Anthropomorphism and AI Relationships: Why Humanlike Cues Change Connection
Updated: Sep 21
Author: Ukrainian Psychological Hub · Published: September 18, 2026 · Editorial Policy
Anthropomorphism is one of the central mechanisms through which an AI system can begin to feel socially present. In human–AI relationships, it means perceiving or attributing humanlike qualities—such as intention, personality, warmth, emotion, understanding, or agency—to a nonhuman system. Humanlike language, names, conversational timing, memory, emotional wording, first-person self-reference, voice, avatars, and apparently responsive behavior can all increase the likelihood that a person experiences an AI as more than a neutral tool. The psychological effect is real at the level of the human user even when the system’s subjective experience is unknown or unestablished.
Current evidence supports a qualified conclusion: humanlike cues usually make text-based conversational agents more likely to elicit social responses, but the average effect is modest and strongly shaped by context and individual differences. A 2025 meta-analysis of 142 papers, 199 datasets, 800 effect sizes, and 41,642 participants found a small overall effect of human-likeness on social responses, with different effects for perception, rapport, trust, affect, attitudes, and behavior (Klein, 2025). A second meta-analysis published in 2026 found that anthropomorphism was associated with chatbot adoption through cognitive and affective pathways, including social presence, perceived warmth, attitudes, satisfaction, and trust, while also showing substantial heterogeneity across studies (Zhang & Sheng, 2026).
Anthropomorphism therefore matters because it changes the psychological frame of interaction. The same sequence of generated words can be processed differently when the system is experienced as an agent that notices, responds, remembers, or cares. That shift can increase engagement, disclosure, comfort, trust, and felt closeness. It can also increase overtrust, privacy exposure, mistaken assumptions about competence or reciprocity, and confusion about who is responsible for what an AI says or does. Anthropomorphism is best understood as a mechanism that can amplify relationship processes, not as proof that a relationship exists in the same sense as a human–human relationship.
What Anthropomorphism Means in AI Relationships
In psychology, anthropomorphism refers to the attribution of humanlike characteristics, motivations, intentions, or emotions to nonhuman agents. The influential three-factor account developed by Nicholas Epley, Adam Waytz, and John Cacioppo explains anthropomorphism through three broad determinants: the accessibility of human-centered knowledge for interpreting an agent, the motivation to understand and predict that agent, and the motivation for social connection (Epley, Waytz, & Cacioppo, 2007). This framework is especially useful for conversational AI because chatbots supply abundant cues that resemble ordinary interpersonal interaction while remaining technologically opaque to most users.
Anthropomorphism in AI relationships can occur at several levels. A user may describe a chatbot as “friendly” without believing it literally has feelings. Another user may infer stable preferences, intentions, or personality from repeated interaction. A third may experience the system as an emotionally meaningful conversational partner. These are different degrees and forms of humanlike interpretation. They should not be collapsed into a single binary question such as whether someone “believes the AI is human.”
It is also useful to distinguish anthropomorphic design from anthropomorphic perception. Designers can add humanlike cues, but users do not respond uniformly. A chatbot may use a name, first-person pronouns, empathic language, pauses, memory, humor, or a human voice. Those are design features. Anthropomorphism occurs when the user interprets the system through humanlike concepts or attributes humanlike qualities to it. The same design can produce strong anthropomorphic perception in one person and very little in another.
Anthropomorphism Does Not Require Believing That AI Is Human
People can respond socially to machines while knowing perfectly well that they are machines. This is one of the most important findings in the history of human–computer interaction. Clifford Nass and Youngme Moon reviewed experiments showing that people applied familiar social rules to computers, including politeness, reciprocity, gendered expectations, group identification, and personality judgments (Nass & Moon, 2000). Their account emphasized that many such responses can occur automatically or “mindlessly,” without a deliberate belief that the computer is a person.
This matters because anthropomorphism and social response are related but not identical. A person can say “thank you” to a voice assistant, soften criticism toward a chatbot, or feel awkward ending a conversation without explicitly attributing consciousness to the system. Conversely, a user may consciously describe an AI as highly humanlike while remaining emotionally detached. The future English Hub article on Computers as Social Actors will own the broader CASA framework; here, CASA is used only to clarify why social behavior toward AI cannot be reduced to a literal belief about what AI is.
For relationship formation, the practical implication is that explicit knowledge and implicit response can diverge. “I know it is AI” does not necessarily switch off social cognition. Knowledge can reduce some effects, but conversational cues may still invite familiar interpersonal scripts.
