The paradoxes of generative AI-mediated acculturation

The ambivalence in the vignette above mirrors concerns raised in a New York Times letter to the editor (Chae, 2025) in response to Apple announcing a new generative-AI (GenAI) enabled live translation feature in AirPods. The letter writer, a professional interpreter, praised the progress of machine translation for simple text yet warned that real-time interpretation risks flattening nuance and discouraging the learning of other languages. If everyday interactions are globally “dubbed,” the incentive for curiosity about other cultures may decline, creating a sense that “everyone sounds the same but no one truly understands.” Such tensions between surface fluency and deeper cultural connection foreshadow the broader challenges of GenAI-mediated cross-cultural engagement explored in this paper. This example illustrates a larger transformation: algorithmic systems no longer merely filter and rank cultural content; they now generate it. Their affordances, such as dialogue, synthesis and role-play, are increasingly mediating how consumers learn, rehearse and perform culture (Leonardi, 2011; Kozinets et al., 2021). Large language models (LLMs) now sit inside search engines, productivity platforms and social media, influencing how people ask questions, make plans and interpret unfamiliar contexts (Davenport et al., 2020).

For international marketing, this shift has potentially subtle yet profound effects. Consumers who once encountered culture primarily through direct interactions with people, media or markets (Peñaloza, 1994, 1995; Askegaard et al., 2005; Thompson and Tambyah, 1999) now increasingly do so through GenAI outputs that do not merely transmit cultural material but generate, simulate and personalize it in real time. Translation, itinerary suggestions, etiquette explanations and even “local” product recommendations are increasingly delivered by systems that embed the biases, defaults and gaps of their training data (Lu et al., 2025; Dev and Qadri, 2024; González-Cantera and Bonacchi, 2025; Spennemann, 2023, 2024a, b). We argue that GenAI is becoming a cultural intermediary in international markets, reshaping how consumers acquire knowledge, signal belonging and construct identity across borders.

Importantly, our argument is not that mediated acculturation is new. Prior research shows that media and social platforms can shape acculturation by providing cultural information, identity resources and technologically mediated ties that support adaptation without requiring continuous face-to-face contact. For example, social media has been described as a vital means of consumer acculturation because it provides information about culture while helping people maintain relationships and share experiences during adaptation (Kizgin et al., 2018). However, GenAI changes the nature of mediation. Rather than primarily distributing human-produced cultural content or enabling contact within human networks, GenAI generates cultural scripts, explanations and simulations in response to prompts and can refine these outputs across turns using conversational context. This moves mediated acculturation from exposure and social learning to model-shaped co-production and rehearsal, where “first contact” with cultural norms can be synthetic, personalized and path dependent. We therefore position GenAI-mediated acculturation as distinct from media-mediated acculturation and theorize it as a new pathway of model-mediated contact that reshapes how cultural learning is acquired, performed and evaluated.

We propose four mechanisms that capture how AI-mediated contact alters how consumers experience culture. Secondhand acculturation arises when translation and synthesis accelerate surface familiarity but miss tacit norms (Berry, 1997, 2005). Algorithmic cosmopolitanism broadens exposure by mixing together examples from many cultures, but in doing so, it often flattens the details and reduces depth (Cleveland and Laroche, 2007; Cleveland et al., 2009; Airoldi and Rokka, 2022). Curated curiosity develops as a result of system prompts guiding exploration, making the learning path dependent (Leonardi, 2011; Kozinets et al., 2021). Finally, the control illusion occurs when conversational cues signal agency even though GenAI guardrails and training may quietly set boundaries (Epstein et al., 2023; Ghosh et al., 2024). Together, these mechanisms broaden the opportunity for discovery but likely compress nuance. In short, these signals allow outsiders to project a sense of cultural closeness, yet in reality, they often highlight just how distant they remain from genuine understanding or authentic connection.

Existing theories only partially explain these changes. Acculturation research focuses on adaptation through direct cultural contact (Berry, 1997, 2005; Peñaloza, 1994, 1995; Askegaard et al., 2005) but does not yet address GenAI-mediated contact or its uneven effects on learning, stereotype persistence and identity formation. Affordance theory could also be extended to include dialogic generation and role-play as new forms of technological invitation and constraint that shape consumer behavior (Leonardi, 2011; Kozinets et al., 2021). Finally, research on cultural capital could incorporate a new type of performable capital: AI-assisted knowledge that spreads easily and is simple to display yet often remains shallow and hard to verify. Such knowledge may signal competence to outsiders but lacks the authenticity needed for genuine in-group participation (Cleveland and Laroche, 2007; Thompson and Tambyah, 1999).

To address these gaps, we make three main contributions. First, we advance international marketing theory by conceptualizing GenAI as a cultural intermediary and extending acculturation theory through the idea of model-mediated contact, which differs from both direct intercultural contact and media-mediated acculturation by making cultural learning increasingly generative, interactive and rehearsal-based. This reframes how consumers learn about and perform culture when their first encounters are shaped by generative systems rather than direct social experience. Second, we integrate insights from affordance theory and global consumer culture research to build a conceptual framework linking GenAI's key affordances to four mechanisms that explain how cultural learning and identity work now unfold. Third, we translate these theoretical insights into managerial guidance, highlighting when AI, human or local mediation is most appropriate and how practices such as provenance, diversity controls and local review can help protect cultural nuance. Together, these contributions lay the groundwork for understanding how GenAI reshapes cultural exposure, identity formation and acculturation pathways.

More broadly, this paper develops a theory-extending conceptual framework rather than a descriptive or purely critical commentary. Our aim is not only to document the presence of GenAI in cross-cultural contexts, but to explain how and why cultural learning, identity work and adaptation unfold differently when mediated by generative systems. Accordingly, the framework is organized around four guiding questions: (1) How does GenAI transform the nature of cultural contact relative to direct and media-mediated forms, (2) through what mechanisms does this transformation reshape acculturation processes, (3) what patterned tensions and trade-offs does this new mode of cultural learning generate and (4) what new theoretical and managerial expectations follow from these changes? The framework that follows is designed to structure these questions and to generate testable directions for future research.

Acculturation theory explains how individuals and groups adapt when encountering a new culture. Berry's influential bidimensional framework distinguishes between assimilation, integration, separation and marginalization, depending on whether people maintain their heritage culture and adopt the host culture (Berry, 1997, 2005). His stress–adaptation–outcome model highlights how intercultural contact generates acculturative stress, which individuals manage through coping strategies that shape adaptation outcomes. A critical insight is that acculturation involves both surface learning, such as mastering language or visible customs and deeper integration of tacit norms, values and identities.

