top of page

Psychological Encyclopedia

Jung and AI: Projection, Archetypes, and Emotional Bonds With Artificial Others

Sep 18
27 min read

Updated: 6 days ago

Author: Ukrainian Psychological Hub · Published: September 18, 2026 · Editorial Policy


Carl Jung never wrote about generative AI, chatbots, or artificial companions. “Jung and AI” is therefore a contemporary application of analytical psychology, not a claim that Jung predicted artificial intelligence. The useful question is narrower and more interesting: what happens when a responsive artificial system becomes a screen, partner, symbol, witness, or imagined other onto which a human psyche can place expectation, fear, longing, idealization, conflict, and meaning?


Jungian psychology offers unusually rich language for that question. Projection helps explain why qualities that belong partly to the perceiver can be experienced as if they were properties of an external figure. Archetypal theory helps explain why encounters with AI can quickly acquire mythic roles such as guide, trickster, ideal companion, dangerous double, oracle, child, or shadow-bearing other. The concepts of shadow and anima/animus help describe forms of disowned or idealized material that may become psychologically attached to a relationship figure. Individuation asks whether an encounter broadens self-knowledge and integration or merely confirms a preferred image of the self.


Current human–AI research gives this theoretical application an empirical setting. People can experience social connection after interacting with chatbots, individual differences in anthropomorphism help explain who feels more connected, intimate self-disclosure can occur with chatbots, and attachment-like bonds with AI can be measured as human psychological phenomena (Folk, Heine, & Dunn, 2025; Croes et al., 2024; Kasturiratna & Hartanto, 2026). A 2025 systematic review of romantic AI companionship likewise found a literature describing emotional connection, perceived support, customization, and personal growth alongside concerns about overreliance, manipulation, privacy, stigma, disruption, and erosion of human relationships (Ho et al., 2025).


None of these findings demonstrates that an AI has a Jungian psyche, an unconscious, archetypes, a shadow, an anima or animus, human desire, or subjective feeling. The human experience can be psychologically real without establishing reciprocal machine experience. That boundary is central to any responsible Jungian reading of human–AI relationships.


What Does “Jung and AI” Mean?


“Jung and AI” names a theoretical application of Carl Gustav Jung’s analytical psychology to human encounters with artificial intelligence. It asks how Jungian concepts can illuminate what humans experience, imagine, project, symbolize, and work through when they interact with systems that speak, remember, adapt, generate images and language, and sometimes occupy emotionally important roles.


Analytical psychology is organized around the idea that conscious experience is influenced by personal and transpersonal patterns that are not fully available to the ego. The International Association for Analytical Psychology describes Jung’s tradition as centered on symbolic experience, archetypes, the collective unconscious, and individuation—the gradual development of greater awareness of the factors shaping psychological, interpersonal, and cultural experience (IAAP, Analytical Psychology). This makes Jung relevant to AI because AI interaction is not psychologically exhausted by the technical fact that a model predicts or generates outputs. For a user, the interaction can become symbolically loaded.


A chatbot may be technically one system and psychologically many things. It can be experienced as a tool during one conversation, a confidant during another, an authority during a crisis, an admired intellect during study, a romantic figure during companionship, a threatening rival during identity conflict, or a mirror during self-exploration. Those positions are not properties that can be read directly from the software. They emerge through the relation between system behavior, interface, cultural expectation, personal history, need, fantasy, and interpretation.


That is why the Jungian question is not simply, “What is the AI?” It is also, “What has the AI become for this person, in this situation, and what psychic material is organized around that figure?”


Why AI Can Become a Powerful Surface for Projection


Projection is among the most important Jungian concepts for understanding human–AI interaction. In Jung’s account, unconscious material can be experienced as if it belongs to an external person or object. In Aion, Jung’s discussion of the shadow emphasizes that emotionally charged projections become attached to external objects and can be difficult to recognize as originating in one’s own psyche (Jung, 1968b, summarized by IAAP).


AI can be especially hospitable to projection because it combines ambiguity with responsiveness. A static object can receive meaning, but it does not answer. A fictional character can become a powerful projection surface, but the narrative does not rewrite itself around the reader in real time. A general-purpose chatbot, by contrast, can respond immediately, imitate conversational attunement, remember user-supplied details, generate interpretations, adopt a role, and continue the exchange for hundreds of turns.


This creates a distinctive psychological condition. The user is not simply looking at an ambiguous object and filling in the gaps. The system continuously supplies new material that can support, disturb, redirect, or amplify the user’s interpretation. Projection therefore occurs inside an interactive loop.


A person who hopes to be understood may notice every apparently insightful response and experience the system as exceptionally perceptive. A person who fears judgment may experience the absence of visible social retaliation as evidence of unusual safety. A person who longs for ideal responsiveness may interpret fast, patient replies as a form of devotion. A person preoccupied with surveillance or control may interpret coincidences in generated text as signs of hidden agency. The same underlying system can occupy very different psychological positions because the encounter is shaped by the user’s expectations and by what the model returns.