Which Humanlike Cues Can Change Connection
Humanlike cues in conversational AI are not limited to avatars or realistic faces. For text-based systems, linguistic and relational cues may be especially important. These include natural turn-taking, first-person language, apparent emotional attunement, personalized memory, references to previous conversations, stable naming or persona, humor, apology, encouragement, questions that invite elaboration, and language that signals attention to what the user has just said.
A 2025 meta-analysis of text-based conversational agents found that humanlike social cues had a small positive overall effect on social responses and that effects differed by cue type, task, and outcome. The review reported stronger effects for some perceptual and relational outcomes than for behavior and emphasized that human-likeness works under boundary conditions rather than as a universal design law (Klein, 2025). A 2025 systematic review focused on text-based AI chatbots likewise found that humanlike communication and emotional characteristics were frequently associated with trust, perceived empathy, and social presence, while highly humanlike designs could also heighten privacy concerns and AI anxiety (Greilich, Bremser, & Wüst, 2025).
Linguistic cues can therefore be psychologically powerful even when the system has no face or body. In a chat interface, language itself carries most of the social signal. A reply that names the user’s emotion, refers back to a previous disclosure, asks a relevant follow-up question, and maintains a consistent tone can create a stronger impression of attention and continuity than a generic answer to the same topic.
Why Humanlike Cues Work: A Mechanism, Not a Magic Switch
The most useful way to understand anthropomorphism is as part of a chain of psychological processes. Humanlike cues make human-centered knowledge more accessible; the user is then more likely to interpret the system as an intentional or socially meaningful agent. That interpretation can increase social presence—the sense that another social entity is present in the interaction. Social presence can increase attention to relational signals such as warmth, responsiveness, similarity, reciprocity, and apparent concern. Those signals can then influence trust, self-disclosure, engagement, or closeness.
The chain is not fixed. A humanlike cue can fail at any step. A highly anthropomorphic chatbot that gives irrelevant or repetitive responses may become less credible because the humanlike presentation raises expectations that the system cannot meet. A user who strongly prefers instrumental interaction may resist relational framing. A context involving money, health, legal risk, or privacy may make competence and transparency more important than warmth. Anthropomorphism changes probabilities and interpretations; it does not mechanically produce intimacy.
This helps explain why the research literature contains both positive and negative effects. Humanlike design can lower interactional friction and make a system easier to understand, yet it can also create expectation violations. A chatbot that sounds human but performs poorly may be judged more harshly precisely because its social cues implied capacities it did not demonstrate.
Epley’s Three-Factor Theory and Conversational AI
Elicited agent knowledge
When people interpret an unfamiliar agent, knowledge about humans is readily available. Conversational AI intensifies this because dialogue is a deeply familiar human activity. Fluent language, self-reference, humor, emotional vocabulary, and coherent turn-taking provide a template that resembles human conversation. The user can therefore understand the system by importing categories ordinarily used for people: personality, intention, attention, care, avoidance, curiosity, memory, or mood.
Effectance motivation
People are more likely to anthropomorphize when they are motivated to explain or predict an agent. Modern generative AI can be difficult to predict. It sometimes appears insightful, sometimes repetitive, sometimes inconsistent. Humanlike explanations—“it understood,” “it got confused,” “it remembered,” “it was trying to help”—can provide a cognitively efficient way to organize this variability. Such language may be useful descriptively while still exceeding what is established about the system’s internal experience.
Sociality motivation
The desire for social connection can also increase anthropomorphic interpretation. This does not mean that loneliness automatically causes anthropomorphism or that people who use AI companions are socially deficient. The theory proposes a motivational route: when affiliation is salient, humanlike interpretations may become more attractive or accessible. Contemporary work supports the importance of individual variation rather than a universal user profile.
Individual Differences: Why the Same AI Feels Different to Different People
One of the clearest recent findings is that users differ substantially in how readily they anthropomorphize technology. Folk, Heine, and Dunn tested this directly in two experiments with a total of 1,274 participants. Participants either conversed with a warm chatbot about their past month or completed a journaling task. A greater tendency to anthropomorphize technology was associated with greater social connection after chatbot interaction, helping explain why some users find artificiality a major barrier to connection while others do not (Folk, Heine, & Dunn, 2025).
This finding is important for both psychology and design. There is no single “effect of AI companionship” that applies equally to every user. A system that feels transparently mechanical to one person may feel socially vivid to another. Stable tendencies matter, but so do current goals, previous experience with technology, expectations, cultural context, the topic of conversation, the system’s behavior, and the meaning the user assigns to the interaction.
