula, colonial legacies, global university hierarchies, and institutional power shape which forms of knowledge are legitimized. Rather than transferring decolonial models developed elsewhere without adaptation, theory should remain attentive to how local, Arab, Islamic, Gulf, migrant, and global knowledge traditions interact within the region's distinctive social and educational conditions.

Implications for practice and conclusion

Because participants used GenAI mainly to support localization, communication, accessibility, and learning design, UAE universities should first develop context-specific guidance that moves beyond simply permitting or restricting its use. Such guidance should clarify expectations for verifying factual and cultural accuracy, protecting student and institutional data, and reviewing outputs involving sensitive topics such as religion, gender, family, nationality, and politics. It should also promote two-way transparency by encouraging faculty to explain when GenAI has been used in teaching or assessment materials and how its outputs were reviewed.

Second, culturally responsive GenAI requires more than effective prompting. Universities and developers should prioritize multilingual, locally grounded, culturally representative, and regularly bias-audited systems. Their development, procurement, and evaluation should involve Emirati and expatriate faculty, students, Arabic-language specialists, cultural researchers, instructional designers, and accessibility experts. Regular qualitative and quantitative review should assess not only factual accuracy and Arabic–English performance, but also representation, stereotypes, accessibility, student experiences, and the cultural appropriateness of generated materials.

Third, faculty development should combine GenAI competence with local cultural knowledge, critical reflection, and decolonial awareness. Training should help faculty recognize stereotypes, examine whose knowledge is absent or privileged, and distinguish surface localization from deeper cultural responsiveness. Universities should also support the development of locally grounded Arabic–English cases and resources and recognize this work through time, funding, and institutional reward structures. Overall, GenAI can strengthen CRT in UAE higher education, but only when technological capacity is accompanied by sustained human judgement, local participation, and institutional accountability for cultural and epistemic power.

In conclusion, this study shows that GenAI can help faculty reduce cultural and linguistic mismatches, expand accessibility, and produce contextually relevant teaching materials more efficiently. Yet the same technology can reproduce stereotypes, obscure Western epistemic defaults, and create a misleading appearance of cultural responsiveness. Its contribution to CRT therefore depends less on the sophistication of the technology than on the quality of the human and institutional judgement surrounding its use. Culturally responsive GenAI in UAE higher education requires culturally representative and bias-audited systems, ongoing mixed-method evaluation, faculty development that combines AI competence with local cultural and critical knowledge, meaningful student and community participation, and institutional accountability for unequal power and knowledge structures. When these conditions are present, GenAI may extend faculty capacity without displacing the relational, ethical, and interpretive work at the centre of CRT. Without them, it risks making imported knowledge appear locally familiar while leaving its underlying assumptions and hierarchies largely unchanged.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The authors verify and take full responsibility for the use of generative AI in the preparation of this manuscript. Generative AI was used to support grammar editing, language refinement, and clarity of expression. All content, interpretations, and final revisions were reviewed and approved by the authors.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Nabiha El Khatib, Lebanese International University, Lebanon

Reviewed by: Rajagopal Sankaranarayanan, Texas A and M University, United States

Salmon Oliech Owidi, Tom Mboya university, Kenya