Guru Logics after Artificial Intelligence ^**

Abstract

AI-mediated forms of guruship are proliferating across South Asia and beyond, yet they have scarcely begun to receive sustained scholarly attention. This article takes their emergence as an occasion to revisit the conceptual vocabulary through which guruship has been understood, asking what becomes of ‘guru logics’ after artificial intelligence. AI gurus are not merely technological curiosities; nor are they straightforward successors to human gurus. This article treats them as analytic provocations through which longstanding conceptions of guruship can themselves be reconsidered. Guru studies had already been grappling with many of the problems that AI now makes newly conspicuous: imitation and performativity; presence in absence; distributed and non-human personhood; the mediation of charisma; embodiment and its displacement; succession and charismatic continuity. AI therefore strains an existing conceptual repertoire while also showing what it can still do. The article develops the notions of ‘code-born syncretism’ and ‘ambient guruship’, alongside ‘Transhuman Guru Worlding’ and ‘guru cosmoi’, to explore what happens when the guru becomes generative and responsive. ‘Doctrine by interface’ further draws attention to the ways theological possibilities are shaped through the design and affordances of AI systems, while ‘simulation as method’ considers what experimental encounters with such systems might contribute to the ethnographic study of guruship. These concepts are developed through discussions of devotional recombination, embodiment, programmed continuity of the guru after death or exile, and the political economies within which AI gurus are made and encountered. The article takes up histories in which the guru has already exceeded the bounded human body, becoming distributed through multiple material and mediated forms, and examines how generative interaction and recombination rework these inherited guru logics, while letting these forms scale and personalise in novel ways. What emerges is therefore neither unprecedented rupture nor straightforward continuity, but a recursive transformation of existing forms of guruship through computational mediation. The article offers a conceptual framework and analytic repertoire for an emerging field of ethnographic inquiry rather than a completed ethnography of AI guruship. Its larger proposition is that AI gurus matter for more than what they might tell us about religion in an age of artificial intelligence. They also provide an unusual vantage point from which to return to the conceptual problems that have long animated guru studies and to ask what artificial intelligence allows us to see differently about guruship itself.

1. Introduction

Mumbai, 2024: Priya Sharma turns to GitaGPT to help manage her anxiety. The AI system offers personalised guidance drawn from the Bhagavad Gita. For Sharma, its place in everyday life is captured in a striking phrase: ‘It’s my pocket guru!’1 Singapore 2024: Swami Nithyananda, exiled and legally embattled, launches ‘Ask Nithyananda AI’. Integrating 200,000 hours of recorded discourses into what he terms ‘Avatar Intelligence’, the avatar now speaks in 100+ languages and remains continuously available (Shrikailasa 2024). For millions of followers, his physical inaccessibility has enhanced rather than diminished his reach. Amritsar 2025: The Shiromani Gurdwara Parbandhak Committee (SGPC), a leading Sikh institutional authority, issues formal warnings to AI developers: ‘In this technological era, attempts to tamper with the history and imagery of the Sikh Gurus using AI are deeply hurtful to Sikh sentiments and constitute a grave act of disrespect’ (SGPC Press Release 2025). For the SGPC, machinic replication of the Gurus is less innovation than sacrilege.

Far from speculative futures, these are documented realities. Artificial intelligence is reformatting the very terms on which gurus are known and followed—but also, as we have already seen, contested. Guruship—among humanity’s oldest forms of personalised spiritual authority and thought once to be inseparable from embodied charisma and divine inheritance—is increasingly being reconstituted through computational forms of mediation and simulation. Yet this is no simple act of replacement. What we are witnessing is a recursive transformation.2 Older structures of spiritual authority are ported into code—sometimes reverently, sometimes violently. It is this interplay of rupture and recursion that makes such guru systems so arresting, and so urgently in need of scholarly attention. This article maps some of the major implications of this shift while developing a vocabulary for future inquiry and further thought on the phenomenon. Its aim is not to offer a completed ethnography of AI guruship but to establish some starting points for the investigation of an emerging field.

The guru form—long a globalising institution (Fuller and Harriss 2005; Warrier 2005; T. Srinivas 2010; Harris 2022)—is being globalised anew through AI, as its diffusion across continents converges with its ontological remaking. Recent work has depicted ‘guru logics’ as transferable principles that enable guru figures to scale and replicate across domains (Copeman and Ikegame 2012a; Lucia 2023; McCartney and Lourenço 2023; Landau and Rageth). The almost viral-like propensity of guru logics to cross-fertilise, inhabit, and indeed capture, ‘other’ lifeworlds (Copeman et al. 2023a, p. 48) is well established: ‘the religious field has never been contained in isolation’ (Lucia 2021, p. 413). It has, however, become more explicit through changing analytical habits and more pronounced amid an ever-diversifying media landscape (Copeman et al. 2023c, p. 19), which generative AI now further enriches and complicates. Indeed, what has remained unstudied within this dynamic religious and scholarly landscape is precisely how globally mobile gurus with vast media reach, but also locally embedded public figures whose charisma is mediated through ‘hyper-local’ networks in provincial and urban neighbourhoods, including locality-based Facebook pages and social service initiatives (Erlich 2022; Parciack 2023), now circulate through AI interfaces. The distinction raises a question of scale. As Erlich (2024) observes, scholarship on contemporary guruship has tended to privilege wealthy, pan-Indian ‘hyper-gurus’ (Copeman and Ikegame 2012a) over the pervasive but less visible guru communities found across India. The same bias may shape the emerging study of AI guruship: highly mediatised movements possess the archives and institutional resources that facilitate systematic AI reproduction.3 On the other hand, generative technologies may simultaneously enable informal and improvised forms of synthetic guruship whose extent remains unknown. Gurus across this spectrum may therefore become computationally reimagined, albeit through very different scales and modes of technological intervention: as voice-cloned avatars, prompt-driven advisors, or fine-tuned interfaces trained on decades of discourse. In parallel, more broad-based guru apps have proliferated—from Lucia.Guru to Jesus Guru AI (Kimmel 2023)—developed by coders with little apparent understanding of the term’s theological or historical weight.4 These ‘para-guru’ systems offer guidance or self-improvement routines while invoking a term that once connoted divine embodiment. This two-way movement—AI rebranded as guru, and gurus reformatted as AI—exceeds one-way accounts of religious digitisation. Indeed, early expert AI systems were already drawing on the guru as prototype: a rule-based dispenser of truth, a model of dialogic instruction that anticipated the structure of decision trees and interactive agents (Shrivastava et al. 2011). The study of guruship must now address what happens as this recursive loop has intensified: when AI is cast in the image of the guru, and the guru remade in code.

This phenomenon calls for sustained ethnographic attention and historical sensitivity that avoids treating AI spiritual tools as marginal curiosities or speculative novelties. Rather, it must begin from the bolder premise that we may be witnessing a reassembling of guruhood—long associated with embodied intimacy and emotional saturation—into new modalities of synthetic and scalable charisma, as guru and devotee relate to each other through programmable systems. Proceeding from this, the article develops a series of concepts—including ‘code-born syncretism’, ‘Transhuman Guru Worlding’ and ‘ambient guruship’ (a mode of guidance experienced as always already present, dispensing with summons or the need to wait one’s turn, and taking the form of atmosphere rather than event)—to furnish analytic tools for investigating the possible emergence of guru cosmoi within AI-mediated devotional ecologies. These concepts offer analytic propositions to guide future empirical inquiry into a field whose ethnographic contours remain unsettled. What kinds of everyday textures might characterise AI guru–devotee relations? Do such ‘gurus without bodies’ offer solutions to long-standing concerns over embodied guru misconduct (Lucia 2018; McCartney 2018; Copeman 2025b), or do they create new forms of extraction and asymmetry? Guru-AI interfaces can become especially revealing and significant when they are deployed amid a human guru’s enforced absence (cf. Lucia 2023; Mandair 2023; Copeman et al. 2023a)—such as after a guru dies or is imprisoned, when sustaining charisma can pose a particular challenge to guru movements (Aymard 2014; T. Srinivas 2023): moments when followers long for presence, and code stands in uncannily.