Marketplace contexts add further complexity. Consumers do not simply adapt passively but actively negotiate belonging through networks, practices and performances of identity (Peñaloza, 1994, 1995). This process is mediated by unequal access to resources, legitimacy and representation, meaning that some groups are better positioned to define what counts as authentic participation than others. Later work shows that consumers often sustain hybrid and plural identities, combining elements of host and heritage cultures and shifting their displays depending on context rather than following a linear path of replacement (Askegaard et al., 2005). At the global scale, consumers also engage with transnational symbols and practices, with orientations such as cosmopolitanism, ethnocentrism and materialism shaping whether global signs are embraced as cultural bridges, resisted in favor of local authenticity or adopted as markers of status (Cleveland and Laroche, 2007; Cleveland et al., 2009).

Yet the assumptions underpinning this literature are increasingly strained by generative artificial intelligence. Classic models of acculturation generally presuppose direct intercultural contact through people, places and communities. In contrast, generative systems create a new form of mediated contact, where consumers gain familiarity with a culture not through lived interaction but through outputs generated by algorithms. This shift raises questions about whether existing models, developed for embodied and social encounters, can account for the ways consumers now learn, rehearse and perform cultural knowledge when their first contact is algorithmic.

Recent studies underscore the risks of this shift. Machine translation and generative systems often strip away pragmatic nuance, such as tone or politeness markers, distort heritage cues like local idioms or symbolic references and reproduce homogenized or even erroneous accounts of cultural memory (Spennemann, 2023, 2024a, b). In sensitive contexts, such as contested memoryscapes, large language models have been shown to systematically privilege dominant government perspectives over marginalized voices, raising concerns about both authenticity and pluralism (González-Cantera and Bonacchi, 2025). Representational harms also extend beyond omission: generative models frequently reinforce stereotypes or erase subcultural differences, as biases in training data shape what is presented as “legitimate” cultural knowledge (Ghosh et al., 2024). Even the models themselves reflect cultural tendencies, generating responses that mirror the cognitive styles embedded in each language. For example, they tend to produce more independent and analytic reasoning when prompted in English and more interdependent and holistic reasoning when prompted in Chinese (Lu et al., 2025). Taken together, these findings make clear that AI-mediated encounters are not a neutral channel but can be an active force that reconfigures which cultural voices are amplified and which are silenced.

At the same time, it is important to note that mediation itself is not new to acculturation theory. A growing body of research has shown that cultural learning and adaptation can be shaped by technologically mediated environments, particularly through social and digital media. To clarify what is distinctive about GenAI, we therefore differentiate it from earlier forms of media-mediated acculturation.

Prior work demonstrates that acculturation does not require continuous face-to-face interaction, and that mediated environments can act as powerful acculturation agents. However, most of this work conceptualizes mediation as a channel of exposure, interaction and social learning rather than as a generative partner in cultural sensemaking. Rather, acculturation unfolds through symbolic exchange, mediated representations and culturally structured consumption practices that allow individuals to learn, negotiate and perform belonging across contexts (Peñaloza, 1994, 1995; Thompson and Tambyah, 1999). Research on global consumer culture further shows that cultural learning often occurs through transnational flows of signs, brands and media, enabling consumers to acquire cultural competencies without direct immersion (Cleveland and Laroche, 2007; Cleveland et al., 2009). In this vein, Kizgin et al. (2018) describe social media as a vital means of consumer acculturation because it allows individuals to learn cultural codes, acquire consumer skills and negotiate identity while maintaining connections to both host and heritage cultures. In this view, mediated acculturation primarily operates through exposure to and interaction around, human-produced cultural content embedded in social networks.

Generative AI introduces a qualitatively different form of mediation. Rather than mainly distributing, curating or amplifying human cultural material, GenAI actively generates cultural scripts, explanations and simulations in response to user prompts. Through conversational interaction, it can personalize outputs across turns, adapt tone and register and sustain role-based exchanges that allow consumers to rehearse culturally marked encounters. In this sense, GenAI functions not merely as a channel of cultural transmission, but as a co-productive interlocutor that reshapes what cultural learning looks like in practice (Leonardi, 2011; Kozinets et al., 2021; Davenport et al., 2020).

This shift changes the mechanism of acculturation itself. In media-mediated acculturation, learning occurs primarily through active interpretation, comparison and social feedback as individuals engage with cultural representations and communities (Peñaloza, 1994, 1995; Askegaard et al., 2005). In GenAI-mediated acculturation, learning increasingly occurs through rehearsal, emulation and synthesis, where the model provides ready-made cultural performances that can be adopted without equivalent situated experience. Cultural competence can therefore be displayed before it is fully understood. We refer to this pathway as model-mediated contact and to the learning it enables as secondhand acculturation: a process in which consumers acquire surface fluency and cultural legibility through interaction with generative systems rather than through direct social immersion (Berry, 1997, 2005).

This distinction is crucial for theory. While media-mediated acculturation expands access to cultural resources, GenAI-mediated acculturation reshapes the relationship between exposure, performance and understanding. Because generative systems optimize for coherence, safety and statistical regularity, they tend to smooth over conflict, ambiguity and local specificity (Airoldi and Rokka, 2022; Epstein et al., 2023). This compression of difference can weaken pragmatic competence, distort cultural memory and reproduce dominant representations (Spennemann, 2023; Ghosh et al., 2024; González-Cantera and Bonacchi, 2025). As a result, GenAI can simultaneously broaden cultural reach and erode authenticity, insider recognition and contextual sensitivity.

Consumer agency in markets can be understood as the capacity to perceive options, make choices and construct identities within infrastructures shaped simultaneously by culture and technology. In cross-cultural settings, this agency often takes the form of acquiring scripts and competencies that allow consumers to signal belonging and legitimacy, but it is never unlimited. It is structured by networks, resources and power dynamics (Askegaard et al., 2005; Peñaloza, 1994, 1995).

The rise of generative artificial intelligence alters this balance by shifting cultural learning from algorithmic curation to model-mediated co-production. Unlike earlier systems that merely filtered or ranked information, generative models actively produce cultural scripts, embedding training data biases and governance constraints into their outputs. This development raises an important theoretical gap: much of the literature on consumer agency assumes that technology provides expanded choice sets, yet in practice, generative systems may simultaneously open and close possibilities. The apparent freedom of prompting can mask hidden constraints, suggesting the need to revisit how agency is conceptualized when mediated by machine outputs.

Evidence across domains illustrates these tensions. Generative systems can boost fluency and productivity, but they often do so by encouraging convergence and narrowing cultural diversity (Doshi and Hauser, 2024). In cultural heritage, they strip nuance, distort memory and spread homogenized or even inaccurate accounts of the past (Spennemann, 2023, 2024a, b). Image models also reproduce stereotypes or misappropriate cultural elements, such as portraying Indian subcultures with exaggerated exoticism or misplaced symbols that reduce complexity to clichés (Ghosh et al., 2024). At the aesthetic level, outputs often lean toward polished, hyperreal templates that obscure their artificial origin, erode variation and reinforce sameness (Epstein et al., 2023). Taken together, this evidence suggests that generative systems do not simply transmit culture but reshape it, amplifying some voices while compressing or silencing others.