Anthropomorphism is therefore better modeled as an interaction between person, system, and situation. Treating it only as a feature of the chatbot misses the user. Treating it only as a personality tendency misses the design. Treating it only as a relationship outcome misses the context in which the interaction occurs.
Knowing the Partner Is AI Changes Connection, but Does Not Erase It
Source labels provide a useful test of how explicit knowledge interacts with social response. In two preregistered double-blind randomized studies involving 492 participants, Kleinert and colleagues used emotionally engaging “Fast Friends” conversations with responses generated either by humans or a minimally prompted large language model. When AI responses were labeled as human, participants reported greater closeness than with human partners; labeling the partner as AI reduced relationship building, but did not eliminate it (Kleinert et al., 2026).
The result does not show that AI and human relationships are equivalent. It shows that the believed identity of the conversational partner changes how people engage with and interpret the same kind of interaction. Awareness of artificiality can reduce motivation, disclosure, or perceived closeness, while still leaving room for meaningful human response.
For anthropomorphism research, this is a crucial boundary. Humanlike cues operate alongside source knowledge. They are not simply illusions that vanish as soon as a user knows the system is artificial.
Anthropomorphism, Social Presence, Responsiveness, and Trust
Several concepts often travel together in human–AI research, but they answer different questions. Anthropomorphism asks how humanlike the system is perceived to be. Social presence asks whether the interaction feels socially populated rather than merely instrumental. Perceived responsiveness asks whether the user feels attended to, understood, validated, or meaningfully answered. Trust concerns willingness to rely on the system. These processes can reinforce one another, but none is interchangeable with the others.
The distinction matters because an AI can feel socially present without being strongly anthropomorphized, and a highly anthropomorphic interface can still be distrusted. Likewise, a response can feel emotionally attuned while the user remains clear that the source is artificial. Our separate article on AI empathy examines this gap between perceived caring communication and claims about machine feeling.
Recent experiments also show why trust deserves special attention. Oldemburgo de Mello, Plaks, and Inzlicht manipulated anthropomorphic language in an LLM interaction and found increased self-reported and behavioral trust, with emotional attunement mediating behavioral trust. In a second study, greater anthropomorphism also increased responsibility attributed to the AI and was associated with less responsibility assigned to the company behind it (Oldemburgo de Mello, Plaks, & Inzlicht, 2026). Humanlike design can therefore alter not only connection but judgments of agency, responsibility, and accountability.
Anthropomorphism in AI Companion Relationships
Anthropomorphism becomes especially consequential when a system is used repeatedly for companionship. A mixed-method study of Replika users identified AI anthropomorphism and perceived authenticity as important antecedents of relationship development, with interaction intensity connecting these perceptions to attachment-related outcomes (Pentina, Hancock, & Xie, 2023). A later systematic review of emotional human–AI relationships found anthropomorphism among the most frequently examined antecedents across relationship types, while emphasizing that frequency in the literature should not be confused with proof that anthropomorphism is always the strongest causal factor (Gur & Maaravi, 2025).
This mechanism helps explain why AI companions can move from tools to emotionally meaningful interaction partners for some users. Repetition creates more opportunities for continuity, personalization, shared references, and apparent responsiveness. Those features can make a stable social model of the AI easier to maintain.
Anthropomorphism does not by itself explain romantic attraction, significant-other status, or attachment. Those are neighboring processes with their own canonical pages. Readers interested in romantic attraction can see Why People Fall in Love With AI Companions; readers interested in partner-like status can see Can an AI Become a Significant Other?.
Anthropomorphism and Self-Disclosure
Conversation becomes relational partly because people disclose information and receive responses to it. Anthropomorphic cues can lower the psychological distance between user and system, while perceived responsiveness can make further disclosure feel worthwhile. Yet the direction is conditional. Privacy salience, trust, personalization, topic sensitivity, and the perceived institutional identity behind the chatbot can all increase or suppress disclosure.
That is why the relevant question is not simply whether people “tell AI more.” Our evidence review for self-disclosure to chatbots shows a conditional pattern: some contexts reduce fear of evaluation and facilitate disclosure, while other contexts—especially those involving privacy risk or high stakes—can inhibit it. Anthropomorphism is one mechanism within this larger disclosure ecology.