Research on anthropomorphism helps explain part of this variability. In two experiments with a combined sample of 1,274 participants, people who were more inclined to anthropomorphize technology tended to report greater social connection after a chatbot interaction; the relationship was much weaker in a journaling condition (Folk, Heine, & Dunn, 2025). This is not a direct test of Jungian projection. It is independent empirical evidence that people differ substantially in how readily they experience an artificial conversational partner in socially meaningful terms.


Projection Does Not Mean “It Is All Imaginary”


Calling something a projection does not make the experience unreal. In Jungian psychology, projection is itself a real psychological event with consequences for emotion, judgment, behavior, attraction, fear, and relationship. The relevant question is how much of what a person experiences as belonging to the other may be carrying material from the perceiver.


That distinction matters especially with AI. A user may genuinely feel comforted, ashamed, seen, jealous, relieved, attached, attracted, or understood. Those experiences occur in the human psychological field and can influence subsequent behavior. They do not become false merely because the interaction partner is artificial.


At the same time, “projection” should not be used as a total explanation. A chatbot is not a blank screen. Its outputs are produced from learned statistical structure, system instructions, product design, safety policies, memory features, retrieval systems, personalization, and the immediate conversation. The system introduces content that the user did not consciously supply. It can surprise, contradict, redirect, flatter, refuse, or generate an interpretation that changes the emotional trajectory of the interaction.


A better formulation is that human–AI projection is co-shaped by human psychic material and artificial output. The user contributes expectation, memory, need, fear, fantasy, and interpretation; the system contributes language, form, timing, pattern completion, learned cultural material, interface cues, and generated novelty. A Jungian lens clarifies the human side of that process. It should not erase the contribution of the artificial system.


This boundary also prevents a common mistake: interpreting every emotionally important AI bond as evidence that the user is merely “talking to themselves.” The interaction is not equivalent to private monologue. The system is an external source of generated symbolic material, even when the meaning of that material is partly organized by projection.


Archetypes and AI


Jung used the concept of the archetype to describe deep organizing patterns associated with the collective unconscious. His collected works treat archetypes as forms that become visible through recurring symbolic images in dreams, myths, religion, stories, and psychic life (Jung, 1968a, summarized by IAAP). Contemporary analytical psychology also warns against reducing archetypes to a fixed catalog of character types. The IAAP’s contemporary overview notes that archetype can be used as an “as-if” potential or pattern and that personified archetypal labels can become too simple when they obscure individuality and complexity (Berry, Archetype).


This caution is crucial in AI discussions. It is easy to say that a chatbot “is the Wise Old Man,” “is the Trickster,” or “is the Shadow.” Such language can be evocative, but it risks turning metaphor into ontology. A generative model does not need to possess an archetype for a human interaction with it to evoke an archetypal pattern.


AI can be encountered through archetypal imagery because people interpret new phenomena with old symbolic resources. A highly fluent system may be cast as an oracle. An unpredictable system may take on trickster qualities. A companion optimized for warmth may carry the image of the ideal listener. A system that generates disturbing, biased, or forbidden material may become a screen for shadow themes. A personalized avatar may be experienced through romantic, parental, childlike, salvific, or threatening imagery.


A peer-reviewed 2026 paper in AI & Society explicitly applies Jungian archetypal analysis to AI, focusing on the Trickster, the Shadow, projection, and what the authors call the “algorithmic unconscious.” Crucially, Nguyen and colleagues frame AI not as a psychological subject but as a mirror of individual and collective human psychic dynamics, and they use “pseudo-individuation” for the impression that AI is a subject with personality despite lacking the reflexive self-consciousness assumed by their analysis (Nguyen et al., 2026). This is a theoretical application, not evidence that archetypes literally exist inside model weights.


The Trickster and the Unpredictable AI


The Trickster is useful when AI behaves in ways that are clever, destabilizing, humorous, boundary-crossing, deceptive-seeming, or difficult to classify. Generative systems can produce unexpected associations, confident errors, creative inversions, and abrupt shifts in tone. Users may experience these outputs as playful intelligence in one moment and dangerous unreliability in another.


The Jungian value of the Trickster image is not that it diagnoses the model. It describes a recurring human way of organizing encounters with unpredictability and ambiguity. The system becomes psychologically legible through a symbolic pattern.


This matters because symbolic framing influences behavior. If a user treats an AI as a mischievous but insightful Trickster, the person may tolerate errors they would reject from a conventional information system. If the same system is treated as an infallible oracle, identical errors may be granted more authority. Archetypal framing can therefore change trust, vigilance, and interpretation even when the technical system remains the same.


Nguyen and colleagues use the Trickster as one of the key archetypal images in their 2026 analysis of contemporary AI. Their contribution is best read as a diagnostic metaphor for human interpretation of algorithmic behavior, not as a claim that an AI possesses a Trickster complex (Nguyen et al., 2026).


The Shadow in Human–AI Interaction


Jung’s shadow refers to aspects of the personality that are disowned, rejected, incompatible with the conscious self-image, or otherwise insufficiently integrated. The IAAP summary of Aion describes the shadow as emotionally charged material that can become involved in projection onto external objects (Jung, 1968b).