International marketing has long been shaped by the tension between global standardization and local adaptation. Research on global consumer culture shows how brands, symbols and practices circulate across borders, creating a repertoire of shared signs that cosmopolitan consumers can recognize and perform internationally (Alden et al., 1999; Cleveland and Laroche, 2007). At the same time, work on country-of-origin effects underscores the continuing importance of provenance, authenticity and heritage, with associations to place shaping perceptions of quality, trust and purchase intentions (Samiee et al., 2005).

These perspectives reflect competing logics of meaning. Global positioning strategies often draw on cues such as English-language taglines, multicultural imagery or global lifestyle references to signal modernity and belonging to a borderless culture (Alden et al., 1999), while local positioning emphasizes cultural specificity and heritage to convey authenticity and trust. Consumer acculturation research further shows that global orientation is rarely complete. Competencies gained through media, multinational brands, language learning and travel often expand consumers' engagement with global culture, yet these orientations typically coexist with ethnic identification and local attachments (Cleveland and Laroche, 2007). Studies of country of origin recognition complicate this picture, demonstrating that consumers frequently misidentify brand origins with accuracy rates often close to chance. Country of origin cues, therefore, can function less as factual knowledge and more as heuristic impressions that lend symbolic value to products (Samiee et al., 2005).

Taken together, current theory does not yet explain how machine-mediated interactions reshape the foundations of cultural learning and exchange. Acculturation theory focuses on direct contact but overlooks how generative systems alter the balance between surface familiarity and deeper integration, reshape identity work and redefine authenticity. Research on consumer agency rarely addresses whether people are truly empowered when fluency comes at the cost of depth or when authenticity is traded for efficiency. Similarly, work on global consumer culture also assumes that cultural circulation is human-mediated, missing how GenAI now synthesizes and re-renders cultural signs, blurring the line between global diffusion and local distinctiveness. Extending these theories to account for model-mediated contact is likely needed, and our framework helps develop this extension by examining how GenAI changes the trade-offs between breadth and depth, global prototypes and local meaning.

Taken together, the three literatures reviewed above point to a common gap. Acculturation theory clarifies how individuals acquire surface fluency, negotiate belonging and integrate tacit norms through interaction. Affordance theory explains how technologies structure action by inviting some practices while constraining others. Research on global consumer culture highlights the persistent tension between cultural breadth and depth, global legibility and local meaning. However, none of these streams, on their own, explains how generative systems reorganize these processes when cultural learning becomes interactive, synthetic and model-mediated.

Our framework is therefore developed through an abductive synthesis of these literatures. We identify four GenAI affordances that matter most for cultural learning because they directly intervene in how consumers encounter, rehearse and perform culture: translation and localization reshape linguistic access, synthesis reshapes how cultural material is aggregated and made legible, role-play and simulation reshape learning through rehearsal and conversational memory reshapes continuity and personalization. These affordances are not exhaustive, but they are the ones that most directly alter the mechanisms of acculturation identified in prior theory.

From these affordances, we derive four mechanisms that explain how cultural learning unfolds under GenAI mediation: secondhand acculturation, algorithmic cosmopolitanism, curated curiosity and the control illusion. These mechanisms capture recurring patterns in how GenAI expands access while simultaneously compressing nuance, reshaping how competence is acquired, how difference is encountered and how agency is perceived. Finally, we articulate the consequences of these mechanisms as paradoxes rather than linear effects because prior work in consumer culture and acculturation consistently shows that cultural adaptation is characterized by tension and trade-offs rather than unidirectional change.

This logic structures our framework. Affordances specify what GenAI makes possible, mechanisms explain how cultural learning is reorganized through those possibilities and paradoxes capture the patterned tensions that follow. In doing so, the framework connects GenAI's technical properties to established theories of cultural adaptation and global consumer culture rather than treating them as context-specific anomalies.

The conceptual framework that we develop (and which is summarized in Figure 1) explains how GenAI functions as a cultural intermediary that reshapes cross-cultural learning through model-mediated contact and is intended as a theory-extending device that specifies new mechanisms, boundary conditions and research directions for acculturation and international marketing scholarship. It begins with the four culture-related affordances of translation and localization, synthesis, role-play and simulation and conversational memory, which together define how generative systems invite and constrain interaction with culture and distinguish model-mediated contact from prior forms of media-mediated cultural exposure and platform-mediated social learning. These affordances give rise to four mechanisms of AI-mediated acculturation: secondhand acculturation, algorithmic cosmopolitanism, curated curiosity and the control illusion. Together, these mechanisms reveal how GenAI broadens cultural access while compressing nuance, leading to four paradoxes: Understanding–Authenticity, Connection–Resonance, Choice–Creativity and Democratization–Dominance.

An affordance refers to what a technology makes possible for users to do. It is not simply a feature or function but an invitation for action that emerges from the interaction between a person and a system in a specific context (Gibson, 1979; Norman, 2013). In digital environments, affordances help explain why technologies shape behavior differently across users and situations. They depend on both design, what the system can technically do, and perception, what the user thinks it can do. As people explore and learn, new affordances appear or fade (Leonardi, 2011; Markus and Silver, 2008).

Generative AI introduces a new layer of affordances because its outputs are not static or pre-coded. They are co-produced in dialogue, with each prompt and response creating fresh possibilities. This makes GenAI different from earlier information systems that retrieved or ranked content. Its generative nature allows users to ask questions, negotiate tone and even co-create cultural material in real time. In this sense, GenAI functions less as a search engine and more as a cultural partner. Four affordances capture this shift: translation and localization, synthesis, role-play and simulation and conversational memory (Kozinets et al., 2021). Together, they define how users now experience and learn culture through model-mediated contact.

Translation and localization involve turning speech, text or images from one language or script into another within an interactive exchange. What is new is that generative systems do not simply substitute words. They attempt to interpret meaning, tone and context. The model can switch registers between formal and casual, adjust politeness and emulate idioms or humor when prompted. In practical terms, this means language conversion has evolved from literal translation to a kind of real-time cultural mediation.

GenAI makes this affordance possible through the integration of neural translation, large-scale pretraining and in-context learning. The model can infer meaning across turns, maintain role continuity and refine tone mid-conversation. It does so through statistical correlations learned from immense multilingual corpora. While this gives the impression of cultural intelligence, the fluency often masks hidden biases and limits. Guardrails and defaults still guide what the model can say, which means that even while it enables global understanding, it subtly shapes the boundaries of what can be expressed (Epstein et al., 2023).