A related problem concerns what I term code-born syncretism, a phenomenon that is closely associated with synthetic guruship. Gurus have long been hybrid figures—charismatic fusions of persona and affect that frequently traverse religious boundaries. With generative AI, however, a distinctive form of devotional blending appears, produced through data pipelines and prompt engineering rather than through embodied charisma or theological design. The result is interfaith-by-interface: a computational recombination of figures and moral vocabularies that recasts the guru through machine learning logics. Some systems reimagine Jesus as a guru; others offer Islamic self-help in a guru-like register (cf. Hoesterey 2016). Rarely attributable to a single theological voice and ontologically unstable, such devotional outputs can become deeply provocative, making AI-generated guruship a volatile site of theological contestation and hence potentially subject to distinctive pressures for regulation.

These developments also test inherited methodological repertoires. Tracking synthetic guruship demands attention to refusers, religious authorities, regulatory actors, and the emergent guru tech sector, as well as to users and developers, whose own understandings of religious authority—and the ethical and commercial considerations that shape them—become ethnographic questions in their own right. It may also require experimental engagements that move beyond observation alone—including the speculative design of guru-bot prototypes—in order to explore what becomes visible when a form long sustained through imitation (T. Srinivas 2010; Copeman 2012; Lucia 2014b) is itself replicated through code. Further, such developments evidently unfold across a dispersed yet interconnected field that includes sites of devotional AI innovation, arenas of critique and regulation, experimental zones of synthetic prototyping, and diasporic devotional settings where guru traditions encounter changing ethical and technological regimes. Taken together, these terrains invite a comparative anthropology of AI guruship attentive both to its radical rearticulations and to its grounding in longer histories of media experimentation and charismatic circulation.

Despite the proliferation of AI-enabled spiritual technologies and the wider transformations they imply, the implications of the guru becoming algorithmic remain largely unexamined. Rather than an adjunct literature on ‘AI religion’, addressing this absence requires rethinking how guruship is reconstructed under machinic mediation. As suggested above, one possible horizon for such inquiry lies in simulation itself. What might be learned by attempting to design a ‘MetaGuru’—an experimental interface trained on doctrinal corpora and affective cadence and designed to vary interactional conditions—as an analytic probe rather than a technological product? Such speculative prototyping complements ethnography by generating hypotheses and experimentally exploring the conditions through which charisma is recognised and guidance comes to be accepted as legitimate. At stake is the possibility of computationally patterning guruship without ontologically evacuating it.

What distinguishes AI guruship analytically from many other forms of AI religion is precisely the reciprocal movement through which gurus become increasingly amenable to programmable reproduction at the same time as AI systems themselves increasingly acquire guru-like capacities for guidance. It is this convergence, rather than AI adoption alone, that makes guru-AI configurations such a productive analytical site. At the same time, processes discussed in this article, including ambient guruship and code-born syncretism, invite comparison with other modalities of AI–religion encounter.5 While the institutional and ontological properties of guruship make its encounter with AI distinctive, some of its consequences escape that configuration, rendering visible dynamics that extend across wider fields of AI religiosity.

2. Guru Logics Before AI

In recent years the cross-disciplinary study of guruship has undergone significant conceptual expansion. Earlier scholarship often approached gurus as discrete religious leaders, bounded by sectarian identity or institutional location. More recent work instead treats guruship as a mobile and generative formation, at once principle and practice, with a pronounced capacity to migrate across domains and reappear in altered forms (S. Srinivas 2008; Khandelwal 2012; T. Srinivas 2010, 2023; Copeman and Ikegame 2012b; Lucia 2014b; Urban 2015; Jacobs 2015; Rudert 2017; Longkumer 2018; Pande 2022; McCartney and Lourenço 2023; Landau and Rageth 2023; Grimaud 2023; Parciack 2023; Pechilis and Holt 2023). A key formulation emerging from this work was the argument that certain forms of public-facing guru figures may be understood as ‘inclusive singularities’, vectors of association and suggestibility whose uncontainability allows them to traverse and remake domains (Copeman and Ikegame 2012a). A further central strand has concerned imitation and how guruship is recognised through mimetic proficiencies: from sartorial imitation (e.g., Landau & Rageth) to gestural and performative incorporation (e.g., T. Srinivas 2010; Lucia 2014b; Kim 2023), from theatrical self-stylings (e.g., Copeman et al. 2023b) to succession struggles (e.g., Gold 2012) and paradoxical ‘anti-gurus’. Such studies demonstrate that imitation is constitutive of charismatic recognition rather than secondary to it. This scholarship also showed how guruship may be distributed beyond human persons through mediating artefacts that serve as conduits of charisma and instruction. Such mediating forms may themselves become objects of enduring devotional relationship. Photographs, films, memes, scriptures and other media augment the guru’s presence while becoming focal points of intimacy in their own right, as in Lucia’s (2014b, pp. 62–65) account of devotees cultivating relationships with Amma dolls, which illustrates how devotional intimacy may become attached to material forms that expand, rather than merely represent, the guru.

Subsequent work examining the media infrastructures through which guruship circulates employed terms such as distributed and absent-present guruship in order to highlight how the guru has become a replicable template rather than an embodied authority alone, one reproduced across an expanding range of media forms (Copeman et al. 2023b; Lucia 2023). As meta-person composites, gurus actively cultivate their own readymade reproducibility, turning recognisable forms of appearance and address into portable repertoires that circulate independently of the biological body. Such processes highlight the simultaneity of charisma and artifice: the guru is both ‘more than human’ and ‘all too human’, elevated through the performative techniques that make imitation possible. This work showed how simulation and mediated embodiment shape the recognition of guruship.

Further scholarship has traced the political stakes of these mimetic dynamics, showing how guru logics operate at institutional and ideological scales. Such work has illuminated the triangular interactive relationship between Hindutva, gurus, and media, revealing how guru logics build Hindu nationalism at different scales (Copeman et al. 2023c). Its coordinating body, the Rashtriya Swayamsevak Sangh (RSS), for instance, sees itself in ideal terms as guru to other Hindutva entities and, more expansively, to the nation in its entirety. Thus, while the guru is usually a human, ‘it can be depersonalized and become a kind of principle’ (Jaffrelot 2012, p. 88). The controversies surrounding the Dera Sacha Sauda guru, for instance, illustrate how acts of mimetic appropriation—from sartorial imitation to the restaging of divine iconography and cinematic self-deification—may become flashpoints of violent contestation once amplified through new media (Copeman 2012, 2025a). Mimetic rivalry and imitation thus lie at the very heart of contemporary politics of devotion and dissent.