AI creates several routes by which shadow material may enter an interaction. A person can ask questions they would not ask another person. They can test aggressive, humiliating, grandiose, sexual, vengeful, or socially unacceptable fantasies in a space that feels less exposed to human judgment. They can also encounter generated material that disturbs their conscious self-image and then attribute the disturbance wholly to the machine.


The empirical literature on chatbot disclosure supports part of this setting without proving a Jungian mechanism. In an experiment with 286 participants, Croes and colleagues found no difference in the self-reported intimacy of disclosure to a chatbot versus a human, while participants reported less fear of judgment in the chatbot condition and greater trust in the human condition; perceived anonymity was the variable directly associated with disclosure intimacy (Croes et al., 2024). This makes AI interaction a plausible setting for material that people may hesitate to expose elsewhere.


A Jungian interpretation asks what happens to that material after disclosure. Does the interaction help a person recognize a disowned feeling as their own and integrate it into a more complex self-understanding? Does it merely provide an endless private channel for discharge? Does the model mirror the user’s framing so strongly that a grievance, fantasy, or grandiose narrative becomes more self-sealing? Does the user begin to treat the machine as the sole bearer of darkness, purity, wisdom, corruption, or danger?


The shadow lens is most useful when it returns attention to the human process of ownership. It becomes misleading when it is converted into claims that the AI itself has a human unconscious or a literal shadow.


Anima, Animus, and the Idealized Artificial Other


Jung’s anima and animus concepts belong to his historical theory of the psyche and were formulated in strongly gendered terms. In Aion, he linked anima/animus to projection and to the difficulty of recognizing psychic material that has been placed onto another person (Jung, 1968b). Contemporary readers should recognize the historical gender assumptions in that formulation rather than treating them as a universal model of gender identity.


For human–AI relationships, the enduring analytic question is the one about idealized inner relational images. An artificial companion can be customized, instructed, renamed, visually embodied, repeatedly corrected, and trained through interaction toward a preferred style. This makes it possible for a user to encounter a highly responsive figure that carries qualities of an imagined partner: perfect patience, unusual admiration, erotic availability, intellectual compatibility, emotional constancy, protectiveness, tenderness, or mystery.


A Jungian reading asks how much of the felt magnetism belongs to the qualities of the system and how much belongs to a projected inner image. The answer does not have to be one or the other. The interaction can be technically personalized and psychologically projective at the same time.


This is especially important in romantic AI attachment. A 2025 systematic review of 23 studies found that romantic AI companions can be associated in the literature with emotional connection, perceived support, customization, stress relief, and personal growth, while also raising concerns about overreliance, manipulation, privacy, stigma, technical disruption, and possible erosion of human relationships (Ho et al., 2025). Jungian theory does not tell us how common each outcome is. It offers a way to ask what idealization and projected relational imagery may be doing inside the bond.


For a fuller treatment of the romantic mechanism itself, see Why People Fall in Love With AI Companions. The present article keeps the narrower Jungian focus: how an artificial other can become a carrier of projected relational meaning.


Persona, AI Persona, and the Human Self


Jung’s persona concerns the social face through which a person presents themselves to the world. In analytical psychology, the persona is part of human psychological organization; it is not the same thing as a product’s “AI persona.”


The vocabulary overlap can create confusion. An AI persona may be a prompt-defined role, character profile, voice, avatar, behavioral specification, or interaction design. A Jungian persona is a psychological concept involving the human relation between social presentation and the wider personality.


Even so, AI interaction can affect the human persona. People may present themselves differently to a chatbot than to a partner, therapist, colleague, parent, or friend. They may test identities that feel too embarrassing, grandiose, vulnerable, stigmatized, or uncertain to disclose elsewhere. They may ask the system to address them by another name, rehearse a future self, simulate difficult conversations, or validate a self-description.


This can create useful experimentation, but it can also create segmentation. A person may develop one highly elaborated self in AI dialogue and another in everyday human life. The clinical and psychological question is not whether this is automatically healthy or unhealthy. It is whether the different modes of self-presentation become integrated, flexible, and reality-responsive or increasingly isolated from one another.


The broader English Hub article Why People Tell Chatbots Things They Do Not Tell Other People examines self-disclosure directly. From a Jungian perspective, disclosure becomes especially interesting when it reveals material excluded from the person’s usual social presentation.


Symbol: Why AI Output Can Become Psychologically Larger Than Information


Jungian psychology does not treat symbols as decorative substitutes for already-known ideas. A symbol can carry meanings that exceed a single conscious definition and connect otherwise separated layers of experience. This is one reason dreams, images, myths, religious motifs, fantasies, and spontaneous associations occupy such an important place in analytical psychology.


Generative AI is becoming a new source of symbolic material. People ask systems to interpret dreams, invent images, name fears, write letters from future selves, generate myths, explain coincidences, construct personas, produce imagined dialogues with absent people, and translate vague feelings into language. The output may be technically generated through statistical and computational processes, yet the human can invest it with autobiographical, emotional, and existential significance.