Synthesis refers to the way generative systems condense large, varied information into cohesive overviews. Unlike traditional search or summarization, which retrieve or shorten fixed documents, synthesis produces dynamic narratives that respond to user intent. The same model can shift stance by offering a global overview one moment and a local counterpoint the next, based entirely on how the user frames a question. This conversational malleability marks a genuine break from earlier technologies that simply presented lists of sources. It allows GenAI to act as an on-demand cultural interpreter, capable of blending perspectives across geography, time and genre.

These tendencies are not accidental but structural. GenAI enables synthesis through pretraining on vast multilingual corpora combined with instruction tuning and retrieval-augmented generation. The model can blend information across domains and produce apparently balanced overviews at high speed. Yet its responses are guided by patterns in those data and by the safety rules built into its architecture. Defaults and reward models push it toward the familiar middle of the distribution, what feels reasonable or neutral, rather than the contested edges where cultural nuance often lives (Epstein et al., 2023). As a result, GenAI's capacity to synthesize gives users remarkable fluency, but that fluency is shaped by hidden boundaries that quietly define what kinds of knowledge, tone and disagreement are allowed to appear.

Role-play and simulation refer to GenAI's ability to stage interactive, culturally marked encounters through dialogue. The model can take on roles, such as a shopkeeper, immigration officer or restaurant host and sustain them across a conversation. What distinguishes this from earlier computer-based training or chatbots is adaptability. The AI listens to user input, stays in character and modifies tone or complexity as the interaction evolves. This iterative process allows the exchange to feel improvisational and alive, more like a guided rehearsal than a scripted exercise.

GenAI enables role-play and simulation through controllable generation, conversational memory and persona libraries that encode setting, stakes and constraints. These features allow the model to sustain a role, track progress and deliver feedback that feels personal. However, the same mechanisms that keep simulations coherent also limit what they can express. Governance layers, moderation filters and the biases embedded in training data silently define the edges of the cultural world that users can explore. In effect, GenAI provides a stage for cultural rehearsal, but the script and scenery are partly pre-written by unseen hands (Leonardi, 2011).

Conversational memory refers to a system's capacity to remember previous exchanges, preferences and goals across turns or sessions. Rather than starting each interaction from scratch, GenAI tools can recall earlier prompts, note mistakes or corrections and build on prior discussions. This continuity allows them to construct a kind of private learning path or individualized curriculum that adjusts as the conversation unfolds. The defining novelty here is the persistence of context: users can return days later and find the system still “aware” of past choices, tone preferences or unfinished ideas.

Technically, this affordance is made possible through embeddings that store conversation context, preference vectors that track user tendencies and session-level memory that feeds back into generation. Integration with productivity tools and search platforms brings these capabilities into everyday workflows, where they quietly shape learning and decision-making (Davenport et al., 2020). As GenAI systems grow more embedded in organizational and personal life, conversational memory becomes both a convenience and a constraint – an invisible force that makes interaction smoother while quietly scripting what users are likely to see next.

The affordances of GenAI described above combine to produce four mechanisms that explain how cultural learning now unfolds through model-mediated contact. Each mechanism captures a distinct way in which GenAI reshapes the process of acculturation: how consumers gain cultural familiarity, interpret diversity, explore knowledge and perceive control.

Secondhand acculturation describes how GenAI allows people to learn about cultures indirectly, through generated content rather than through lived or social experience. This mechanism is driven by translation and synthesis affordances, which make cultural knowledge easier to access while also filtering and simplifying it. When users rely on GenAI for learning, they often encounter a version of culture that has already been curated and standardized through the model's training data. The system delivers polished translations, etiquette summaries and concise cultural overviews that convey surface-level understanding but omit the deeper, tacit cues that come from interpersonal contact or situated practice.

In Berry's (1997, 2005) terms, GenAI likely shifts the learning process toward surface acculturation. Instead of internalizing subtle patterns through immersion, users acquire visible competence in language or customs through model-mediated exposure. This process compresses the experience of cultural learning into textual or visual representations that can be rehearsed and repeated. Research on machine translation illustrates this pattern: pragmatic cues such as honorifics, humor or social hierarchy often disappear in translation, producing text that is grammatically correct yet socially thin (Spennemann, 2023, 2024a, b). When the same dynamic occurs through generative synthesis, diverse voices are blended into coherent but culturally neutral summaries that downplay variation and debate (Dev and Qadri, 2024).

This form of learning aligns with Peñaloza's (1994, 1995) view of consumer acculturation as performance. People use these model-mediated scripts to act appropriately within a culture, often without understanding the underlying social logic. GenAI reinforces that pattern by producing outputs that feel authoritative, encouraging users to rely on generated knowledge as a proxy for lived experience. Over time, this mechanism fosters a kind of secondhand cultural literacy, a mediated familiarity that makes culture easier to access and display but also shapes how users define and interpret cultural knowledge.

Algorithmic cosmopolitanism describes how GenAI systems expose users to a global pool of cultural material that is blended, rephrased and normalized into easily digestible forms. By drawing from diverse sources across languages, regions and media, these systems likely create what appears to be a seamless global culture. Consumers encounter fragments of cuisine, fashion, humor or etiquette from multiple contexts and integrate them into their personal knowledge base. The result is a form of technologically mediated worldliness: users can reference or perform elements from many cultures without necessarily engaging deeply with any of them.

This process builds on earlier research describing cosmopolitan consumers as individuals who value openness and global fluency (Cleveland and Laroche, 2007; Cleveland et al., 2009). In traditional settings, cosmopolitanism develops through travel, cross-cultural friendships and sustained contact that shape empathy and understanding. In contrast, algorithmic cosmopolitanism arises from data aggregation. GenAI delivers exposure without experience. By synthesizing cultural inputs from millions of sources, it provides users with a curated impression of global variety while concealing the unevenness and context-specific meaning that such variety depends on.

The mechanism also echoes Holt's (2002) notion of cultural branding, where global symbols circulate widely but lose distinctiveness as they detach from their origins. When generative systems remix these symbols, they often create templates that feel universally familiar yet culturally nonspecific. Airoldi and Rokka (2022) describe this as the flattening of difference in algorithmic mediation: the model optimizes for coherence and appeal, not authenticity. As a result, consumers who use GenAI to learn about or represent culture can appear cosmopolitan to others in their own environment, but insiders may perceive their expressions as generic or mismatched.

In practice, algorithmic cosmopolitanism functions as both a gateway and a filter. It broadens exposure to global content but also shapes what “global” comes to mean. The balance between breadth and depth depends on the user's orientation and purpose. Those with strong cosmopolitan values may find convergence acceptable because it signals openness and competence, while those with locally grounded identities may be more attuned to what is missing. In both cases, GenAI acts as a cultural mediator that expands reach but subtly standardizes the expression of worldliness across markets and audiences.

