The reason for choosing a digital, rather than conventional, diet ethnography was that participants were hard-to-reach via conventional methods in China during COVID-19 pandemic (35). This study was completed in three phases, as summarized in Figure 1. Phase 1 included a feasibility study in China where university students reported eating occasions via a social networking mobile application. Phase 2 included the China cohort and phase 3 included the New Zealand cohort of the main digital diet ethnography, respectively. In this section, we provide necessary details of the theoretical framework; role of the ethnographers; the food environment setting; recruitment, data collection, and data analysis methods. For full details of the methods, please refer to Supplementary file 1. We followed the Consolidated Criteria for Reporting Qualitative Research (COREQ) guidelines to maintain thorough and transparent reporting throughout the study (36) (Supplementary file 2).

In our digital diet ethnography, a theoretical framework was chosen only to inspire our inductive grounded theory analysis. In the New Zealand cohort, Bandura's Social Cognitive Theory (SCT) was integrated with Bronfenbrenner's Ecological Model to construct a composite theoretical framework, tailored to the university student population (37). In the China cohort, Contento's (38) theoretical framework was employed because both intrapersonal and environmental factors were considered (23).

Reflexivity acknowledges that researchers' backgrounds, experiences, and assumptions inevitably shape data collection, interpretation, and analysis in ethnographic work. The New Zealand cohort was conducted by a male postgraduate student at the University of Auckland as part of his master's thesis. His own nutrition literacy, recent university student identity, and shared experiential context with participants influenced which themes he perceived as salient and how he interpreted participants' rationales for food choices. Although he produced detailed pre- and post-data-collection positionality statements that explicitly addressed his personal and intellectual goals, social identity, unconscious biases, and the implications of a male researcher studying predominantly female participants' eating behaviors, he did not maintain an ongoing reflexive journal during data collection and analysis, limiting the ability to fully trace potential researcher effects in real time.

The China cohort was conducted by a different male postgraduate researcher at the University of Auckland as part of his Ph.D. thesis. Having lived in China for the first 20 years of his life, originating from the same city (Shijiazhuang) as the study campus, and having previously completed a comprehensive food environment audit on that exact campus (29), he possessed deep contextual familiarity with local food culture, dietary practices, and campus-specific constraints, enabling particularly accurate and nuanced interpretation of participants' choices and rationales. His subsequent 10 years of tertiary education in English-speaking countries provided a strong command of English-language dietary behavior theories, allowing seamless translation and theoretical mapping of Chinese-language themes without loss of cultural nuance. Being a university student himself at the time of data collection further reduced social distance and the risk of misinterpretation of student lifestyle factors. Taken together, the researchers' insider student status, cultural knowledge, and prior campus-specific experience strengthened contextual credibility and thematic resonance across both cohorts, while explicit positionality reflections helped mitigate potential biases stemming from gender, personal eating habits, or theoretical preconceptions.

The New Zealand cohort was conducted at the University of Auckland, with most students falling between the ages of 21 and 23 (39). Food service was provided by privately-owned food vendors. The campuses were in urban areas in Auckland where external food vendors surrounding the campus were readily accessible. The China cohort was conducted in two urban campuses of Hebei Medical University (HBMU), located in Shijiazhuang City, China. Most students came from middle- to low-income areas. Food service was provided by university-owned, government-subsidized food outlets. All students were boarding mandatorily. In both cohorts, a convenient sample of participants was recruited from the general student population.

This study followed standard ethical procedures for digital ethnography (40), with approvals obtained from the University of Auckland Human Participants Ethics Committee (Reference UAHPEC23986, 4 April 2022, for three years) for the New Zealand cohort and the Research Ethics Committee of HBMU (Approval 202300037) for the China cohort. All participants received detailed Participant Information Sheets and signed informed consent forms that outlined the study aims, procedures, potential risks and benefits, voluntary nature of participation, confidentiality measures, de-identification processes, and the right to withdraw at any time. To protect participant welfare, the researchers avoided exploiting any information that could harm self-esteem, emotions, or well-being; maintained awareness of personal and interpersonal contexts; and worked under qualified supervision throughout. Data were collected via the secure, password-protected Indeemo platform, accessible only by the researcher(s); all identifiable information was de-identified prior to analysis, storage, or presentation. Ethnography data were disclosed publicly only with participant permission and full anonymisation. All data were stored on password-protected university servers, with Indeemo data automatically deleted after project completion.