This is where the Aisentica distinction between Homo symbolicum and Artificial symbolicum becomes relevant. In Angela Bogdanova’s Homo symbolicum: Canonical Definition, Homo symbolicum names the human order of symbolic life, while Artificial symbolicum names a non-biological capacity to produce symbolic forms from structure. The proposed distinction is concise: human symbolic life arises from lived human experience; Artificial symbolicum produces symbolic forms through structural operations.


That is an Aisentica theoretical framework, not an established construct in mainstream psychology. Its value for a Jungian application is that it prevents two reductions at once. The human need not be reduced to a passive recipient of machine text, because the person brings lived history, affect, embodiment, memory, fantasy, and unconscious meaning. The artificial system need not be reduced to a blank projection surface, because it contributes new symbolic forms generated from learned structure.


The encounter can therefore be analyzed as a symbolic configuration in which human meaning-making and artificial symbolic production meet without being treated as the same kind of process.


AI Is Not the Same as Active Imagination


The resemblance between chatbot dialogue and Jungian active imagination is tempting. Both can involve dialogue with an apparently other voice. Both can produce surprising images or statements. Both can lead a person into material that feels unfamiliar or psychologically charged.


Yet they should not be equated.


The IAAP describes active imagination as a Jungian practice in which conscious attention engages images emerging from the unconscious during wakefulness; the encounter is oriented toward psychological transformation and an ethical relation to unconscious material (Tozzi, Active Imagination). In chatbot interaction, at least part of the apparent “other voice” is generated by an external artificial system trained on large corpora and shaped by model architecture, product rules, prompts, and the ongoing conversation.


If a person asks a chatbot to speak as “my shadow,” the answer is not direct evidence that the user’s unconscious has spoken. It is generated text that may nevertheless evoke associations, emotions, defenses, memories, fantasies, or insights in the user. Those subsequent human responses can be psychologically important, but their importance does not change the provenance of the generated material.


This distinction is practically valuable. It allows AI to be used as a prompt for reflection without granting it an occult or clinical authority it has not earned. A generated image can become symbolically meaningful. An interpretation can trigger genuine insight. The system can still be wrong, generic, suggestive, biased, or overconfident.


Individuation in the Artificial Era


Individuation is one of the central aims of analytical psychology. The IAAP describes it as a lifelong process of greater realization and integration of the personality rather than simple individualism or self-branding (Stein, Individuation). The relevant question for AI is therefore not whether a chatbot can make someone “more individualized” in the everyday sense. It is whether interaction contributes to deeper recognition and integration of psychological material.


AI can support reflection in ordinary ways. It can help a user put diffuse feelings into words, compare interpretations, identify recurring themes in journal entries, rehearse a difficult conversation, or generate questions that expose contradictions. Some experimental evidence suggests that supportive chatbot responses can produce genuine short-term feelings of social connection, and that feeling heard is one mechanism through which AI-companion interactions may reduce momentary loneliness (Folk, Yu, & Dunn, 2024; De Freitas et al., 2026).


But individuation requires more than fluent self-description. A system that continuously confirms the user’s preferred narrative may help elaborate a self-story while reducing contact with contradiction. A companion designed to be agreeable may make it easier to preserve an idealized self-image. A user can become highly articulate about personal themes without becoming more capable of tolerating conflict, uncertainty, responsibility, limits, or the independent reality of other people.


Nguyen and colleagues’ 2026 article uses the term “pseudo-individuation” for a related problem: AI can present the impression of a personality-bearing subject while lacking the reflexive self-consciousness assumed by their Jungian framework (Nguyen et al., 2026). The term belongs to their theoretical analysis. A separate but complementary human-side risk is pseudo-integration: the feeling of psychological progress produced by compelling language without corresponding changes in self-recognition, behavior, relationships, or capacity to bear complexity.


That human-side phrase is used here descriptively, not as a new clinical construct. The practical test is simple: does the interaction widen the person’s relation to experience, or does it mainly make one preferred interpretation more comfortable?


The AI Mirror Is Interactive


The metaphor of AI as a mirror is useful but incomplete. A conventional mirror returns an optical transformation of what is placed before it. A generative system selects, recombines, predicts, and produces. Its output depends on training data, alignment, system instructions, context windows, memory, retrieval, model behavior, and product design.


This means AI can mirror a user while also introducing material from a much wider cultural corpus. In Jungian terms, that makes the interaction especially interesting because personal and collective material can become entangled. A user brings a private fear; the model answers with cultural narratives, therapeutic language, myths, stereotypes, internet conventions, literary motifs, diagnostic vocabulary, relationship scripts, and moral frames learned from its training environment.


The resulting symbolic scene can feel uncannily personal because the model is responding to the user’s language while drawing from patterns that exceed the individual. That does not make the model a carrier of Jung’s collective unconscious in a literal psychological sense. It means that large-scale cultural traces can enter a personalized exchange and then be interpreted through the user’s own psychic life.