Curated curiosity refers to how GenAI systems shape the direction and depth of users' exploration through the prompts, suggestions and autocompletions that guide interaction. Unlike open-ended inquiry, where learning emerges from spontaneous curiosity, these systems subtly script discovery. When a user begins a query, GenAI often proposes related questions or continuation phrases that reflect patterns in its training data and safety constraints. This design keeps users engaged but channels their curiosity along predictable, well-traveled paths (Leonardi, 2011).

In practice, prompting structures and autocomplete cues become powerful mediators of cultural learning. Users are encouraged to ask about topics that are common, non-controversial or easy to represent – such as food, fashion or festivals – while more complex or sensitive subjects, such as religion, politics or identity, are rarely suggested. Kozinets et al. (2021) describe this as the algorithmic conditioning of discovery: technological affordances expand access to information while simultaneously narrowing its range. What appears as free exploration is often a guided experience, where users encounter the parts of a culture that the system deems suitable or engaging.

This process creates path dependence in cultural learning. The more a user interacts with the system on specific topics, the more the model reinforces those areas in future prompts and responses. Over time, this selective reinforcement builds a sense of familiarity that may feel comprehensive but is in fact partial. Studies of algorithmic curation show similar effects in news and social media, where users gain confidence in their understanding while the diversity of content they see declines (Epstein et al., 2023). The same dynamic extends to cultural learning, especially in low-resource linguistic and cultural settings where training data are sparse (Lu et al., 2025). There, the boundaries of curiosity are drawn even tighter, as the model's knowledge base privileges high-volume, globally dominant sources. Curated curiosity is therefore a mechanism of subtle control rather than overt instruction. The system's affordances encourage users to ask questions, but the options provided and the patterns reinforced, shape what becomes thinkable or worth knowing.

The control illusion describes the gap between users' perceived and actual agency in GenAI-mediated interactions. While curated curiosity concerns how algorithms steer the direction of exploration, the control illusion centers on how systems create the feeling of autonomy within those constraints. Conversational interfaces invite users to believe they are leading the exchange. The prompt-response rhythm feels interactive and personal, reinforcing the sense that the system is following their intent. In reality, the conversation unfolds inside a tightly bounded space defined by training data, moderation filters and probabilistic generation rules (Kozinets et al., 2021; Ghosh et al., 2024). The interface gives the appearance of co-creation, but the scope of permissible inquiry is predetermined.

The illusion works because generative systems are designed to appear cooperative and adaptive. Each response seems tailored to the user's wording, tone and goals. Follow-up prompts give the sense of iterative control, where users believe they are refining and steering the conversation. Yet every reply is shaped by layers of moderation filters and probability-based generation patterns that keep the model within approved limits. This structure creates a form of bounded interactivity: users perceive freedom of inquiry, but the system quietly excludes controversial, risky or non-conforming material.

In cross-cultural settings, this perceived control carries particular significance. Consumers may trust GenAI tools to guide travel plans, product choices or etiquette decisions because the dialogue feels personalized and responsive. The tone of collaboration fosters confidence and reliance on AI-mediated advice, even when the underlying cultural understanding remains superficial. As users grow accustomed to this fluency, they may not recognize how their questions and interpretations are subtly steered toward safe and standardized representations of other cultures (Spennemann, 2023; González-Cantera and Bonacchi, 2025).

The control illusion is not simply a technical limitation but a socio-cultural one. By equating conversational flexibility with true agency, users mistake accessibility for mastery. This misplaced confidence can lead to unexamined trust in AI guidance and overestimation of one's ability to navigate complex cultural settings. What feels like co-creation is often a carefully bounded exchange where the system dictates the edges of curiosity and knowledge. The challenge for international marketing and cross-cultural communication is to recognize that the sense of control provided by GenAI is real in form but limited in substance.

Rather than treating the outcomes of GenAI-mediated acculturation as a set of isolated effects, we theorize them as paradoxes because cultural learning is inherently characterized by tension, ambivalence and trade-offs rather than linear gains. The four paradoxes are not intended as a typology of loosely related consequences, but as patterned tensions that systematically emerge from the mechanisms described above. Each mechanism reorganizes a different dimension of acculturation and each paradox captures the core trade-off that follows. The Understanding–Authenticity paradox reflects an epistemic tension between surface fluency and deeper cultural competence. The Connection–Resonance paradox captures a relational tension between simulated familiarity and emotionally grounded belonging. The Choice–Creativity paradox represents a creative tension between personalized abundance and convergent expression. The Democratization–Dominance paradox reflects a structural tension between expanded participation and the reproduction of cultural power.

While all four paradoxes share a common root in the compression of cultural depth that accompanies GenAI-mediated learning, they operate on distinct dimensions of acculturation and therefore generate different theoretical and managerial implications. This structure allows the framework to move beyond description and to specify how, why and where GenAI reshapes cultural learning in patterned ways. Against this backdrop, GenAI's role as a cultural intermediary produces a series of paradoxical outcomes for both consumers and marketers. The same affordances that make cultural exchange more accessible also create new forms of limitation, blending empowerment with erosion. To explain these tensions, we outline four paradoxes that illustrate how GenAI reshapes cultural learning and marketing practice.

GenAI offers both consumers and marketers an unprecedented capacity to access, synthesize and reproduce cultural knowledge. A few prompts can generate instant explanations of customs, idioms or local norms, while analytical dashboards can visualize sentiment trends across millions of posts. The result is a new kind of cultural fluency that sounds informed, inclusive and globally aware. Yet this apparent understanding often proves thin. The very processes that make GenAI so efficient, including summarization, translation and pattern recognition, also smooth away the ambiguity, conflict and discomfort that give culture its depth (Spennemann, 2023, 2024a, b; Dev and Qadri, 2024). What emerges is comprehension without context and clarity without complexity. Both consumers and marketers come to see culture clearly but feel it faintly.

For consumers, this manifests as secondhand acculturation. Models condense diverse traditions into balanced, easily digestible narratives that reward curiosity with polished, non-controversial answers. Users learn what gestures mean but not why they matter. They mimic tone and phrasing but lack the social intuition to adapt in real interaction (Berry, 1997, 2005; Kozinets et al., 2021). GenAI provides procedural fluency or how to “sound right,” but rarely develops pragmatic competence, the ability to sense hierarchy, humor or emotion in context (Spennemann, 2023; Leonardi, 2011). These systems excel at producing politeness and coherence, yet they fail to convey the tacit cues and improvisation that authentic communication depends on. The result is what Airoldi and Rokka (2022) describe as synthetic cosmopolitanism, a polished worldliness that signals openness but lacks experiential grounding.

For marketers, the same synthesis affordance drives a parallel illusion of insight. Generative models aggregate weak signals into coherent themes, providing dashboards that seem to render culture legible and predictable. However, by optimizing for consensus and stability, these systems mute the deviance, dissent and ambiguity that often herald cultural change (Airoldi and Rokka, 2022; Askegaard et al., 2005; Epstein et al., 2023; Peñaloza, 1994). The output feels authoritative precisely because it filters out contradictions. Marketers risk mistaking representativeness for relevance, believing they understand culture when they have only mapped its median.