The China cohort ethnographer pilot-tested the feasibility of digital diet ethnography among 6 students at HBMU between 17 and 23 Jun 2021. Participants provided photos, videos, and text of seven days' dietary intake. Findings from the feasibility study were used to improve the ethnographers' skills of approaching participants and managing their responses.

Participants for the New Zealand cohort were recruited by convenience sampling via University of Auckland social media channels, emailed enquiries to the student researcher, and campus flyers. Eligibility required being a current University of Auckland student (≥18 years), mobile-device literate, willing to record eating occasions, and not enrolled in nutrition degrees. Interested students received a participant information sheet, consent form and background questionnaire, followed by remote onboarding (Indeemo account setup, task instructions, and a tutorial video) and a Slack workspace for ongoing communication and researcher contact. In total, 10 students consented, 1 withdrew, leaving 9 completers; participants received a NZD 100 supermarket voucher. Because recruitment relied on self-selection and digital responsiveness, the New Zealand sample is likely biased toward students who are digitally comfortable, time-available, and willing to be observed. These traits may correlate with higher food autonomy or confidence in sharing meals publicly and may therefore shape both participation and reporting behaviors. In comparative interpretation we therefore treat New Zealand findings as reflective of engaged, technology-willing students rather than the entire campus population, and we interpret cross-cohort contrasts with caution given these selection and engagement differences.

China cohort recruitment used convenience sampling via posters placed at entry points to the three main university canteens. Recruitment quotas were apportioned across canteens and posters remained in place until the target number of expressions of interest was reached (36 expressions overall), from which 10 participants were randomly selected to maximize perceived fairness. Eligibility mirrored the NZ cohort: full-time students, ≥18 years, digital device literate; nutrition majors and those with eating disorders or metabolic disease were excluded (41). Participants received two structured remote training sessions via Tencent Meeting with ethnographer modeling of responses to normalize recording and were supported with WeChat group communication for the entire duration of the ethnography. Incentives were CNY 1,200 paid upon satisfactory completion. The on-site poster strategy and later random selection reduced some self-selection bias but likely attracted students who habitually dine on campus and those available during recruitment windows. The relatively larger end-payment may have influenced adherence and reporting intensity differently than the New Zealand voucher. Furthermore, mandatory boarding and subsidized, university-operated canteen systems mean that China participants' day-to-day food exposure is structurally distinct from the New Zealand cohort, which both shaped observed behaviors and reduces direct comparability. We therefore frame cross-cohort contrasts as context-sensitive: differences may arise from genuine cultural and institutional factors but also from recruitment modalities, incentive structures, and participant selection processes.

Data were analyzed using a hybrid inductive–abductive approach that drew on grounded-theory principles and the six-phase thematic procedures described by Terry et al. (42). In practice, inductive codes were generated from participants' transcripts and media (videos, photos, captions), and these initial codes were iteratively compared and, where appropriate, mapped to sensitizing concepts from corresponding theoretical frameworks in each cohort. Deductive codes from the frameworks were therefore used as helpful analytic guides, while inductive codes were retained whenever participant material did not fit existing constructs or represented novel meanings. The primary ethnographer completed initial coding for each cohort; coding lists were discussed in regular meetings with a supervising researcher, and unresolved disagreements were adjudicated by a second senior investigator. Although the two cohorts used different theoretical starting points, both ethnographers adhered to the same qualitative content analysis procedures, maintained harmonized coding practices. We monitored conceptual saturation by coding meal entries in chronological order and tracking cumulative unique codes: saturation was pragmatically defined as fewer than 1 new code per 10 successive meal entries combined with the absence of new substantive themes across two consecutive weeks. Using this criterion, initial coding reached saturation approximately by day 13 (range 11–17) for on-campus diets for both cohorts, and day 15 (range 11–19) for at-home diets in China cohort. Nevertheless, all data were coded in full to allow exploratory theme development. Automated Indeemo transcripts were manually verified where ethnographers reviewed every original video and audio entry, corrected transcription errors, and logged corrections in NVivo to ensure data fidelity prior to analysis.