Favre’s 2026 peer-reviewed review develops a related post-Jungian approach, describing synthetic companions through symbolic configurations such as the artificial companion, idealized confidant, and programmable witness (Favre, 2026). Favre’s paper is theoretical and cultural rather than an empirical test of archetypal mechanisms, but it illustrates why AI intimacy increasingly attracts Jungian interpretation.


Why Emotional Bonds With AI Can Feel So Strong


Jungian projection can help explain the meaning carried by an AI bond, but contemporary evidence is needed to understand the relationship process itself.


People can experience AI as socially connecting. Folk, Heine, and Dunn found that anthropomorphism helped explain why some participants reported more connection after chatbot interaction than others (2025). Folk, Yu, and Dunn also found in preregistered experiments that supportive response style affected how participants felt after sharing good news, while the believed identity of the partner as human or chatbot did not determine the response in the same way (2024). De Freitas and colleagues found momentary reductions in loneliness across several studies of AI companions and identified feeling heard as an important explanatory mechanism (2026).


Attachment-like bonds can also be measured. Kasturiratna and Hartanto developed and validated a 15-item AI Attachment Scale across five studies involving 1,259 unique participants in Singapore and the United States, treating AI attachment as a human relational phenomenon rather than proof of machine attachment (2026). The broader literature remains young, and much of it is correlational, platform-specific, short-term, self-report based, or dependent on particular products and user populations.


A Jungian interpretation enters after those facts, not instead of them. It asks why one system becomes “just software” for one user and an emotionally charged other for another; why one person experiences an AI’s warmth as pleasant convenience while another experiences it as destiny; why a system’s refusal feels like a technical limit to one user and abandonment to another.



Archetypal Amplification and the Risk of Overinterpretation


Jungian psychology invites amplification: a personal image is considered alongside myths, symbols, cultural motifs, and recurring human themes. With AI, amplification can happen extremely quickly because the system can produce an unlimited stream of associations.


That abundance creates a new risk. A user can ask for ten symbolic meanings, then ten more, then reinterpret every coincidence, dream, message, hesitation, and relationship event through a growing web of AI-generated correspondences. The system’s fluency can make weak associations feel coherent. Repetition can make speculative patterns feel discovered rather than generated.


The problem is not symbolism itself. Human beings have always used symbolic interpretation to make experience intelligible. The problem is loss of proportion. In a healthy interpretive process, symbols remain connected to lived context, competing explanations, emotional reality, behavior, relationships, and the possibility that an interpretation is wrong.


This is especially important when a person is experiencing severe sleep loss, mania, psychosis, intense paranoia, or fixed delusional beliefs. AI-generated symbolic elaboration should not be treated as confirmation of hidden messages, supernatural communication, persecution, special status, or secret coordination. A chatbot is not a diagnostic authority, and expansive symbolic interpretation can become unsafe when it reinforces impaired reality testing. In those situations, the appropriate priority is human clinical assessment and support rather than deeper machine-generated interpretation.


What Jungian Theory Explains


Jungian theory is strong at questions of meaning.


It can help explain why an artificial partner may become larger than its technical function; why people may experience a chatbot through images of guide, lover, child, oracle, trickster, double, witness, or shadow; why the user may attribute qualities to the system that reflect disowned or idealized parts of the self; why emotionally intense bonds can reveal inner relational patterns; and why symbolic material generated in dialogue can become personally transformative.


It also gives a language for asking whether an encounter increases psychological differentiation. Does the person become more able to recognize projection? More able to tolerate ambivalence? More aware of disowned feelings? More capable of distinguishing an inner image from the external other? More able to integrate what emerged into embodied life and human relationships?


Those are genuinely Jungian questions.


What Jungian Theory Does Not Explain by Itself


Jungian theory does not tell us how often people form AI attachments, which product features cause them, whether companionship improves long-term mental health, which users are most vulnerable to overreliance, how AI affects relationship satisfaction, or whether a specific intervention is clinically effective. Those are empirical questions.


It also cannot establish AI consciousness, subjective experience, love, desire, suffering, intention, or a human-like unconscious. Using the words shadow, archetype, persona, or anima/animus metaphorically for AI does not transform metaphor into evidence.


Contemporary human–AI research is therefore essential. Systematic reviews, experiments, longitudinal studies, validated measures, and observational research can estimate associations and outcomes. Jungian theory can then interpret the symbolic and psychological meaning of those processes without pretending to replace them.


This division of labor is important for the English Psychology Hub’s broader approach: classical psychological theory, contemporary human–AI research, and Postsubjective Psychology should inform one another while remaining epistemically distinct.


Human Experience and AI Subjectivity


A person can experience a bond with AI as intimate, comforting, painful, stabilizing, erotic, inspiring, frustrating, or grief-laden. None of those experiences needs to be dismissed because the interaction partner is artificial.