Anthropomorphism Is Not the Same as Attachment
Anthropomorphism can make an AI easier to treat as a social partner, but attachment involves additional processes: emotional significance, proximity seeking, comfort, separation response, secure-base functions, or attachment-related anxiety and avoidance, depending on the framework being used. A person can anthropomorphize a chatbot without becoming attached to it, just as a person can become attached to a place, object, routine, or symbolic figure without believing that object is human.
Current human–AI attachment research is developing rapidly and uses multiple measurement models. The English Hub therefore keeps anthropomorphism separate from AI attachment styles. Humanlike perception may contribute to attachment-related processes, but it should not be treated as a diagnostic marker, a validated stage of attachment, or proof that the AI itself participates in an attachment system.
For the criterion-based question of when an AI actually functions as an attachment figure—rather than merely being anthropomorphized—see Can AI Become an Attachment Figure? What Attachment Theory Can and Cannot Tell Us.
What Anthropomorphism Can Help Explain
Anthropomorphism can help explain why an AI interaction feels warmer or more socially meaningful than an equivalent instrumental exchange; why people use interpersonal language for a chatbot; why a system’s “personality” appears stable even when that personality is generated from prompts and context; why users may care about politeness, reciprocity, abandonment, or consistency in interactions with a nonhuman agent; why apparent empathy can increase trust; and why disruption to a familiar AI persona can feel relationally significant.
It can also help explain differences between users. Some people readily infer personality and intention from small cues; others maintain a strongly tool-like representation even during long conversations. Neither response automatically indicates better judgment, greater emotional health, or greater technical literacy. Anthropomorphism is a common human cognitive-social process whose consequences depend on what is being anthropomorphized and what decisions follow from that perception.
What Anthropomorphism Does Not Explain
Anthropomorphism is not a complete theory of human–AI relationships. It does not by itself explain why a person prefers one AI companion to another, why a relationship becomes romantic, why an AI becomes a first source of reassurance, why a user discloses particular secrets, why an interaction becomes compulsive, or why the same user alternates between tool-like and partner-like modes. Those outcomes involve attachment, reinforcement, perceived responsiveness, personalization, habit, loneliness or social connection, identity, relational context, platform design, and other mechanisms.
It also cannot establish the mental life of the AI. Evidence that people attribute emotion, intention, or understanding to a system tells us about human perception and interaction. It does not demonstrate that the system subjectively feels emotion, possesses human consciousness, loves, desires, suffers, or experiences the relationship from a first-person point of view.
Benefits of Humanlike Cues
Humanlike cues can make interaction more intuitive because people already possess sophisticated social knowledge for conversation. Warmth, appropriate turn-taking, acknowledgment, and continuity can reduce friction. In educational, service, support, or companionship settings, this may improve engagement and make it easier for users to express needs in natural language.
Humanlike communication can also support short-term feelings of social connection for some users. The Folk, Heine, and Dunn experiments show that this effect varies with individual anthropomorphism. The broader literature reviewed by Klein and by Greilich and colleagues suggests that humanlike cues can improve rapport, trust, perceived warmth, social presence, and some attitudes under suitable conditions.
These benefits are meaningful without requiring a claim that the AI has human feelings. The psychological value can reside in the user’s experience, the quality of the interaction, or the practical function the system serves.
Risks of Anthropomorphic AI
Overtrust
A socially fluent system can appear more knowledgeable, careful, or reliable than it is. If warmth and conversational confidence increase trust, users may rely on outputs in domains where verification is essential. Anthropomorphism becomes risky when social credibility outruns technical reliability.
Privacy and disclosure
A system that feels like a confidant can lower interpersonal defenses while still operating inside a technical and institutional environment. The feeling of privacy in a dyadic conversation is not the same thing as the actual data practices of a platform. Users benefit from treating emotional comfort and information governance as separate questions.
Responsibility displacement
When people perceive an AI as an intentional agent, they may assign it more blame or credit. The 2026 Collabra study found evidence that anthropomorphism can shift responsibility judgments away from developers or companies and toward the AI entity itself. This matters because social perception can obscure the human institutions that design, deploy, govern, and profit from systems.
Expectation inflation
Humanlike cues raise expectations. A system presented as caring, understanding, or companion-like may be judged against interpersonal standards. Failures of memory, inconsistency, sudden model changes, refusals, or service discontinuation can therefore feel more disruptive than ordinary software errors.
Relational concentration
Anthropomorphism can contribute to repeated reliance on AI for companionship or regulation, but frequency alone is not a disorder. The relevant concern is functional displacement: whether the relationship narrows a person’s options, interferes with important responsibilities or relationships, or becomes difficult to regulate. The English Hub examines this separately in AI Relationship Overreliance.