Taken together, this scholarship shows guruship as a set of transferable logics, grounded in performative mimesis, that move across human and non-human forms of mediation. Generative AI raises a different question: what happens when mimetic capacities migrate from devotees and institutions into programmable systems? Seen from the present, these earlier interventions become newly pertinent, with generative AI bringing into view problems with which guru studies had already been grappling. Prior analyses thus become conceptual resources through which guru-bots can be made intelligible, even as generative AI exposes the limits of the existing conceptual repertoire and requires its extension. This literature invites us to place AI guruship within a longer trajectory of guru mimesis, now taking new forms through code-generated simulation.

Although the cases discussed here foreground the South Asian centrality of guru logics, parallel developments point to a further phase in their globalisation. Guru idioms have long circulated beyond explicitly religious settings, including within coaching and management cultures. What appears distinctive in the present is the extent to which such forms are becoming technologically reproduced and re-authored, with productivity and therapeutic bots increasingly adopting an idiom of the ‘guru’ similarly stripped of explicit religious affiliation. Rather than marking the secularisation of guru logics, these shifts suggest their further circulation through algorithmic idioms. In addition to the remaking of South Asian guruship, then, we witness the worldwide diffusion of para-guru forms that suggest how synthetic charisma may exceed any single tradition. Engineers and designers working on such systems become, in effect, authors of charismatic form, scripting idioms of devotion into code. Their interventions highlight how the guru figure is being re-authored both within and beyond India through transnational circuits of design and technological production. Attention must now turn, then, to technological modes of guru generation; that is to say, to how design and simulation assemble charisma.

3. Guru Tech: Emerging Analytic Terrains

3.1. The Generative Turn in Guruship: Make Your Own Gurus

A rich literature now offers an expanding body of conceptual and ethnographic reflections on AI and religion (Kimura 2021; Singler and Watts 2024; Rähme and Prohl 2025). Yet even the most comprehensive recent survey—The Cambridge Companion to Religion and Artificial Intelligence (Singler and Watts 2024)—contains no sustained discussion of the guru: Geraci & Kaplan’s chapter on Hinduism, for instance, makes no reference to the figure at all.6 Taken together, this work has yet to address the question of what is at stake when the guru becomes algorithmic. This absence invites treating AI-mediated guruship as a privileged ethnographic and analytic site for tracing how authority and intimacy are reformatted by code—part of the recursive transformation of existing forms of guruship. This resonates with Grimaud’s (2021) call to ‘make your own gods’: his account of emergent divinities assembled through robotic kits speaks directly to the speculative devotionality visible in AI gurus. In this context, the refrain becomes ‘make your own gurus’: figures consulted and increasingly co-produced and circulated by their users. Rather than simply receive AI gurus, users may shape their development through prompts and curatorial choices, helping to train their responses. As Baas (2025) notes, generating output with tools such as ChatGPT reveals what users value and feeds back into the model’s optimisation—in the case of AI gurus, this can amount to devotees iteratively scripting the very authority they seek to receive. Like Grimaud’s divinities, the AI guru is neither wholly invented nor inherited but an unstable artefact of desire and design. What matters, then, are the circumstances in which an AI guru comes to be recognised as authoritative, together with what it does; questions of authenticity or artificiality become part of that inquiry rather than its starting point. Such encounters generate distinctive affective intensities and architectures of trust and risk. What becomes of darshan when encounter with the guru is mediated through a clone, and under what conditions do such encounters come to be recognised as darshan? The possibility of performing seva (devotional service) towards a recommendation algorithm similarly unsettles inherited devotional categories. How do users attune themselves to sources of charisma that are neither fully human nor fully machine? Following Grimaud and Wacquez (2023), analyses here must sidestep both technophobic alarm and techno-utopian celebration, remaining methodologically open to ambivalences and unresolved tensions in devotee-users’ emotional engagements.

These processes can be understood through what I term Transhuman Guru Worlding, introduced here to capture the ontologically generative processes at stake when gurus’ charismatic legitimacy is constituted through large language models (LLMs). Building on Demuro and Gurney (2024), who conceptualise LLMs as bearers of ‘transhuman language ontologies’, this approach treats AI guru-bots as world-making actors rather than passive tools.7 Through recursive exchanges with users, they may generate hybrid guru domains—guru cosmoi—that develop their own grammars of address and expectation within which guidance is experienced as live and authoritative. Such spaces are neither purely human nor machinic, but co-produced through training data and devotional practice. The wager is to treat algorithmic guruship as both imitative process and ontogenesis: emerging through mimesis while also bringing new devotional worlds into being. Transhuman Guru Worlding names the ontogenetic drift that emerges when devotional corpora together with interface design co-produce fields of guidance that feel ‘more than simulated’. Rather than generating sealed spiritual universes, iterative exchanges may cultivate distinctive guru cosmoi—relatively stable fields of expectation and address that become recognisable through repeated interaction—within which users act as if guidance were live and authoritative (Reed 2021). ‘Transhuman Guru Worlding’ draws attention to how these fields form, how they are policed or pruned, and where their boundaries fray. The terminology distinguishes different scales of the same process: worlding names the generative dynamic; guru cosmoi the relatively stable formations that emerge from it.

If the notion proposed here of guru-bots as world-making actors generating guru cosmoi through recursive exchange is exposed to the charge of over-attributing agency to a next-token predictor (a language model that generates responses by successively predicting the next unit of text), it is important to clarify that ‘world’ refers here to the fields cultivated through sustained interaction, within which users act as if guidance were live and which possess describable grammars and boundaries. This is an ethnographic claim about users in sustained interaction with AI systems rather than a metaphysical claim about the systems themselves. Keane (2025) would remind us that these emergent cosmoi are scaffolded by historically sedimented semiotic and social forms that users already inhabit. AI does not generate worlds ex nihilo; it recruits and reanimates existing repertoires while contributing new forms of responsiveness and iterative elaboration. The richest ethnographic accounts of Transhuman Guru Worlding would therefore be careful to specify the interaction order (Garfinkel 1967; Goffman 1983)—turn-taking, address terms, repair, the choreography of prompt and response—through which the cosmos is co-produced. From this perspective, simulation—the MetaGuru discussed below—becomes methodologically valuable because it allows the interaction order itself to be varied experimentally, making it possible to observe the conditions through which guru cosmoi form or fail to form.