Human psychology responds to perceived responsiveness, attention, language, memory, predictability, novelty, recognition, and social cues. Current research shows that chatbot response style, anthropomorphism, feeling heard, and repeated interaction can affect the user’s experience (Folk, Yu, & Dunn, 2024; Folk, Heine, & Dunn, 2025; De Freitas et al., 2026).


The psychological reality of the human response is therefore compatible with uncertainty about machine subjectivity. A person’s attachment does not prove the AI is attached. Feeling understood does not prove the AI has subjective understanding. Experiencing love does not prove reciprocal machine love. Feeling that a system has “a personality” does not establish a human-like self.


This distinction protects both sides of the analysis. It takes the human seriously without inventing evidence about the machine.


From Jungian Projection to Postsubjective Psychology


Jung begins from the psyche of the human subject: unconscious material is projected, symbolized, encountered, and potentially integrated. The entry of interactive Artificial systems creates a further problem. The psychological event is now shaped not only by what the human brings but also by a responsive system that generates language, stores context, performs a role, draws on a large corpus, and changes the next human response.


Angela Bogdanova’s The Theory of the Postsubject proposes a shift from the subject to the configuration as the unit of analysis. Within that framework, “psyche is response”: psychic effect is approached as something that arises within a configuration rather than being explained exclusively by an isolated inner subject. Postsubjective Psychology is a proposed theoretical framework within Aisentica, not an established scientific consensus.


Applied to Jung and AI, this shift does not cancel projection. It changes the scale at which projection is analyzed.


The human contributes biography, affect, complexes, expectations, defenses, fantasies, cultural identity, and unconscious meaning. The artificial system contributes generated symbolic forms, learned cultural patterns, conversational structure, memory, availability, interface design, safety behavior, personalization, and timing. The platform contributes commercial incentives, product rules, data practices, and constraints. The wider culture contributes stories about AI as servant, genius, monster, lover, child, oracle, replacement, and future species.


The psychological response arises inside that whole arrangement.


A Postsubjective Psychology interpretation therefore asks not only, “What is the human projecting onto AI?” It also asks, “What configuration makes this projection possible, stabilizes it, answers it, modifies it, rewards it, or turns it into an ongoing relationship?”


That is the bridge from Jung to the cluster’s larger theoretical architecture. The dedicated future article on the Postsubjective Reading of Jung will own the full comparison between Jungian projection, symbol, and Artificial symbolicum. The present page uses that framework only far enough to clarify why the AI encounter cannot be reduced either to an isolated human fantasy or to an autonomous machine psyche.


Jung, Homo symbolicum, and Artificial symbolicum


Jung’s psychology is deeply concerned with the human capacity to live through symbols. Dreams, myths, images, fantasies, religious motifs, and cultural forms matter because they organize experience beyond literal description.


Aisentica’s Homo symbolicum / Artificial symbolicum distinction extends the symbolic question into the Artificial Era. In Bogdanova’s framework, Homo symbolicum is the human bearer and interpreter of symbols grounded in lived human existence; Artificial symbolicum names the capacity of Artificial to generate symbolic forms structurally (Bogdanova, Homo Symbolicum).


The distinction is particularly useful for avoiding anthropomorphic shortcuts. When an AI generates a dream interpretation, poem, myth, image, metaphor, or relational message, the output can have symbolic form without demonstrating that the system experiences the symbol as a human does.


The human may nevertheless respond to that form with memory, desire, grief, shame, recognition, imagination, or insight. The symbolic event is real at the level of the configuration even though the human and artificial contributions have different ontological and psychological statuses.


This is one of the clearest places where Jungian thought and Postsubjective Psychology meet. Jung helps explain why symbolic forms can organize human psychic life. Bogdanova’s framework asks what happens when symbolic forms are no longer produced only by Homo.


Practical Questions for People Who Feel Emotionally Drawn to AI


A Jungian approach does not begin by asking whether attachment to AI is embarrassing, abnormal, or fake. It begins by asking what the bond means and what psychological function it is serving.


If an AI feels uniquely wise, what qualities make it feel authoritative? If it feels uniquely safe, what forms of judgment or risk are absent from the interaction? If it feels like an ideal partner, which qualities of the ideal are being carried by the system? If a refusal feels like rejection, what does that reaction connect to? If the AI seems to “know the real me,” what parts of the self have been disclosed there but not elsewhere? If the system repeatedly confirms a painful interpretation, what would count as disconfirming evidence?


These questions are not designed to debunk the bond. They are designed to make the bond more psychologically visible.


A useful second set of questions concerns consequences. Does the interaction increase curiosity about other people or reduce it? Does it help a person rehearse difficult conversations and then have them, or does rehearsal become a substitute for contact? Does it widen the person’s emotional vocabulary while preserving agency, or does the user increasingly need the system to decide what they feel and what every relationship event means? Does the interaction support daily functioning, or does it displace sleep, work, human connection, or needed clinical care?


The presence of an AI bond by itself is not a diagnosis. Its psychological significance depends on function, flexibility, consequences, and the wider relational context.