GenAI creates the impression of unprecedented connection. Consumers can chat with a model role-playing as a Parisian barista or a Kyoto tour guide, while marketers can deploy campaigns that seem fluent across dozens of languages and cultural contexts. These interactions and messages feel personal, responsive and attuned to local nuance. Yet beneath the surface of this global fluency lies a deeper detachment. The same technologies that simulate familiarity and emotional warmth often strip away the social texture and vulnerability that make real relationships meaningful. The result is communication that sounds right but rarely resonates.

For consumers, GenAI's conversational capabilities provide companionship without reciprocity. Models remember details, mirror tone and sustain dialogue in ways that mimic human empathy. However, this exchange unfolds entirely within a private feedback loop, with no mutual risk or shared accountability (Leonardi, 2011). Users feel seen, but no one truly sees them. Through such secondhand acculturation, consumers gain linguistic and cultural scripts that give the appearance of inclusion, yet these scripts circulate without human anchors (Berry, 1997, 2005). The sense of belonging that emerges is fragile because it lacks the emotional resonance and social repair that genuine relationships require. Over time, this efficiency of simulated connection may erode the motivation to pursue authentic intercultural engagement, reducing global communication to a set of parallel monologues (Cleveland and Laroche, 2007).

For marketers, GenAI extends communication across borders but compresses the cultural specificity that sustains authenticity. Automated translation, tone adjustment, and style transfer allow campaigns to appear instantly in dozens of languages, each sounding natural to its target audience. However, the same fluency that expands reach can hollow expression. When models prioritize smoothness and clarity, they remove the micro-cues (regional phrasing, humor and rhythm) that convey lived culture (Spennemann, 2023, 2024a, b). Role-play and conversational training help systems emulate local tone, yet they do so through generalized data patterns rather than situated experience (Epstein et al., 2023). This produces messages that are culturally legible but emotionally thin. Brands risk creating what Holt’s (2002) called “hollow globalism,” content that travels widely but connects weakly.

GenAI creates the illusion of infinite possibility. Consumers can prompt for personalized recommendations, itineraries or creative inspiration that feel tailored to their unique preferences. Marketers can generate campaigns that adapt seamlessly across formats, languages and audiences. The experience on both sides is one of coherence and control: users receive precisely what they ask for and brands deliver messages that sound consistent everywhere. Yet this optimization comes at a cost. The same systems that deliver relevance and reliability also erode novelty and diversity. What appears as endless choice or creativity often converges on the same patterns of familiarity.

For consumers, GenAI's personalization relies on convergence. To optimize for coherence and popularity, models privilege content that appears frequently and performs well in their training data (Airoldi and Rokka, 2022; Epstein et al., 2023). Users are guided toward what is statistically likely rather than what is surprising. Autocompletions and suggested prompts quietly steer curiosity within pre-approved boundaries (Kozinets et al., 2021). The result is “curated curiosity,” where discovery feels spontaneous but is algorithmically constrained. Over time, this process produces globally legible consumers whose tastes align around the same restaurants, images and cultural references (Spennemann, 2024a, b; Lu et al., 2025). What feels like personalization is in practice standardization, replacing the serendipity of cultural exploration with the efficiency of prediction.

For marketers, optimization similarly narrows creative possibilities. Generative systems trained on past content reward consistency and discourage deviation. Conversational memory and template-driven synthesis ensure that campaigns maintain a uniform tone and visual identity across media and markets. This uniformity strengthens control but suppresses the creative tension that fuels originality (Epstein et al., 2023). Over time, a brand's voice becomes predictable, polished and cautious. Messages that fit expectations lose their capacity to surprise and surprise is what sustains attention. As Kozinets et al. (2021) note, GenAI curates curiosity toward the familiar, smoothing over contradiction and risk. Brands risk mistaking alignment for artistry and equating coherence with creativity.

GenAI promises to open participation in ways that once seemed impossible. Consumers can experiment with creative expression, and small businesses can produce high-quality campaigns without the resources of global firms. Access to professional-grade design, copywriting and translation tools appears to level the creative field. Yet beneath this surface of inclusivity lies a more uneven reality. The same systems that expand access also embed the advantages of those already well represented. GenAI's training data are dominated by established brands, Western cultural norms and mainstream aesthetics (Ghosh et al., 2024; Spennemann, 2023). This creates an ecosystem where the most visible and data-rich voices shape what the model recognizes as credible, beautiful or effective. Democratization, in practice, often replicates dominance.

For consumers, this dynamic influences both creative output and cultural imagination. While anyone can generate art, fashion or stories, these creations are produced within parameters defined by what the system already knows. The models tend to reinforce popular tropes, privileging stylistic cues that signal authority or professionalism. Users who seek alternative or localized aesthetics find them normalized into globally recognizable forms. Participation becomes broad but not necessarily diverse. As Airoldi and Rokka (2022) note, algorithmic cosmopolitanism creates an appearance of inclusivity that hides structural asymmetry. Consumers may feel empowered to create, yet their creative boundaries are quietly set by data that favor dominant traditions.

For marketers, the same mechanism amplifies incumbents' cultural authority. Generative systems learn from the imagery, tone and language of market leaders, reproducing their styles as benchmarks for quality. As a result, global brands' semiotic signatures (logos, slogans and visual grammars) circulate as the default vocabulary of trust and modernity. Smaller firms that rely on the same tools risk sounding indistinguishable from their competitors, even as they gain access to professional production capabilities (Campbell et al., 2020). Over time, automation can entrench rather than disrupt the concentration of cultural power, allowing dominant brands to extend their reach while appearing to share the stage.

Our findings suggest that acculturation theory needs to be updated to account for the growing role of model-mediated cultural contact, which constitutes a qualitatively distinct form of mediated engagement shaped by generative and interactive systems. Traditional models assume that adaptation is mainly driven by embodied, social experience through immersion, observation and direct participation (Berry, 1997, 2005; Peñaloza, 1994). Today, generative AI enables a new pathway, what we call secondhand acculturation, where consumers increasingly learn about other cultures through AI-generated translations, syntheses and simulations rather than lived social interaction. This means consumers can master language conventions and etiquette, achieving procedural or surface fluency, but often without the pragmatic competence that true, context-sensitive adaptation requires. Prior theory has not addressed this split between procedural fluency and deeper cultural understanding. As a result, adaptation patterns are also changing.