Ethnographers experience significant barrier to draw valid conclusions when participants divert from natural behaviors, known as the Hawthorne effect (43, 44). In this study, we implemented mitigation strategies including training, prolonged data collection, and participant-as-observer posting, which may have reduced reactivity over time. Nevertheless, some Hawthorne-type reactivity remains possible. Participants were encouraged to respond with a video recording of their eating occasion instead of simple text. Two training sessions were provided by the China cohort ethnographer to familiarize participants with study expectations and Indeemo tasks. Both cohorts collected data over a long period of time, allowing rapport and trust to be established between the ethnographer and the participants (45, 46).

Among all practices, perhaps those minimizing the participants' feeling of being observed were the most effective (47–49). For example, the China cohort ethnographer created a group chat via a mobile social networking application, WeChat (Tencent, Shenzhen, China), where all participants had access to. The ethnographer published every eating occasion as his own responses, every day for 90 consecutive days, exactly as the participants were expected to do. When the participants became the “observer”, their responses exhibited more natural settings.

Participant adherence during the digital diet ethnography was adequate and provided satisfactory video, graphic, and text data. A total of nine students, all females aged between 19 and 29 years (mean age 22 years), completed the New Zealand cohort digital diet ethnography. A total of 10 participants (six females) with an average age between 20 and 23 years (mean age 21 years) completed the China cohort. This section summarized participant behaviors during the digital ethnography and the results of qualitative content analysis. Findings from the New Zealand cohort were presented first, highlighting the four themes and how their sub-themes related to university student dietary behaviors. Then, three major themes and associating codes from the China cohort were presented, including on-campus and at-home datasets. A comparative synthesis between the two cohorts was then presented. Remarkable extracts from participant responses were included where appropriate.

Nine University of Auckland students participated in the New Zealand cohort of this digital diet ethnography (all females; age range 19–29 years), producing a three-month dataset, describing everyday food practices. Thematic analysis of the New Zealand arm identified four principal themes: time is of the essence, cash is king, the on-off campus conundrum, and miscellaneous influences. Each principal theme contained several interrelated subthemes that capture how students made food choices within their academic and social lives, as shown in Table 1.

A total of 10 participants (six females) with an average age of 22 years (range 21–23 years) completed the China cohort, producing a three-month dataset, describing everyday food practices. Thematic analysis revealed three major themes that structured students' dietary behaviors: social and environmental determinants, food-related determinants, and intrapersonal determinants, as shown in Table 2. These themes manifested differently across on-campus and at-home settings. Below we described each domain and its major subthemes, using illustrative extracts where appropriate.

Across both cohorts, three broad patterns consistently influenced student food choices: structural constraints (time and food availability), economic pressures, and moment-to-moment sensory and physiological drivers. In New Zealand, these pressures most often translated into trade-offs between convenience and quality. Students with limited time tended to choose privately sold, ready-to-eat options on campus and only traveled off campus when schedules and budgets allowed. Cost considerations operated as a constant cost-benefit calculus, prompting sharing strategies, discount-seeking and substitution toward cheaper, energy-dense items. In China, the same categories of determinants were evident, but their practical expression followed different institutional logics. Campus canteens and mandatory dormitory living concentrated social influence and shaped routine meal patterns, while parents exercised primary control over food at home. In both settings, sensory appeal, hunger, and comfort repeatedly overrode abstract nutrition knowledge, but these immediate drivers played out against distinct backdrops of vendor models, meal provisioning and daily schedules.

Despite these shared determinants, the mechanisms generating them diverged in ways that have clear cultural and institutional roots. New Zealand students generally reported greater individual autonomy and flexibility. They negotiated between on-campus convenience and off-campus options depending on time and money whereas Chinese students experienced more constrained autonomy arising from collective living arrangements, government-subsidized canteen structures, and parental decision-making at home. Consequently, interpersonal influence was a central force in both cohorts but with different primary actors: peers/roommates in China on campus; parents in China at home; more individualized peer influence and autonomous decision-making in New Zealand. These contrasts suggest that although the categories of determinants are similar across contexts, the pathways by which they shape behavior are context dependent.