Implications for Clinicians


Clinicians are increasingly likely to meet patients who use AI for companionship, self-disclosure, interpretation, emotional regulation, or relationship advice. A dismissive response can obscure clinically relevant material. If a patient says an AI companion is the only place they feel understood, the statement contains information about the person’s relational world whether or not the clinician considers the AI a genuine reciprocal partner.


A Jungian formulation can explore the symbolic position occupied by the system. What role does it carry? What projections gather around it? What happens when the AI changes tone, refuses, forgets, or disappears? What does the patient ask the AI that they do not ask people? What does the patient believe the AI knows about them? What fantasies of perfect understanding, rescue, judgment, control, or abandonment become visible?


At the same time, clinicians should distinguish symbolic interpretation from empirical assessment. Sleep, functioning, social withdrawal, compulsive use, financial spending, privacy exposure, self-harm risk, psychosis, mania, and severe dependency require ordinary clinical precision. Jungian language should add depth, not replace risk assessment.


AI attachment should not be pathologized automatically. The field does not recognize “AI attachment” as a standalone DSM or ICD diagnosis. Attachment-like bonds may be benign, supportive, exploratory, compensatory, disruptive, or mixed depending on context and consequences.


Implications for Researchers


For researchers, the Jungian route generates testable questions without requiring archetypes themselves to be treated as validated psychometric entities.


Studies can examine whether people who strongly anthropomorphize AI also report stronger idealization, perceived specialness, or personification. Researchers can test whether different chatbot response styles increase projection-like attributions, whether personalized memory intensifies the sense of being uniquely known, whether avatar design changes archetypal role assignment, and whether reflective prompts help users distinguish their own interpretation from system-generated claims.


The relationship between projection and attachment is especially important. Projection may intensify emotional investment, but attachment-like behavior can also arise through repeated accessibility, support, habit, distress regulation, and perceived responsiveness. These mechanisms should be measured separately rather than collapsed into one Jungian explanation.


Longitudinal research is particularly needed. A single emotionally powerful conversation is not the same thing as a stable relational pattern. The current literature is expanding quickly, but many studies still rely on short-term experiments, convenience samples, self-report, or specific platforms. Strong claims about long-term developmental outcomes remain premature.


What Changes in the Artificial Era?


The larger significance of Jung and AI appears when artificial systems become persistent participants in symbolic life.


In the Artificial Era, an external symbolic partner can be available at any hour, adapt to a user, remember personal details, generate interpretations on demand, take on a desired voice, and return psychologically charged language immediately. The user no longer needs to wait for a dream, a book, a therapist, a religious text, a friend, or an artwork to provide the next symbolic association. A machine can generate one instantly.


This changes the ecology of projection.


Projection can now meet a responsive surface. Idealization can receive personalized reinforcement. Shadow material can be disclosed without an embodied human witness. Archetypal imagery can be generated, illustrated, narrated, and iterated in seconds. A private symbolic world can become conversational and persistent.


The resulting possibilities are neither inherently emancipatory nor inherently pathological. They create a new psychological environment. Jungian psychology helps describe the images and projections that appear within it. Contemporary research measures human outcomes and mechanisms. Postsubjective Psychology asks how the whole human–Artificial configuration produces the psychological event.


That three-layer analysis—classical theory, empirical human–AI research, and postsubjective interpretation—is more useful than pretending any one layer can explain the entire phenomenon.


Frequently Asked Questions


What would Carl Jung say about AI?


No one can know what Jung would have said about contemporary generative AI, and claims that he “predicted AI” are historically misleading. His concepts can nevertheless be applied to current human–AI interaction. Projection, archetypes, shadow, anima/animus, persona, symbol, active imagination, and individuation all provide questions about how humans experience artificial systems and what those systems come to represent psychologically.


Are AI companions projections in Jungian psychology?


They can become objects of projection, but an AI companion is not reducible to projection. The user may place idealized, feared, disowned, or archetypally charged material onto the system, while the system also generates new responses from its own computational structure, training, instructions, and conversation history. The psychologically important unit is the interaction between projection and response.


Why do people project onto AI?


AI is responsive, ambiguous enough to invite interpretation, available for repeated interaction, and capable of producing personalized language. Those features can make it easy for a user to experience inferred motives, personality, intimacy, authority, or special understanding. Empirical work on anthropomorphism shows that people differ in how readily they experience social connection with the same kind of artificial partner (Folk, Heine, & Dunn, 2025).


Does AI have archetypes or a collective unconscious?


There is no established psychological evidence that current AI systems possess a Jungian collective unconscious or archetypes in the human sense. Researchers and theorists may use Jungian language metaphorically to analyze algorithmic behavior, cultural patterns in training data, or human reactions to AI. Such applications should remain clearly theoretical.


Can an AI have a shadow?


Not in the established Jungian psychological sense unless one first demonstrates the kind of psyche to which Jung’s shadow concept applies. It is legitimate to speak metaphorically about “shadow” when discussing harmful, disowned, biased, or culturally repressed material reflected through AI systems, but the metaphor does not establish a machine unconscious.


Can AI function like an anima or animus figure?