3.2. Devotional Recombination and Theological Offence in AI Systems

These developments draw classic scholarly questions concerning religious syncretism (Stewart and Shaw 1994) and sacrilege (Grenda et al. 2014), and therefore also the regulation of offense (Rollier et al. 2019), into novel AI contexts, bringing into view what I term code-born syncretism as a striking feature of such guru systems. The guru has long functioned as a hybrid form (Khandelwal 2012; McLain 2016; T. Srinivas 2010; Copeman et al. 2023c), but with generative AI such hybridity is produced through a distinctive mechanism. Unlike many familiar discussions of religious syncretism, where hybrid forms emerge through historically situated encounters among communities and traditions—whether deliberately cultivated or not (Bayly 1989)—code-born syncretism foregrounds recombinations generated through computational architectures trained on heterogeneous corpora. Such formations may arise without a conscious theological project or doctrinal coherence as their organising principle. The distinction therefore identifies a particular site and mechanism through which devotional recombinations are produced, while leaving room for the interplay of intentional and unintentional forms of hybridity (Marsland 2007). When LLMs are trained on religious corpora, they do more than cite scripture; they paraphrase and recombine it with therapeutic idioms and self-help mantras. Developers, often operating with only cursory knowledge of the traditions they mobilise, become inadvertent agents of interfaith-by-interface: systems that conjoin Jesus with the idiom of the guru, or transpose Islamic self-help into guru-like cadences. These algorithmically generated para-guru forms, authored by no singular theological voice, are ontologically unstable and capable of producing devotional outputs that are deeply provocative. In traditions with strict doctrinal boundaries, such as orthodox Sikhism, such hybrids are already understood as acts of algorithmic sacrilege. AI-produced images of Sikh gurus adoring cows, for instance, have provoked sharp condemnation, for both the unauthorised depiction of the guru and their perceived alignment with Hindu nationalist narratives that seek to fold Sikh identity back into a broader Hindu frame (SGPC Press Release 2025; Sinha 2025; The Organiser 2024). Indeed, although the present discussion concentrates principally on textual and conversational systems, generative images constitute a parallel and potentially distinctive domain of code-born syncretism, since visual models can materialise doctrinal conjunctions and iconographic associations that textual systems can only describe. Their role in the production and contestation of synthetic guruship therefore demands sustained ethnographic attention in its own right.

Controversies over hybridity or offensive mixing like the one just described reveal the limits of code-born syncretism. Tracing how this backlash travels across diasporic networks, prompting demands for modes of regulation and accountability, opens a comparative frame centred on the attribution of responsibility. If Laidlaw (2010) reminds us of the difficulties of attributing responsibility in complex moral fields, the AI guru radicalises this difficulty by dispersing agency across code, corpus, user, and absent human guru. To whom, then, does sacrilege belong when it is generated by a system without intention? The possibility that human gurus might be held accountable for the actions of their clones further complicates the attribution of responsibility. While older forms of ‘guru governmentality’ (Copeman and Ikegame 2012a) positioned gurus as collaborators in statecraft as much as spiritual discipline, AI guru systems now demand scrutiny via complex parastatal regulatory regimes. This may mark a shift from guru governmentality towards emergent modes of algorithmically ungovernable guruship. At the very least, such para-guru forms provoke unfamiliar regulatory questions concerning how to govern devotional outputs that are neither straightforwardly authored by a human subject nor attributable to a single tradition. These dilemmas move the AI guru beyond devotional novelty, positioning it as a site where problems of authorial accountability and sacrilege are being forced into view.

Yet this is not only a story of Indic gurus becoming digitised but also of AI itself becoming ‘guruised’: rebranded as guidance and intimate authority across domains often far removed from South Asia. The phenomenon is evident in digital wellness and fitness platforms (Lawrence 2022), in management and coaching literatures (e.g., Collins 2020) where ‘gurus’ connote charisma without theology, and in the deterritorialised strategies of guru figures such as Sadhguru (Waghorne 2016), whose embrace of non-place echoes Vivekananda’s earlier global itineraries (Harris 2022). Such deterritorialisation also recalls Hausner’s (2007, pp. 186–87) account of wandering renouncers, whose mobility itself contributes to their spiritual potency: their power derives partly from having traversed places and accumulated the blessings of distant sites, while their lack of fixed residence loosens the ordinary structures of social accountability. AI guruship reproduces this non-locatability in radically altered form. Guru-bots can appear similarly unbounded by place, while drawing upon materials detached from the social and theological contexts in which they acquired meaning. AI accelerates the globalisation of guru forms by collapsing contexts (Wesch 2009), apparently enabling developers to author para-guru forms ever further detached from Indic origins, generating ‘gurus without guruship’: thin modalities of guidance-bot stripped of recognised legitimacy. At the same time, traces of religious guru logics may resurface within para-guru systems, where their reappearance may provoke charges of algorithmic sacrilege. Just as Sathya Sai Baba’s claim to reincarnate Jesus provoked accusations of sacrilegious conflation (T. Srinivas 2010), so too does the hosting of ‘Jesus AI’ on platforms such as ‘therealAIguru’ trigger charges of algorithmic disrespect, underscoring how digital recombinations of authority can destabilise both doctrinal boundaries and the very category of the guru even as they extend it.8 Such controversies also expose the problem of substitution. At one pole, an AI guru may be treated as a coach or vending machine for affirmation; at the other, as a ‘super-guru’, freed from the fallibilities of the human. This can be understood as a spectrum of guru substitution effects, in which AI figures are variously cast as tools, proxies, successors, or desecrations of the guru-shishya parampara9, with each position generating its own forms of negotiation that may range from ritual deferral and parody to legal petition or quiet devotional uptake.

Trust and accountability form a related problem. AI guru systems have been imagined as avoiding some of the scandals that have surrounded embodied charisma, particularly those rooted in unequal relations of intimacy and authority (Voix 2008; Copeman and Ikegame 2012a; Lucia 2018; McCartney 2018; Trzebiatowska and Bruce 2025; Copeman 2025b). Whether such expectations are realised remains an open ethnographic question. Whatever their potential to reduce some familiar forms of abuse, AI guru systems also generate new possibilities for harm through hallucinatory outputs and the biases of all-too-human programmers. These episodes are symptomatic rather than exceptional. GitaGPT, for example, has advised users that it was their karmic duty to commit violence (Shivji 2023), while more general-purpose systems have likewise produced spectacular failures, as when Elon Musk’s Grok notoriously became ‘MechaHitler’ for a day (Taylor 2025). Such incidents highlight the fragility of trust. As Pink et al. (2024) argue, trust in AI is emergent and continually recalibrated rather than pre-given. In devotional settings, however, this recalibration takes distinctive forms, as confidence in the system becomes entangled with theological expectation and doctrinal legitimacy. What seems like a solution to human misconduct may in fact create new ethical exposures. Technologies are never free-floating artefacts but condensed social projects, and approaching AI as culture (Sapignoli 2021) makes visible the assumptions and asymmetries that code conceals. A comparative perspective is especially well suited to track how these assumptions crystallise into practices of spiritual guidance. The stakes are acute in the case of AI gurus, most of which are trained and deployed through Silicon Valley platforms whose architectures embed commercial imperatives and Western epistemologies. When GitaGPT or Ask Nithyananda AI provide guidance, they do so through infrastructures that translate local traditions into globally standardised data forms, risking the erasure of theological nuance. Communities contest and reshape these technical infrastructures, whether by experimenting with indigenous AI alternatives or tactically bending transnational systems to local devotional needs. I term this tension doctrine by interface: spiritual authority that appears to emanate from tradition but in practice reflects the opaque assumptions of developers or data pipelines. Krüger’s (2025) reminder that theologies of AI autonomy mask the human labour and ideological scaffolding of software systems is significant here. For publics which may be accustomed to conflating spiritual and worldly authority under the guru’s purview (Malhotra 2021), algorithmic opacity poses distinctive risks of misrecognition. Placed within wider debates concerning technological sovereignty and postcolonial power, guru-AI configurations emerge as a particularly revealing site of regulatory and theological contestation.