Taken together, these convergences and divergences form the empirical basis for the comparative interpretation that follows. The Discussion therefore examines how cultural norms, institutional food systems and economic constraints interact to produce the observed patterns of diet autonomy and choice in each setting, and it considers implications for university food policy and interventions that are sensitive to these contextual differences.

This digital diet ethnography provides the first in-depth qualitative comparison of dietary behavior determinants among university students in two markedly different food-environment contexts: a decentralized, commercialized New Zealand campus and a centralized, government-subsidized Chinese medical university. Despite structural differences, both cohorts revealed similar core dietary behavior determinants: time scarcity, financial constraint, convenience-seeking, taste/familiarity preferences, and social influence. These drivers often operate as powerful barriers to healthy eating while interacting dynamically with the respective food environments to produce context-specific patterns (50–52). New Zealand students exhibited high food autonomy but frequent reliance on cheap, convenient, energy-dense options, while Chinese students displayed lower day-to-day autonomy yet greater exposure to interpersonal and cultural norms that shaped choices. The richness of the 90-day dataset (approximately 2,700 eating occasions per cohort) and the ethnographic depth enabled identification of potentially modifiable factors, including availability, affordability, social norms, convenience infrastructure, and sensory appeal. Consistency of such factors across cultures reinforces the utility of ecological models for understanding university student dietary behavior (38, 51). The findings extend previous qualitative work in Western (50, 53, 54) and Asian settings (55, 56) by demonstrating how macro-level food-system differences (profit-driven vs. welfare-oriented) translate into micro-level lived experiences among university students (23).

Time emerged as the dominant constraint in both cohorts. New Zealand participants repeatedly framed food choice as a trade-off against academic workloads, with convenience trumping health during lectures, assignments, and exam periods. Chinese students similarly prioritized speed and proximity but within the structural reality of fixed canteen hours, long queues, and rigid class schedules. In this setting, meal-skipping or instant noodles were common coping strategies (57). The key difference lay in recovery opportunities: New Zealand students could access 24/7 off-campus options, whereas Chinese students on campus were captive to operating hours and had fewer alternatives when canteens closed. When Chinese participants returned home for summer break, time pressure eased and dietary variety increased markedly. Such environmental shift highlights how rigidly structured campus routines, rather than inherent student laziness, could drive convenience-seeking behavior (23, 52).

Cost was described as “cash is king” in New Zealand and an ever-present consideration in China despite government subsidies. High food autonomy among New Zealand students allowed sophisticated cost-minimization strategies, yet healthy foods were repeatedly labeled “too expensive,” leading to deliberate trade-offs (53, 54). Chinese students benefited from canteen prices substantially lower than off-campus equivalents yet still perceived healthy options as relatively costly within their limited budgets, suggesting that subsidies reduced absolute prices but did not eliminate relative cost barriers (29). Both cohorts therefore converged on energy-dense, nutrient-poor choices as rational economic responses to perceived expensiveness (58), confirming that affordability remains a critical determinant even in subsidized systems when healthier items carry premium pricing or preparation time costs.

Social factors operated differently yet with comparable strength. New Zealand participants were heavily influenced by flatmates, friends, and partners, but decisions remained ultimately individual. Chinese students exhibited stronger collectivistic patterning: roommates synchronized meals, shared food purchases, and exerted normative pressure. Furthermore, family control was near-absolute when students returned home. This reflects broader cultural orientations, i.e. individualism in New Zealand vs. collectivism in China. However, the behavioral outcome was similar: social context frequently overrode personal health intentions (59). The China cohort provide particularly strong evidence of social conditioning as a modifiable determinant, with peer purchasing behavior and family food parenting practices exerting powerful effects that persisted beyond the campus environment (60).