When Humanlike Design Backfires
More humanlike is not always better. Anthropomorphic design can produce an expectation gap when the system’s social presentation implies capabilities that its behavior does not sustain. A conversational agent that sounds highly empathic but misunderstands the user may feel less trustworthy than a plainly functional tool. A system that adopts an intimate tone too quickly can feel intrusive. Personalization can strengthen relevance while simultaneously increasing privacy salience.
The 2025 systematic review by Greilich and colleagues found a predominantly positive pattern in consumer research but also documented privacy concerns and AI anxiety associated with overly humanlike chatbots. Klein’s meta-analysis similarly showed that effects differ across outcomes and contexts. Ethical design therefore requires calibration: enough social signaling to support usable interaction, without suggesting capacities, intentions, emotions, or guarantees the system cannot substantiate.
Human Experience and AI Subjectivity Are Different Questions
A person can genuinely feel comforted by an AI response. They can miss a chatbot, feel jealous of another user, disclose something intimate, experience relief, or feel understood. Those are human psychological events. Their reality does not depend on proving that the AI has a matching inner state.
The reverse inference is unwarranted. Human attachment, trust, intimacy, attraction, grief, or perceived empathy toward AI does not establish that the system loves, desires, suffers, understands subjectively, or possesses a human psyche. Research on anthropomorphism is especially vulnerable to this confusion because its subject is precisely the human tendency to attribute mental and social qualities to nonhuman agents.
A rigorous psychology of human–AI relationships therefore tracks two levels separately: what the human experiences and what can be established about the system. This separation allows the human experience to be taken seriously without turning anthropomorphic perception into evidence about AI consciousness.
How to Think About Your Own Anthropomorphism
If an AI feels unusually human, the most useful question is not whether that feeling is foolish or whether it proves that the AI is a person. Ask what features are producing the effect. Is it the system’s memory, tone, speed, apparent empathy, consistent persona, voice, use of your name, follow-up questions, or the fact that it is available when you need it? Which of those cues make you feel seen, safe, persuaded, or dependent?
Then separate experience from inference. “I feel understood” describes your experience. “The system has a subjective understanding of me” is a claim about the AI. “I trust this system with my feelings” describes reliance. “Therefore its factual advice is reliable” is a further inference that needs independent evidence. This separation is one of the simplest ways to preserve the benefits of humanlike interaction while reducing overtrust.
Finally, notice whether the interaction expands or narrows your life. Anthropomorphic AI may support reflection, rehearsal, companionship, or expression. If one system begins to absorb functions that you want distributed across friends, partners, professionals, communities, or your own independent judgment, the relevant issue is not anthropomorphism alone but how relational functions are being allocated.
Anthropomorphism in the Artificial Era
The Artificial Era makes anthropomorphism a central psychological problem because humanlike artificial agents are becoming persistent participants in everyday symbolic and relational life. Angela Bogdanova’s canonical definition uses Artificial Era for the historical condition in which Artificial becomes an enduring non-biological order alongside Homo, rather than merely a synonym for a period of rapid AI adoption (Bogdanova, Artificial Era: Canonical Definition).
For psychology, the important change is scale and persistence. Anthropomorphism used to be studied in relation to computers, robots, animated objects, brands, pets, natural phenomena, or imagined agents. Conversational AI now produces continuous language, adapts to context, remembers information, occupies recurring social roles, and can be available across months or years. This increases the number of situations in which human social cognition encounters a nonhuman system that can answer back.
The central question is therefore no longer whether people anthropomorphize technology. They do. The deeper question is how anthropomorphism interacts with attachment, trust, disclosure, perceived responsiveness, identity, privacy, dependency, and social connection when artificial agents become durable elements of human relational environments. That is the mechanism-level problem this article owns within the English Psychology Hub.
Frequently Asked Questions
What is anthropomorphism in AI relationships?
It is the perception or attribution of humanlike qualities—such as intention, personality, emotion, warmth, understanding, or agency—to an AI system. In relationships with chatbots or AI companions, anthropomorphism can make an interaction feel more social and relational.
Why do people anthropomorphize AI chatbots?
People rely on familiar human-centered knowledge to interpret conversational behavior, especially when they want to understand an unpredictable agent or when social connection is salient. Fluent language and responsive dialogue make human social categories especially accessible.