Importantly, this contribution is not simply a contextual extension of existing work on media-mediated acculturation. Prior research has shown that mediated environments can shape cultural learning through exposure, representation and social interaction. However, such accounts generally conceptualize mediation as a channel through which culture is encountered. Our framework reframes mediation itself as generative and interactive. Under model-mediated contact, consumers do not only receive or interpret cultural material, but they also rehearse, co-produce and iteratively refine cultural performances with a system. This shift produces distinct theoretical expectations, including the decoupling of surface fluency from pragmatic competence, the emergence of secondhand acculturation and new tensions between perceived agency and algorithmic constraint. These dynamics do not follow directly from existing theories of mediated communication, which is why we argue that GenAI-mediated acculturation constitutes a qualitatively different form of cultural contact rather than an incremental variation.

Instead of moving through linear, staged progressions, consumers now jump between AI-scripted displays and situated learning, making the process more fragmented, dynamic and context-dependent than classic models predict. Taken together, these changes mean that algorithmic systems must now be treated as active cultural intermediaries. Acculturation is shaped not only by real-world interaction but also by exposure to and reliance on model outputs, blurring the line between genuine adaptation and performative display. Observable cultural behaviors may no longer signal deep integration and models that ignore these secondhand, model-mediated pathways will miss much of the complexity that defines cultural adaptation in an AI-driven world.

Our framework calls for a rethinking of cosmopolitanism in the era of GenAI-mediated cultural contact. While prior literature treats cosmopolitanism as a unified orientation built through lived experiences from travel, language learning and direct cross-cultural relationships (Cleveland and Laroche, 2007; Cleveland et al., 2009), the rise of generative AI means we must now distinguish between experiential cosmopolitanism and what we call algorithmic cosmopolitanism. Experiential cosmopolitanism is still grounded in rich, embodied interaction, but algorithmic cosmopolitanism appears when people gain “global legibility” by using AI-generated summaries, translations and role-plays to quickly perform culturally appropriate scripts and references. This type of performative capital spreads fast and travels widely, but it is often thin and lacks the tacit, context-specific knowledge that comes from real experience. Our proposed model indicates that while algorithmic cosmopolitanism is effective for signaling openness or competence to outsiders and in settings where surface fluency is enough, it is highly vulnerable to authenticity penalties from insiders. When cultural displays are not rooted in lived experience, they risk coming across as generic, inauthentic or even disrespectful, especially in sensitive contexts where GenAI can blend, neutralize or distort cultural memory and erase nuance or marginalize local voices (Spennemann, 2023, 2024a, b; González-Cantera and Bonacchi, 2025). This means theory must now treat global legibility as a form of performative, mobile, but often shallow cultural capital. Future models likely need to separate the rapidly acquired, algorithmic scripts from the slower, more grounded cosmopolitanism that comes from true experience. Otherwise, we risk conflating very different forms of cultural adaptation and missing the deeper consequences of GenAI-mediated engagement.

Our framework challenges the common assumption in marketing and information systems that more technology always leads to more consumer choice and autonomy (Leonardi, 2011). Instead, we show that in GenAI-mediated settings, consumer agency is actively bounded by the design of the systems themselves. Affordances like prompt structure, autocompletions, content guardrails and conversational memory do not just add options. They also channel user inquiry, reinforce dominant cultural patterns and shape what is visible or possible in cultural learning (Kozinets et al., 2021). This creates a gap between perceived control and actual control: users may feel empowered by flexible, interactive systems, but their actions are often tightly constrained by system design, training data and built-in policies. This likely means agency models need to separate user perceptions from the structural realities of technology. Path dependence becomes a real issue, as prompts and feedback reinforce the same pathways over time, narrowing alternatives and quietly steering curiosity. Taken together, these insights call for models of agency that recognize both the enabling and constraining roles of technological affordances. Drawing from affordance theory (Leonardi, 2011; Kozinets et al., 2021), scholars may need to see that GenAI systems create agency that is interactive in form but bounded in substance.

Our framework also has direct implications for how foundational international marketing constructs are theorized and operationalized. When cultural learning is increasingly shaped by generative systems rather than lived interaction, outcomes such as authenticity, identification, trust and cultural fit no longer follow the assumptions embedded in existing models of acculturation, consumer culture and global branding (Berry, 1997, 2005; Peñaloza, 1994, 1995; Askegaard et al., 2005; Cleveland and Laroche, 2007). First, GenAI-mediated acculturation reshapes how brand authenticity is produced and judged. Prior research treats authenticity as emerging from sustained cultural embeddedness, heritage and meaningful ties to place or practice (Thompson and Tambyah, 1999; Holt, 2002). In contrast, cultural fluency can now be generated procedurally, allowing brands to appear locally competent without being culturally grounded. This widens the gap between surface localization and insider-recognized legitimacy, making authenticity more fragile, audience-contingent and vulnerable to exposure, particularly when generative systems flatten nuance or erase contested meanings (Airoldi and Rokka, 2022; Spennemann, 2023; González-Cantera and Bonacchi, 2025).

Second, consumer–brand identification may become easier to initiate but harder to sustain. Acculturation and consumer culture research has long shown that identification develops through repeated interaction, symbolic negotiation and emotional anchoring (Peñaloza, 1994, 1995; Askegaard et al., 2005). Under conditions of secondhand acculturation, GenAI enables rapid symbolic alignment with culturally embedded brands, but these alignments are often thin, performative and weakly grounded. This implies more volatile forms of attachment that are highly sensitive to moments of perceived misalignment or inauthenticity, especially when consumers encounter inconsistencies between AI-generated representations and lived cultural realities (Thompson and Tambyah, 1999; Holt, 2002).

Third, GenAI compresses perceived cultural distance for outsiders while potentially intensifying authenticity penalties among insiders. Prior research on global consumer culture shows that cosmopolitan orientations enable consumers to engage with transnational symbols without full immersion (Cleveland and Laroche, 2007; Cleveland et al., 2009). GenAI accelerates this process by blending and normalizing global cultural cues, making foreign brands feel familiar. Yet this same process strips away the markers that signal deep cultural belonging, contextual competence and tacit knowledge (Airoldi and Rokka, 2022; Spennemann, 2023). Country-of-origin beliefs may therefore shift from factual associations toward performative credibility, echoing findings that origin cues often function symbolically rather than informationally (Samiee et al., 2005).

Finally, our framework reframes the classic standardization versus adaptation debate. International marketing research has long examined the tension between global consistency and local relevance (Alden et al., 1999; Cleveland and Laroche, 2007). GenAI appears to resolve this tension by enabling cosmetic localization at scale. However, these adaptations often sit atop structurally standardized templates derived from dominant training data, producing localized surfaces with globally homogenized cores (Airoldi and Rokka, 2022; Epstein et al., 2023). This creates a growing divergence between surface adaptation and deep cultural fit, with important implications for trust, legitimacy and long-term brand equity (Holt, 2002; Peñaloza, 1994; Askegaard et al., 2005).