Both cohorts prioritized taste, texture, and familiarity over nutrition messaging. New Zealand students sought “comfort” and “reward” foods during stress, often articulating guilt but prioritizing immediate sensory pleasure. Chinese participants displayed even stronger attachment to familiar flavors and regional dishes, with nostalgia and childhood memories frequently cited. Exposure to new foods was limited in both settings by risk aversion and habitual choice, but the Chinese cohort showed greater willingness to try novel items when socially facilitated or during holidays. These findings support Köster's psychological perspective on food choice diversity (61) and demonstrate that sensory and emotional functions of food consistently outweigh cognitive health knowledge across cultures (62).

New Zealand students possessed high behavioral control yet frequently chose not to exercise it in favor of health, citing fatigue, stress, and competing priorities. Chinese students on campus had lower control due to fixed menus and limited cooking facilities but showed surprisingly high acceptance of this constraint, framing it as normal rather than problematic. When autonomy increased at home, healthier patterns emerged for some but not all, highlighting that autonomy alone is insufficient without supportive environmental conditions (63). Nutrition knowledge was present but rarely actioned in either cohort, which was consistent with global evidence that knowledge-behavior gaps persist in young adults (64–66).

The most profound differences arose from institutional structure rather than culture per se. New Zealand's commercialized, fragmented foodscape produced high variety but pervasive unhealthy default options and price signaling that penalized health (10, 23). China's centralized, subsidized system delivered superior affordability and reduced food insecurity but at the cost of variety, flexibility, and nutritional quality, creating a different but equally potent set of constraints (29). Both systems need improvement to make healthy choices the easy, or default, choice, confirming that individual-level interventions alone are inadequate when structural determinants dominate decision-making contexts (67, 68).

Overall, the cross-cohort comparison demonstrates that university students' dietary patterns are primarily shaped by pragmatic adaptations to common contextual constraints rather than individual shortcomings. These constraints exert remarkably consistent influences on eating behavior despite different institutional and structural approaches to food provision in the New Zealand and Chinese university settings. Interventions must therefore target the shared modifiable determinants, including time-compatible healthy options, genuine affordability of nutrient-dense foods, social norm change, and sensory appeal, while exploiting each food environment system's strengths to shift “default” dietary choices toward healthier options (69, 70).

These findings have direct implications for university food policy and student health promotion in both New Zealand and China. Universities could strengthen campus food environments by prioritizing the accessibility, affordability, and appeal of nutritious options through procurement policies aligned with public health nutrition guidelines and sustainability principles, thereby reducing student reliance on low-cost, energy-dense foods (87). New Zealand institutions would benefit from stronger policy coordination and regulatory oversight to ensure vendors offer nutritious, culturally responsive choices (88). Chinese universities can implement system-level reforms to enhance menu diversity and nutritional quality, especially with centralized food-service systems (89). Complementary interventions should foster greater student food autonomy: in New Zealand through improved nutrition education, cost-effective meal-preparation support, and food-insecurity alleviation; in China via family engagement, cooking-skill development, and culturally tailored guidance. Finally, both countries should integrate environmental sustainability into food pricing, messaging, and offerings while ensuring such initiatives remain economically accessible to students.

The findings of this ethnography offer methodological implications for future qualitative and quantitative studies of university student dietary behavior. To minimize bias and enhance data quality, future digital or face-to-face diet ethnographies should prioritize prolonged pre-recruitment engagement to build rapport and incorporate barrier-free data collection (90). Researcher training in task design, culturally sensitive communication, and empathy remains essential to mitigate observer effects in context-specific settings (91). Demographics, socioeconomic status, and hometown should be routinely collected as predictors for subsequent quantitative analyses (53). Digital technologies, including mobile ethnography apps and wearables, substantially reduce the observer effect while enabling simultaneous capture of quantifiable intake metrics and rich qualitative rationales, despite ongoing ethical discussions (45, 92–94). Potential next quantitative steps are therefore to (1) assess diet quality and health status using validated instruments such as Healthy Eating Index, (2) implement or leverage digitized campus food-service systems for precise intake recording (95), and (3) conduct controlled intervention trials evaluating food environment modifications on student dietary outcomes.