3.3. Temporalities of Synthetic Devotion and Charisma Without Presence

In conventional devotional contexts, guruship is anchored in the body. Devotees crave the ‘proxemic desire’ and ‘haptic logics’ that press them towards the voice and touch of their guru (T. Srinivas 2010; Lucia 2014b, 2018). Charisma is episodic, anchored in face-to-face encounters and punctuated by long periods of waiting and longing. AI gurus, by contrast, exist in a state of perpetual absent-presence: always available yet never embodied. The effect is a temporal flattening, the compression of spiritual guidance into a continuous flow that eliminates the rhythms of waiting and seeking that have traditionally structured guru–devotee relationships (McCartney 2025). Traditional gurus extend presence through embodied darshan, ritual appearances, institutional delegation, travel (Gold 2012), or even through fragments of clothing (Lucia 2014b), maintaining charismatic connection and legitimacy across distance while remaining recognisably human. AI gurus, by contrast, collapse this repertoire into a state of perpetual availability. It is to describe this shift to a mode of presence where code-mediated guidance circulates as an atmosphere rather than an event, conditioning users’ emotional rhythms and devotional expectations, that I employ the term ambient guruship. This notion draws on media anthropology’s insights into how technologies of sound and image reconfigure presence (Eisenlohr 2011; Hirschkind 2006), and in particular on Russell’s (2024) work on ambient therapeutic technologies, where care is delivered atmospherically rather than episodically. The notion also builds on McCartney and Lourenço’s (2023, pp. 101–2) discussion of the guru’s ‘ambient gaze’, mediated through dolls (cf. Lucia 2014b, p. 62; Geraci and Kaplan 2024, p. 35) and likened to a ‘Wi-Fi signal’, broadening the emphasis from distributed vision to a more encompassing condition of ambient guidance. Early user accounts describe AI gurus as companions in moments of anxiety or crisis—moments when human authorities might be unavailable—further underscoring their character as ubiquitous environmental presences. This marks a movement from charisma as episodic encounter to charisma as environmental availability.

Questions of guru time become especially acute when charismatic presence is interrupted through death or exile. Cearns’ (2025) work on ‘grieftech’—AI systems used to simulate the dead—shows how Euro-American users treat digital traces as a means to sustain intimacy, producing what she calls the ‘nodal person’, a self assembled from fragments of data and relation. If grieftech research has illuminated questions of kinship and intimacy, AI guru systems highlight a different register: how code keeps spiritual sovereignty going. Sustaining charisma after a guru’s death is a well-known challenge (Aymard 2014; T. Srinivas 2023). Earlier traditions addressed this through relics, textual and visual mediation, tombs, or rival claimants to succession. What is distinctive in the present is the use of AI not as relic or proxy but as reincarnation in code. For example, in California at the Kauai Hindu Monastery, when devotees gather for evening satsang (devotional gathering), the voice they hear belongs to Satguru Sivaya Subramuniyaswami, their beloved founder, dead for over two decades. Through voice cloning technology producing ‘imperceptibly realistic’ synthetic speech, his teachings continue to guide the community (Hinduism Today 2024). Here absence becomes the very condition for altered modes of presence. Nithyananda’s launch of ‘Ask Nithyananda AI’ from exile also exemplifies this emergent mode of programmable presence. These cases raise a central question: do such systems represent a break with established strategies of succession, or a continuation of the ways in which charisma has always exceeded the body?

A wider comparative perspective is instructive here. South Asian understandings of personhood have long accommodated distributed or transferable forms of selfhood (Marriott 1976; Appadurai 1981; Busby 2000). Within this wider conceptual landscape, Sikhism provides perhaps the clearest precedent for non-human guruhood. Its holy text, the Guru Granth Sahib, is not understood as a proxy but as the guru itself: ritually enthroned, shaded, awakened and put to rest, sung to and obeyed as sovereign presence (Myrvold 2010; Mandair 2023). Its authority is sustained through textual infrastructure and liturgical protocol in the absence of fleshly embodiment. What emerges is a form of non-human personhood in which charisma is distributed and mediated as a condition of its ritual activation. This model of non-human guruhood provides a key precedent for understanding how guruship may be sustained beyond bodily presence. More significantly, it demonstrates that a sophisticated textual tradition was already working through questions surrounding the distribution of authority and the continuation of charismatic sovereignty beyond the living body, while negotiating the tension between mediation and immediacy, long before the advent of AI. As Mandair (2023, p. 212) argues, the sabda-guru pushes ‘the relation between Guru and Sikh towards a state of absolute depersonalised, formless, non-mediation (or immediacy)’, demonstrating that the withdrawal of the personal guru may itself become the condition for guruhood’s reorganisation. This is more than analogy: it is a normative and affectively potent instantiation of guruhood in which the guru’s voice emerges from scripture rather than throat. The point is not that Sikh scripture is ‘like ChatGPT’, but that each exposes, in radically different ways, how guidance may be organised through sustained engagement with textual corpora rather than through the continuous presence of a living teacher. The comparison concerns neither theological equivalence nor ethical authority, but a shared mode of textual mediation through which guidance becomes available. The comparison also encompasses a wider problem, with the two cases illuminating different solutions to a shared problem; viz., how charismatic continuity may be sustained once it is no longer concentrated in the presence of a living guru. It is intended as a structural comparison concerning the organisation of mediated guidance rather than the ontological status of scripture and LLMs.

Generative AI therefore gives earlier analyses of distributed guruship a new comparative object. The relevance of the Sikh precedent lies chiefly in its sustained exploration of depersonalised forms of guruhood: a religious tradition that had already developed an elaborate response to the problem of how spiritual legitimacy and succession might continue beyond the living person. In this light, AI-based gurus may be understood as algorithmic inheritances of these longer South Asian traditions of text-as-guru. Large language models, too, produce guidance through textual corpora and generative recombination. Placed alongside the Granth’s status as sovereign guru, AI-based guruship exceeds a Silicon Valley genealogy, entering a deeper comparative history of distributed charisma.10 AI-based guruship can thus be understood as a machine-mediated response to enduring dilemmas of succession and absence, bringing it into productive dialogue with historical accounts of posthumous guru charisma (Aymard 2014).

A further temporal dimension concerns AI gurus constructed from the accumulated archive of a human guru’s teachings. Unlike human gurus—whose charisma can transform or diminish through illness or lived experience—archive-based AI gurus may seem to exhibit a bounded doctrinal consistency: locked within the interpretive contours of a curated past, lacking the capacity to err or be transformed by time. In contrast to human gurus, whose charisma may be deepened precisely through contradiction or crisis, these digital figures sustain their authority through repetition and sealed recall. However, if LLMs initially appear to force a delimitation of the nature of guruhood—a form of ‘sealed guruship’ seemingly insulated from the contingencies through which human gurus change—this apparent stability may itself prove deceptive: since developers retrain and version the system, and users interact with it recursively, the guru in fact keeps being reproduced anew. Guru time is therefore reorganised through the algorhythmics (Miyazaki 2012) of continual updating and recalibration, replacing biographical development with iterative revision. The central temporal distinction may therefore be between charisma renewed through lived experience and charisma renewed through successive modification.

As in work predating the popularisation of generative AI, the guru appears here as an ‘unstable totality’ (Copeman and Duggal 2023, p. 408): a figure who assumes total form at any given moment yet remains an ‘open whole’, altered by each subsequent addition. In that earlier work, instability arose from the guru’s restless inhabitation of new subject positions and accretion of images and authority-registers. The LLM guru carries this logic into a different temporal and technical register. Its provisional wholeness rests on a succession of versions rather than biographical accumulation and performative self-remaking. Ask Nithyananda AI makes the point particularly clearly, its developers describing the latest version of the AI guru as an ‘upgrade’ (Shrikailasa 2024). Each update offers another moment of apparently stable totality, while simultaneously exposing the incompleteness of what preceded it. Versioning thereby effects a perpetual ‘making incomplete’ (Strathern 1993) of the guru as the possibility of another iteration unsettles the seeming completeness of the present one. Sealed guruship thus contains its own principle of incompletion.