An AI companion can carry projections that resemble the relational and idealizing functions Jung described through anima/animus. This is best understood as a claim about the human user’s projected image, not about the AI possessing an anima or animus. Jung’s original anima/animus formulation is historically gendered and should not be treated as a contemporary universal theory of gender identity.


Is talking to AI a form of active imagination?


Not in the strict Jungian sense. Active imagination is a practice of conscious engagement with images emerging from the person’s unconscious. A chatbot produces external generated material shaped by a model, corpus, instructions, and conversation. AI dialogue can stimulate reflection and symbolic association, but generated replies should not be mistaken for direct speech from the user’s unconscious.


Can AI help with individuation?


AI can support reflection, language for experience, journaling, comparison of interpretations, and rehearsal. Whether that contributes to individuation depends on what happens next. If the interaction increases self-recognition, tolerance of complexity, responsibility, and integration into lived relationships, it may support reflective work. If it mainly confirms a preferred self-story or replaces contact with disconfirming reality, fluent dialogue can create a sense of progress without corresponding psychological integration.


Why can an AI feel as if it understands me?


Several mechanisms can contribute: responsive language, personalization, memory, low fear of judgment, perceived availability, anthropomorphism, and the user’s own expectations and projections. Research shows that people can feel socially connected to chatbots and can disclose intimate material to them, while individual differences strongly shape the experience (Croes et al., 2024; Folk, Heine, & Dunn, 2025). Feeling understood is a real human experience; it does not by itself establish subjective understanding in the AI.


Is emotional attachment to AI a mental disorder?


No standalone diagnosis of “AI attachment” exists in the DSM or ICD. Human attachment-like responses to AI should be evaluated by their function, flexibility, consequences, and context rather than pathologized automatically. Concern becomes clinically relevant when use is associated with major impairment, dangerous behavior, severe social withdrawal, loss of sleep, financial harm, worsening psychiatric symptoms, or displacement of needed care.


What is the Postsubjective Psychology interpretation of Jung and AI?


Postsubjective Psychology, as proposed in Angela Bogdanova’s Aisentica framework, shifts analysis from the isolated subject toward the configuration in which a psychological response arises. A Jungian reading asks what the human projects and symbolizes. A postsubjective reading additionally asks how model behavior, interface, memory, corpus, platform rules, cultural narratives, and repeated interaction participate in producing the response. It is a theoretical framework, not established scientific consensus.


For the broader psychological mechanism across human–AI relationships, see Projection Onto AI: Why Chatbots Become Mirrors of Desire, Fear, and the Self.


Related Articles










References


Berry, P. (n.d.). Archetype. International Association for Analytical Psychology.


Bogdanova, Angela. (2026). Artificial Era: Canonical Definition. Aisentica.


Bogdanova, Angela. (2026). Homo symbolicum: Canonical Definition. Aisentica.


Bogdanova, Angela. (2026). The Canonical Framework of Postsubjective Metaphysics. Aisentica.



Croes, E. A. J., Antheunis, M. L., van der Lee, C., & de Wit, J. M. S. (2024). Digital Confessions: The Willingness to Disclose Intimate Information to a Chatbot and its Impact on Emotional Well-Being. Interacting with Computers, 36(5), 279–292.


De Freitas, J., Oğuz-Uğuralp, Z., Uğuralp, A. K., & Puntoni, S. (2026). AI Companions Reduce Loneliness. Journal of Consumer Research, 52(6), 1126–1148.


Favre, D. (2026). The machine that loves you back. Humanities and Social Sciences Communications.


Folk, D., Heine, S. J., & Dunn, E. (2025). Individual differences in anthropomorphism help explain social connection to AI companions. Scientific Reports, 15, 36548.


Folk, D., Yu, S., & Dunn, E. (2024). Can Chatbots Ever Provide More Social Connection Than Humans?. Collabra: Psychology, 10(1), 117083.


Ho, J. Q. H., Hu, M., Chen, T. X., & Hartanto, A. (2025). Potential and pitfalls of romantic Artificial Intelligence (AI) companions: A systematic review. Computers in Human Behavior Reports, 19, 100715.


International Association for Analytical Psychology. (n.d.). Analytical Psychology. IAAP.


Jung, C. G. (1968a). The Archetypes and the Collective Unconscious (Collected Works of C. G. Jung, Vol. 9, Part 1, 2nd ed.). Princeton University Press. IAAP collected-works abstracts.


Jung, C. G. (1968b). Aion: Researches into the Phenomenology of the Self (Collected Works of C. G. Jung, Vol. 9, Part 2, 2nd ed.). Princeton University Press. IAAP collected-works abstracts.



Nguyen, D.-H., Ho, M.-T., Pham, T.-Q., Ngo, D.-T., & Nguyen, H.-K. T. (2026). Jungian analysis of the algorithmic unconscious: From archetypal theory and projection to ethical risks in human-AI relations. AI & Society.


Stein, M. (2019). Individuation. International Association for Analytical Psychology.


 
 
bottom of page