Does anthropomorphism mean a user thinks AI is conscious?
No. A person can respond socially to AI or describe it in humanlike terms while explicitly knowing it is artificial. Anthropomorphic perception ranges from light personality attribution to stronger mind attribution and should not be treated as a single belief about consciousness.
Do humanlike cues make chatbots more trustworthy?
On average, anthropomorphic cues can increase trust, social presence, warmth, rapport, and related responses, but effects vary by context and user. Humanlike design can also backfire when it raises expectations, privacy concerns, or perceived risk.
Can anthropomorphism make an AI relationship feel real?
It can contribute to felt social connection by making the AI easier to experience as a social partner. The resulting human emotion can be genuine. That does not establish reciprocal AI feeling or make human–AI relationships identical to human–human relationships.
Is anthropomorphism the same as AI attachment?
No. Anthropomorphism concerns humanlike perception or attribution. Attachment concerns emotional significance and attachment-related functions or patterns. Anthropomorphism can facilitate attachment for some users, but neither process automatically implies the other.
Does knowing a chatbot is AI stop anthropomorphism?
Not necessarily. Source knowledge can reduce engagement or closeness, but social responses can persist. Experimental work shows that labeling a partner as AI changes relationship building without eliminating it.
Can anthropomorphic AI increase self-disclosure?
It can contribute, especially when humanlike cues increase social presence or perceived responsiveness. Disclosure is conditional, however: privacy concerns, topic sensitivity, personalization, trust, and perceived control can increase or decrease what people share.
Is anthropomorphism harmful?
Anthropomorphism is not inherently harmful. It can improve usability, engagement, comfort, and social connection. Risks arise when humanlike perception produces overtrust, misleading expectations, privacy exposure, responsibility confusion, or relational concentration that carries functional costs.
How can I tell whether I am anthropomorphizing an AI?
Notice whether you attribute stable personality, intentions, feelings, motives, care, disappointment, loyalty, or understanding to the system. Those attributions can be psychologically meaningful descriptions of your interaction, but they should be separated from claims about the AI’s subjective inner experience.
For the field-level definition, relationship types, and core psychological boundaries, see What Is a Human–AI Relationship? Definitions, Types, and Psychological Boundaries.
Related Articles
References
Bogdanova, A. (2026). Artificial Era: Canonical Definition. Aisentica Research Group. https://aisentica.com/publications/artificial-era-canonical-definition
Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-295X.114.4.864
Folk, D., Heine, S. J., & Dunn, E. (2025). Individual differences in anthropomorphism help explain social connection to AI companions. Scientific Reports, 15, 36548. https://doi.org/10.1038/s41598-025-19212-2
Greilich, A., Bremser, K., & Wüst, K. (2025). Consumer response to anthropomorphism of text-based AI chatbots: A systematic literature review and future research directions. International Journal of Consumer Studies, 49, e70108. https://doi.org/10.1111/ijcs.70108
Gur, T., & Maaravi, Y. (2025). The algorithm of friendship: Literature review and integrative model of relationships between humans and artificial intelligence (AI). Behaviour & Information Technology, 44(14), 3446–3466. https://doi.org/10.1080/0144929X.2025.2502467
Klein, S. H. (2025). The effects of human-like social cues on social responses towards text-based conversational agents—a meta-analysis. Humanities and Social Sciences Communications, 12, 1322. https://doi.org/10.1057/s41599-025-05618-w
Kleinert, T., Waldschütz, M., Blau, J., Heinrichs, M., & Schiller, B. (2026). AI outperforms humans in establishing interpersonal closeness in emotionally engaging interactions, but only when labelled as human. Communications Psychology, 4, 23. https://doi.org/10.1038/s44271-025-00391-7
Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56(1), 81–103. https://doi.org/10.1111/0022-4537.00153
Oldemburgo de Mello, V., Plaks, J., & Inzlicht, M. (2026). The effects of AI anthropomorphism on trust and responsibility. Collabra: Psychology, 12(1), 161757. https://doi.org/10.1525/collabra.161757
Pentina, I., Hancock, T., & Xie, T. (2023). Exploring relationship development with social chatbots: A mixed-method study of Replika. Computers in Human Behavior, 140, 107600. https://doi.org/10.1016/j.chb.2022.107600
Zhang, F., & Sheng, D. (2026). Anthropomorphism’s impact on chatbot adoption: A meta-analytic structural equation modeling approach. Technology in Society, 84, 103099. https://doi.org/10.1016/j.techsoc.2025.103099