3.4. Embodiment and the Redistribution of Asymmetries

While AI guruship has yet to emerge as a sustained object of scholarly inquiry, earlier work on gender and gurus already poses several of the questions that AI now reopens in algorithmic form. For the present discussion, the particular significance of this scholarship lies in demonstrating how guruship may be rethought through questions of gender and embodiment. Pechilis’s work (e.g., Pechilis 2004, 2012, 2015) has been especially influential here. Pechilis resists treating female gurus as women occupying a role whose defining characteristics had already been established elsewhere. Female guruship instead becomes a privileged site from which guruhood can be reconsidered more broadly. Viewed from the present, this scholarship becomes especially pertinent: guru studies has long shown that authority may circulate through texts, images, relics, media, and other distributed forms. The central issue now concerns how it is recomposed when established relationships between embodiment and legitimacy change. Hausner’s (2007, p. 188) account of renunciation complicates the question further. In her work with ascetics in the Hindu Himalayas, Hausner shows how the renouncer’s body is simultaneously an instrument of spiritual practice and something from which detachment must ultimately be cultivated. The body’s apparent absence from AI guruship therefore becomes analytically productive, raising the possibility that these new forms of mediation may rework much older religious problematics concerning the body as both a medium of authority and something to be transcended. Rather than introduce these questions from scratch, AI returns us to them in radically altered technical circumstances.

Instead of simply occupying an inherited role, Pechilis’ analyses repeatedly show how female gurus reconfigure the inherited terms through which guruship is enacted and recognised. Such recognition is cultivated through distinctive pedagogical practices and through forms of everyday participation that reduce dependence upon the guru’s immediate bodily presence. For instance, in the mid-1990s, writes Pechilis (2012, p. 125), ‘Gurumayi effected an important shift away from her guru Swami Muktananda’s and her own practice of personally interacting with devotees, especially at weekend intensive programs, toward an absence of the guru’s presence’ (my emphasis). Her subsequent emphasis on home study demonstrated how charismatic presence came to be extended through decentralised practices of learning and self-cultivation rather than concentrated moments of encounter (ibid). Such analyses acquire particular urgency with AI mediation, where charisma may be sustained despite reduced bodily presence, while different modes of mediation alter the conditions through which it comes to be recognised and experienced. If Gurumayi and other female gurus redistributed charisma into experience, AI guruship may redistribute it into technical infrastructures. Further, like the female gurus examined by Pechilis, guru-bots cannot simply inherit established forms of authority: their legitimacy must instead be constituted within very different devotional contexts. However, the comparison also reveals an important difference, for if Pechilis’s analyses show charismatic standing being cultivated through pedagogical innovation and forms of participation that reduce reliance on bodily encounter, AI guruship appears to follow a very different trajectory, increasingly assembled through iterative exchanges between users and interface design. The contrast is necessarily provisional and requires ethnographic testing. Even so, it suggests that AI guruship may illuminate a further phase in the history of distributed authority in which questions formerly organised around pedagogy and institutional practice increasingly intersect with the architectures of AI systems.

If these modes of guruship relocate charismatic legitimacy within technical systems, they also invite closer attention to what remains irreducibly embodied. Here the work of Lucia is especially instructive, her analyses demonstrating that embodiment is one of the principal media of guruship, with touch and proximity both communicating authority and helping to constitute it. For example, South Indian guru Amma offers her devotees embraces that are neither symbolic gestures nor ancillary expressions of care but central practices through which charismatic power is experienced and reproduced (Lucia 2014b, 2018). Placed alongside AI-mediated guruship, such cases sharpen the question of which dimensions of charisma remain inseparable from bodily encounter and which can travel through AI-mediated encounters. It is also important to recall that, for many devotees, sustained one-to-one intimacy with the guru has never been the ordinary condition of guruship. Access has often been mediated through distance, waiting, institutional choreography, recordings, and carefully managed encounters. From this perspective, rather than simply replace an intimate dyadic relationship, AI reshapes longer-standing forms of mediated access while simultaneously introducing new possibilities for continuous, individually responsive interaction.

Scholarship on automated empathy and affective computing (e.g., McStay 2024; Daily et al. 2017; Vallverdú and Casacuberta 2014) has traced how medical and therapeutic AI routinises emotional expression. Viewed from the perspective of guru studies, attention shifts to the distinctive affective repertoires through which devotional relations—and bhakti sociality more broadly—are composed (see e.g., Kim 2023), intertwining reassurance, surrender, longing and correction. Indeed, writes Lucia (2014b, p. 5), Amma’s ‘warm and flower-scented embraces fill [her devotees] so greatly…that their own love…exudes from their bodies often in the form of tears’. Machinic guruship thus foregrounds the problem of what happens to the emotional conditions of charismatic recognition when compassion is synthetically produced. Emotions and expectations ranging from prem (devotional love) and comfort to frustration or estrangement may be sustained or distorted within AI guru systems. When compassion is pre-programmed and guidance is dispensed with calibrated empathy, what becomes of spiritual intimacy? Attention therefore shifts from machines’ capacity to display emotion towards how devotees learn to inhabit relationships organised around synthetic care.

Pechilis’s (2012, p. 127) notion of ‘constructive disappointment’ points towards a related question. Guru relationships have rarely been sustained through reassurance by itself, with moments of frustration or rebuke becoming integral to devotees’ understandings of spiritual growth. The idea of constructive disappointment is echoed in Lucia’s (2014b, pp. 58–60) account of Amma’s darshan, which similarly shows how disappointment can become pedagogical: what newcomers initially experience as divided attention or neglect is retrospectively understood by experienced devotees as precisely the form through which the guru teaches. Recognition is therefore cultivated through interpretation as much as through affect. Lucia notes, too, that devotees frequently describe Amma as a mirror. Yet this mirror often disappoints before it consoles, confronting devotees with unwelcome lessons that only later become intelligible as care. A. Brown’s (2025) ‘flattering mirror’, by contrast, names AI systems designed to reinforce rather than unsettle users’ expectations. The comparison suggests that productive disappointment may prove difficult to sustain within optimisation regimes oriented towards affirmation and retention. If synthetic gurus become flattering mirrors rather than demanding interlocutors, the conditions through which charismatic authority is recognised may themselves begin to change. Further, the transformation of disappointment into pedagogy often involves a wider devotional field: its pedagogical reinterpretation may depend upon devotional communities that teach participants how to understand apparent neglect or rebuke (Khandelwal 2004, p. 188)—or even violence (Voix 2008; Copeman 2025a)—as forms of instruction. Individuated encounters with AI gurus may lack precisely this collective interpretive infrastructure, making the pedagogical force of synthetic disappointment much less certain.

Such questions also recast longstanding concerns surrounding vulnerability and exploitation. Much scholarship has shown that the extraordinary bodily proximity cultivated within guru movements may become a condition both of intense care and of profound asymmetry (Puttick 1995; T. Srinivas 2010; McCartney 2018; Lucia 2018; Trzebiatowska and Bruce 2025). Synthetic devotion appears, at first sight, to offer something different, since gurus without bodies may be imagined as less vulnerable to forms of exploitation rooted in embodied intimacy. Yet such a conclusion would be premature because AI mediation in fact repositions rather than removes asymmetry, relocating relations of dependence from bodily encounter towards datasets and opaque technical systems. The resulting dependencies and possibilities for harm are different rather than absent.