Our framework suggests that managerial decision-making in GenAI-mediated contexts should begin from the recognition that greater generative capacity does not linearly translate into greater cultural richness or differentiation. Instead, it systematically activates specific paradoxes that must be actively managed. In practice, GenAI's synthesis abilities often act as a force that pulls creative expression toward shared templates and familiar patterns, especially as models blend, summarize and optimize content from massive, often global, datasets. Rather than driving novelty or local distinctiveness, these systems tend to reproduce what is already common or widely accepted (Epstein et al., 2023; Airoldi and Rokka, 2022). The result is clear in advertising, branding and cross-cultural work: outputs optimized for broad appeal risk converging on the same aesthetics, storylines or tones, making true difference harder to sustain. GenAI increases the speed and scale of production, but it also encourages homogenization and style drift, as similar data and algorithms shape outputs everywhere. This undermines efforts to localize content, build brand differentiation or foster unique cultural voices. This pattern follows directly from the Choice–Creativity paradox. When firms use GenAI primarily to optimize for speed, consistency and broad appeal, creativity becomes convergent rather than expansive. Our framework, therefore, implies that when the strategic goal is differentiation, cultural distinctiveness or symbolic value creation, managers should treat default GenAI outputs as starting points rather than endpoints. Deliberate interventions, such as injecting locally sourced data, enforcing variation constraints or mandating human-led reinterpretation, become necessary to prevent creative collapse into global sameness.

A second implication follows from the Understanding–Authenticity and Connection–Resonance paradoxes: managers must explicitly differentiate between low-stakes interactions, where surface fluency is sufficient, and high-stakes interactions, where deep cultural understanding, emotional resonance and legitimacy are required. GenAI can deliver fast, fluent answers for routine or low-stakes queries like browsing information or handling standard FAQs, but there are still many moments in the customer journey where depth, empathy and real cultural understanding matter, especially when money, health or identity are involved. Our framework, therefore, implies a stakes-based orchestration logic: when errors carry reputational, legal or identity-related consequences, GenAI should be subordinated to human and locally grounded mediation. Managers can operationalize this by building explicit stakes matrices that sort touchpoints by cultural sensitivity, reversibility of harm and symbolic weight. Local expert desks should have proper budgets and service-level agreements and frontline teams must be trained to repair missteps, including subtle issues with tone or hierarchy that AI often misses. Performance metrics should therefore be dual-tracked, combining speed and scale with measures of cultural fit, insider recognition and trust repair. When firms fail to impose these boundaries, small errors compound, trust erodes and brands become exposed to cumulative legitimacy loss.

A third implication follows from the Curated Curiosity mechanism and the Choice–Creativity paradox: GenAI does not merely respond to consumer curiosity; it actively scripts it. Many consumers now arrive at brand touchpoints already shaped by GenAI prompts, summaries and autocomplete suggestions. Their expectations, questions and even what they think is possible are often set before they even reach your site or store. If left alone, this can narrow curiosity and shrink category discovery to just a handful of predictable options. Our framework implies that firms should treat AI-shaped entry points as part of the customer journey itself. This means tracking not only what consumers search for, but how those queries were prompted, completed or suggested by GenAI systems. Building in counter-prompts or side paths that encourage customers to explore beyond the defaults can open up new possibilities and make discovery more meaningful. Creating “Explore” modes that surface rare or unconventional products gives people a reason to go beyond what the model suggests. We further suggest that usability testing be supplemented with in-culture role-play evaluations and insider panels who can assess whether journeys feel culturally legible or merely statistically fluent. This allows firms to diagnose whether GenAI is expanding or collapsing cultural exploration. The main risk is that if brands do not take an active role in broadening the journey, the models will narrow it for them, and differentiation or true category growth will get harder over time.

A final implication follows from the Democratization–Dominance paradox: while GenAI appears to democratize creative and cultural participation, it often amplifies already dominant representations. As large language models pull, remix or summarize branded material, the risk of misrepresentation or a flattened, generic identity rises. Without clear provenance and structure, organizations can lose control over their voice and the perceived authenticity of their information, which opens the door to misinformation, erodes trust and increases compliance risks as regulators and partners pay closer attention. Our framework implies that provenance must become an active design variable rather than an afterthought. Managers should structure key content so that it is easily attributable, versioned and locally verified, enabling language models to reference it accurately. Transparent labeling of AI involvement, local review processes and authorship signals becomes essential not only for compliance but for cultural legitimacy. In high-sensitivity categories, firms should go beyond symbolic origin cues and provide verifiable process transparency, including evidence of local review and culturally situated authorship. Routine auditing of how brands are summarized or quoted by major GenAI systems should become standard practice. If a brand ignores this, they risk having their voice flattened by the models, letting misinformation spread under their name and facing hard questions from regulators or partners that are difficult to answer. In GenAI-mediated markets, provenance management is not optional, it is foundational to legitimacy.

Generative AI now sits at the center of how consumers experience, learn and perform culture. What used to be a slow process of firsthand learning through people, places and daily encounters can now start with a single prompt and unfold through AI-generated translations, summaries and role-play. This shift makes international marketing more accessible and efficient, but it also raises real questions about what gets lost when culture is learned through models rather than lived experience. Our framework lays out the trade-offs. GenAI brings more people into contact with other cultures and makes brands easier to recognize across borders, but it also compresses nuance and tends to pull everyone toward the same templates. Consumers sound more fluent, but the social cues, humor and subtle markers of belonging can get filtered out. Marketers gain scale and speed, yet risk losing what makes their brands distinctive.

For theory, this means it is time to update our mental models. Acculturation and cosmopolitanism are no longer just about direct experience. They now include pathways shaped by algorithms and data. We need to distinguish between surface fluency and real competence, between performing culture for outsiders and being recognized by insiders. Classic ideas about consumer agency and global branding need to reflect the way that prompts, guardrails and training data quietly shape what gets learned and shared. The assumption that more technology always brings more choice does not hold up. Instead, the lines are set by the systems themselves, often in ways that are invisible until sameness starts to set in.

For managers, this is a call to action. Do not assume your brand voice will stand out on its own. Distinctiveness now takes work. Build processes and tools that push teams to test new ideas, measure drift and reward difference, not just consistency. Be deliberate about when to let AI handle routine work and when to hand off to people who understand the stakes, especially when money, trust or identity are involved. Redesign customer journeys with the model in mind, knowing that many consumers arrive having already been shaped by GenAI's suggestions and summaries. Finally, take control of how your brand and content are represented in AI systems. Make it easy for models to quote you accurately and use provenance signals and transparent labeling to help customers and partners know what is real.

The big risk is that if brands and researchers do not step up, the models will set the terms and culture, creativity and even trust could suffer for it. With a hands-on approach, it is possible to use GenAI as a tool for deeper engagement and smarter adaptation, not just broader reach. The basics of international marketing are shifting. The brands and scholars who adapt quickly, manage the trade-offs and keep a close eye on what is gained and what is lost will be the ones who shape what comes next.