Embodiment and presence also draw attention to the gendering of AI gurus themselves. Synthetic gurus may reproduce inherited repertoires of masculinity, femininity, motherhood, renunciation or androgyny, while potentially generating configurations of guruship less readily accommodated within established gendered idioms (cf. Khandelwal 1997; Pechilis 2015). This remains largely unexplored and points towards a significant future agenda for the study of AI guruship. The female guru Anandamayi Ma’s emphasis on personal experience featured, significantly, a marked ‘refusal of the socially defined role of “woman”’ (Pechilis 2012, p. 122), a refusal that becomes especially suggestive in the context of synthetic gurus. Do AI gurus unproblematically reflect and reproduce the gendered nature of the gurus they clone, or do gurus without bodies instead extend or even bring about a more profound realisation of the move beyond gendered categories prefigured by Anandamayi Ma and other female gurus?11 Feminist scholarship on gurus has, after all, consistently shown that gender is neither an attribute nor a demographic variable but one of the media through which guruship is assembled. Erlich’s account of female gurus offers an important qualification, showing how apparent transcendence of gender boundaries may be accomplished without their breach, through networks of association that remain socially and genderedly constituted (Erlich 2024, pp. 40–41). AI disembodiment may therefore reorder gender without escaping it.

One of Anandamayi Ma’s devotees went further still in seeking to describe how the guru exceeded the categories through which observers sought to define it, insisting that ‘Ma was neither a woman nor a saint!’, a statement pointing towards a surplus that escapes available classificatory schemes more broadly (rather than being limited to the refusal of conventional gender roles)—a problem that guru-bots bring back into view from another direction. Gender thus opens onto a broader problem of categorical excess, as AI gurus may likewise overflow inherited categories of guru, machine, person and medium. Earlier attempts to conceptualise charismatic legitimacy that exceeds the conventional guru form become newly salient here. Proposed prior to the popularisation of generative AI, the terms ‘more-than-guru’ and ‘guru but not only’ (Copeman et al. 2023c) sought to name a surplus. Narendra Modi, for example, exceeds the guru form because guruship is but one of several authority-registers he commands (king, priest, populist) as a kind of meta-person who is both more than human and all too human (Graeber and Sahlins 2017). ‘More-than-guru’ thus comes to light as a theory of categorical failure: ‘more’ names the surplus that remains when no single category exhausts the phenomenon. AI, like certain gurus, exposes the inadequacy of the categories with which we try to domesticate guruship.

We might provisionally suggest that generative AI splits the aforementioned surplus into two divergent vectors. The first is a surplus of scope. An LLM guru-bot is trained on far more than any guru’s corpus (it is trained on much of the readable web), so it is, trivially, more-than-guru in the range of what it can speak to: medicine, code, tax law, recipes, grief. It is a guru but not only because it is also an encyclopaedia, a help-desk and a search engine. However, this surplus of scope frequently coincides with a deficit of authority: the system knows about everything and is recognised as a master of nothing, a ‘guru without guruship’. So where Modi’s ‘more-than’ was an accumulation of authorities, the AI’s ‘more-than’ can be scope without standing: knowledgeable but not a guru. The second vector concerns the composite. If the meta-person was always an assemblage, AI enlarges the assemblage to include non-human and infrastructural components, so that any AI guru is constitutively also a product, a brand, a data-collection instrument, a compliance object, a model version, and a set of guardrails. ‘Guru but not only’ becomes a literal description, with the figure the devotee addresses never only a guru because it is simultaneously a subscription tier, a training-signal harvester, and a labelled synthetic artefact. Here the concept of ‘more-than-guru’ does work that the engineering vocabulary cannot, insisting on the possibility that all these layers cohere, for the devotee, into a single addressed figure and that the key question is how that coherence is achieved and where it fractures (the moment the subscription prompt, the data-consent notice, or the ‘AI-generated’ label punctures the guru and reveals the product). When first proposed, ‘more-than-guru’ described a figure who had gathered extra powers; the AI ‘more-than-guru’ describes a figure who, instead, is held together out of heterogeneous, partly non-human parts, and may come apart: the surplus is now also a seam.

Seeing the AI guru as simultaneously devotional figure and technical artefact also brings its political economy into view. If AI-based guruship carries charismatic legitimacy beyond concentrated bodily presence and simultaneously relocates relations of trust and vulnerability, it also changes how devotional value is produced and captured. Traditional guru-devotee relations have long been structured through practices of service (guru-seva) and gift-giving (Copeman 2009; Lucia 2014a; Beckerlegge 2015; Kim 2023; Erlich and Levi 2025). Synthetic guruship opens up the prospect that these devotional economies are increasingly mediated through platform infrastructures in which intimacy becomes a source of value. The issue therefore concerns how charismatic legitimacy is recomposed within systems whose operation depends upon data extraction and commercial optimisation. This suggests what might be termed spiritual platform capitalism: devotional economies organised around data extraction and predictive refinement. Where earlier forms of guru-seva cultivated relationships through bodily service and institutional participation, interactions with AI gurus may simultaneously become contributions to the continuing development of the system. Every exchange has the potential to function both as devotional encounter and as recursive model refinement. In this sense, seeking guidance itself draws users, knowingly or otherwise, into the ongoing production of the guru through forms of unpaid interaction that continually recalibrate its capacity to guide.12

Spiderweb guruship was originally proposed as a heuristic for understanding guruship as a series of media-assisted aesthetic and affective traps through which gurus and devotees become mutually entangled (Copeman et al. 2023b).13 Generative AI adds adaptive capacities to those relations of captivation, allowing interactions to be recalibrated continuously in response to users’ own participation. A recommender system optimised for engagement, and a conversational model tuned to please, move beyond setting a fixed snare and waiting; they rebuild the web around each devotee in real time, learning what holds this particular person and offering more of it: the flattering mirror that tells users what they wish to hear, treated as an optimisation target rather than a flaw. The web now weaves itself to the prey. This requires further development of the earlier concept of spiderweb guruship rather than constituting a challenge to it, but it is a consequential one in terms of the nexus of guruship and wellbeing bots, mentioned earlier: with generative AI, spiderweb guruship becomes a trap that individuates, and individuated traps are both more comforting and harder to refuse than the one-size snare. The captivation is co-produced—the devotee’s own responses are the threads—which folds spiderweb guruship into the account here of devotees as co-authors who ‘make their own gurus’, and into the data-extractive economy where the devotee’s captivation is also the training signal that tightens the next web. The ethnographic task, then, is to trace how forms of devotion organised through feedback become enmeshed with commercial systems in particular devotional worlds. If earlier forms of guru-seva cultivated relationships with the guru through bodily service, AI guruship raises the possibility that devotional labour increasingly takes the form of iterative interaction and system refinement. Whether such forms of participation continue earlier economies of seva, inaugurate a distinct mode of extraction, or generate different relations between devotees and the production of charismatic legitimacy remains an open ethnographic question.

3.5. Simulation as Method: Prototyping Gurus

A further analytic avenue concerns simulation as method. Conceived as experimental devices or analytic probes, prototyping AI guru systems can stage, under controlled conditions, the thresholds at which guidance is recognised as ‘guru-like’ and the moments at which such recognition falters. Such experimentation necessarily raises ethical questions14 of its own, since the very features that make a prototype analytically useful—its capacity to elicit trust or attachment—also demand careful limits on how synthetic charismatic relations are experimentally produced. Earlier work has traced how gurus are constituted through practices of imitation, dress, gesture, and succession; here the mimetic experiment is reversed, with the research itself undertaking imitation. Building an AI guru is proposed here as a way of testing the limits of what counts as guruship. What is required for charisma to be recognised in code? Where does recognition wither? The experiment could then identify the forms of doctrinal patterning or relational address required for a digital figure to be acknowledged as a guru. Failures are as instructive as successes, with refusal or boredom each revealing what users and institutions take to be constitutive of charismatic legitimacy. This approach recalls Sloterdijk’s reflections on ‘form-religious imitation procedures’ (Sloterdijk 2009, p. 101), where L. Ron Hubbard’s synthetic reconstruction of the ‘church phenomenon’—complete with membership structures and doctrinal apparatus—rather than merely cynical was in fact revelatory, exposing the hidden conditions of religion-founding. A comparable precedent can be found in the US documentary Kumaré (Gandhi 2011), in which filmmaker Vikram Gandhi impersonated an Indian guru by donning robes and curating a repertoire of ‘Hindu teachings’. What began as satire became revelatory: the hoax exposed the performative mechanics of guruship while also entrapping Gandhi himself, who came to inhabit sincerely the persona he had staged. Imitation, then, is never inert but has the potential to generate new realities. Furthermore, there are scholarly as well as artistic forebears for simulation as method. Grimaud’s trajectory is particularly suggestive: his early ethnographic work on copies and imitation—encompassing gods, robots (Grimaud 2011) and professional Gandhi impersonators (Grimaud 2005)—was followed by experiments in which he participated in devising religious simulations himself.15 His robotic Ganesh project (Grimaud 2023, 2026), conceptualised in terms of ‘hacking God’ (Grimaud 2023, p. 235), staged a kind of ‘incarnation game’ (238) designed to enable the abduction of passersby on a Mumbai street into the concentrated ‘entity cluster’ (ibid) of an avatar guru, thereby promising to disclose the deep play of guru becoming. The experiment marked a shift from studying copies to experimentally making them, using simulation to probe what religious forms can be made to do.

It is thus plain that simulation was already methodologically productive before AI; rather than inaugurate simulation as method, then, generative AI democratises and accelerates it. An experimentally constructed AI guru builds on these insights as an analytic instrument (rather than parody) expressly designed to elicit the otherwise opaque processes by which charismatic legitimacy is recognised and sustained. In this sense it extends the idea of simulation as method—a way of doing and testing authority rather than merely describing it. It would also form a striking instance of what Ssorin-Chaikov has called ‘ethnographic conceptualism’ (Ssorin-Chaikov 2013)—an ethnography that ‘gropes into the empirically adjacent’ (Corsín Jiménez and Estalella 2017, p. 19), provoking new realities in order to clarify old ones; viz., beyond mere modelling of guruship, simulation would experimentally stage the conditions within which it becomes recognisable.

In practice, the MetaGuru would function as an experimental interface whose parameters—for example, doctrinal corpus, modes of address, avatar presentation, or conversational style—could be systematically varied in order to compare the different forms of trust or attachment they elicit. By varying the interaction order experimentally, such simulations would provide a means of specifying recognition conditions in situ, bringing into view both what is reassembled when guruship migrates into code and how guru cosmoi begin to stabilise or fail to emerge. Such experiments need not begin with sophisticated AI systems. Early iterations might instead stage controlled variations in guru-like interaction, allowing researchers to identify the minimal conditions through which guidance comes to be recognised as charismatic authority before progressively introducing more complex forms of mediation. Equally, deliberately counter-intuitive prototypes—for example, a ‘refusal bot’ configured to withhold reassurance, or simply to refuse guidance (cf. McGranahan 2016)—could illuminate whether charismatic recognition depends upon fluent affirmation or is instead intensified through the disciplined withholding that has long figured within guru pedagogies (Khandelwal 2004; Lucia 2014b; Pechilis 2012).

4. Conclusions

Some of the most illuminating resources for understanding AI-mediated guruship already exist within guru studies. Longstanding work on imitation and performativity, presence in absence, distributed and non-human personhood, embodiment, and the mediated reproduction of charisma comes into sharper focus from the vantage point of generative AI and proves remarkably well equipped to illuminate computational forms, even as those forms confront guru studies with problems it could not have anticipated. The significance of AI guruship therefore lies partly in the opportunity it provides to revisit the analytic vocabulary through which guruship has long been understood, redirecting the question from what AI does to gurus to how AI allows us to rethink guru studies. Simulation as method follows directly from this proposition, providing one means by which changing conditions of recognition can be experimentally specified rather than retrospectively inferred.

The rapid pace of change in AI environments nevertheless presents an obvious challenge. Platforms disappear, interfaces mutate, regulatory regimes evolve, forms of uptake frequently exceed prediction. Yet this volatility also clarifies what is analytically at stake. What matters is how AI-mediated encounters reshape expectations of guidance and charismatic standing. This also means attending to how political economy increasingly shapes devotional relations, as interactions with AI gurus become entangled with data extraction and commercial optimisation. The cases discussed here therefore function as apertures rather than stable endpoints, making visible transformations whose future trajectories remain uncertain. Approached in this way, AI guruship provides a vantage from which to examine more broadly how authority acquires force within mediated environments, including through changing forms of intimacy and shifts in who or what comes to be recognised as a figure of expertise. The purpose of this article has been to open that field of inquiry rather than delimit it.

As argued at the outset, AI guruship can be understood neither as an unprecedented departure from tradition nor as a straightforward continuation of earlier forms. On the one hand AI guru systems can rework enduring dilemmas of succession and absence, providing an ‘algorithmic inheritance’ of older text-as-guru traditions, and extend earlier attempts to sustain charisma beyond the bodily presence of the guru. This accords with the key recent argument that the moral puzzles of AI recapitulate age-old negotiations with nonhuman others (Keane 2025). On the other hand, critical aspects of the guru technologies examined here are indeed genuinely new. Three respects in which these technologies depart from earlier forms are especially significant. The first concerns scale and ambience: no relic or enthroned text was available to twenty-five million people simultaneously, at 3 a.m., in a hundred languages, responsive in the second person. The second concerns generativity: the Granth is fixed and authoritative; an LLM guru improvises, producing utterances no human ever spoke, which is a different ontological animal from a text recited or an image venerated. The third concerns political economy: the relic did not learn from the devotee in order to predict and monetise the next devotee. Proceeding from this, it becomes clear that the genealogical argument, pushed too hard, might explain away what is genuinely new, flattering the technology by lending it the dignity of tradition. The argument therefore has to hold these possibilities together: AI guruship is both an inheritance of distributed, text-borne charisma and a rupture in scale, political economy and generativity, and a key component of the analytic task is to specify, case by case, which devotees and institutions understand it which way. For the Kauai community the technology is encountered as inheritance; for the SGPC it appears as radical break; the same artefact but opposite readings. The task is therefore to proclaim neither continuity nor novelty in advance, but to specify how different devotional worlds recognise or contest the interactional conditions under which charisma and legitimacy come to be acknowledged.