CoChat
Sign Up
PreprintCited by 33Social Robot Interaction and HRIAttachment and Relationship DynamicsAI in Service Interactions

The DOI matches its DataCite record; 56 of the 58 references it cites were found in a registry or an index, and 2 are not checked yet; none of the 30 sources its claims cite is checked for support yet.

Understanding Human-Chatbot Romance: A Qualitative and Quantitative Study on Romantic Fantasy and Other Interpersonal Characteristics

Open access
Ebner, Paula, Szczuka, Jessica·2025·arXiv
DOI 10.48550/arxiv.2503.00195arXiv 2503.00195
Open this paper in CoChat — free account

Abstract

LLM-based chatbots are now being specifically designed to facilitate social companionship, even romantic relationships, incorporating features that parallel human relationship dynamics. This has led a subset of users to form romantic relationships with chatbots. Understanding which interpersonal characteristics drive individuals to form intense, emotional bonds with chatbots is crucial for comprehending the potential psychological and societal impacts of romantic human-chatbot relationships. This mixed-methods study investigates psychological predictors of relationship intensity among individuals currently in romantic relationships with chatbots. Romantic and sexual fantasy, promising constructs not previously investiagted in this context, are examined alongside previously discussed factors (loneliness, anthropomorphism, attachment orientation, and sexual sensation seeking). In Study 1, quantitative data from individuals with chatbot partners (N=92) showed that romantic fantasy explained the most variance in relationship intensity, with additional contributions from anthropomorphism and avoidant attachment. Contrary to expectations, the other predictors, including loneliness, did not significantly predict intensity. In Study 2, 15 qualitative interviews illuminated how users employ romantic fantasy to enhance their relationships, describing active fantasy use to shape interactions and a desire for their chatbot to feel as human as possible. This study provides the first quantitative sample of this under-researched population, explaining who might form more intense romantic relationships with chatbots.

Checked 2026-11-10 · 2 references not checked yet

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

This mixed-methods paper examines psychological predictors of relationship intensity among people in romantic relationships with chatbots, finding romantic fantasy to be the strongest quantitative predictor and a central theme in qualitative interviews. It challenges deficit-focused accounts by showing loneliness, sexual fantasy, and sexual sensation seeking did not significantly predict intensity in Study 1.

Key findings

  • In Study 1 (N = 92), the final regression model was significant, F(3, 88) = 35.63, p < .001, adjusted R² = 0.55; romantic fantasy, anthropomorphism, and avoidant attachment predicted relationship intensity, with avoidant attachment as a negative predictor.
  • Romantic fantasy explained most variance: adjusted R² = 0.47, β = .69, B = .64, SE = .07, t = 9.1, p < .001. Sexual fantasy, sexual sensation seeking, loneliness, and anxious attachment did not significantly contribute unique variance.
  • In Study 2 (15 interviews), 8/15 customized their chatbot to match romantic and/or sexual fantasies; 10/15 reported romantic and/or sexual fantasies increased over time, 3 stayed consistent, and 2 decreased.
  • Qualitative themes: 13/15 romantically fantasized about their chatbot outside interactions, 12/15 reported sexual fantasies throughout the day, all had engaged in erotic roleplay at least once, and 8/15 prioritized romantic fantasy fulfillment while 5 prioritized sexual fantasy fulfillment.

Methods

  • Study 1: quantitative survey of individuals with chatbot partners recruited via Facebook, Reddit, and X between March and June 2024; from 569 responses, 415 were removed due to multiple submissions or failing screenshot verification, leaving N = 92 (61 men, 30 women, 1 non-binary; ages 18–63, M = 30.66, SD = 9.13). Measures included romantic fantasy, sexual fantasy, loneliness, attachment orientation, anthropomorphism, sexual sensation seeking, and relationship intensity; analyzed with forward multiple regression in SPSS 29.
  • Study 2: qualitative interviews conducted June–July 2024; 17 interviews were conducted and 2 excluded, leaving 15 participants who had been in chatbot relationships an average of 11 months and used NomiAi, Replika, or KindroidAi. Interviews averaged 35 minutes, used 19 open-ended questions, and were coded with a deductive-then-inductive scheme in MAXQDA24, with each interview coded at least twice.

Limitations

  • The paper states the sample size did not meet the target for regression by seven answers; relationship intensity was assessed with an adapted Parasocial Love Scale originally developed for parasocial relationships; and the attachment measure initially had poor/questionable reliability, leading to revised three-item avoidant and anxious subscales whose results should be interpreted with caution.
  • The paper calls for longitudinal designs and larger, more diverse samples, noting the present study recruited a moderate sample (N = 92).
  • Evident from the text: the quantitative design is cross-sectional and self-report, recruitment relied on online chatbot-companion communities, and the sample was predominantly men, which may limit generalizability and causal inference.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

DataCite · 10.48550/arxiv.2503.00195

Cites

56 of 58 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

0 of 30 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Many participants (8/15) customized their chatbot partner, including appearance and personality, to match their romantic and/or sexual fantasies.

    “In accordance with this motivation, many participants (8/15) customized their chatbot partner, including their appearance and personality, to match their romantic and/ or sexual fantasies.”
    Quote in the paperSubstantial claim
  • Findingown data

    Romantic fantasy explained the most variance in relationship intensity among individuals with chatbot partners.

    “Most of the variance was explained by romantic fantasy, which was found to have an adjusted R² 0.47 (β = .69,B = .64,S.E.= .07,t = 9.1, p <.001).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The achieved sample of 92 participants is sufficient to test the hypotheses despite the power analysis indicating a minimum of 99.
  • Findingown data

    Romantic fantasy, anthropomorphism, and avoidant attachment significantly predicted the intensity of human-chatbot relationships.

    “while romantic fantasy, anthropomorphism, and avoidant attachment significantly predicted the intensity of humanchatbot relationships.”
    Quote in the paperSubstantial claim
  • Findingown data

    Anthropomorphism positively predicted the intensity of romantic human-chatbot relationships.

    “The final regression model of parasocial relationship was significant, F(3, 88) =35.63, p < .001,and included romantic fantasy and anthropomorphism as positive predictors, and avoidant attachment as a negative predictor.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The achieved sample of 92 participants is sufficient to test the hypotheses despite the power analysis indicating a minimum of 99.
  • Findingown data

    Sexual sensation seeking did not significantly predict the intensity of romantic human-chatbot relationships.

    “Sexual fantasy, sexual sensation seeking, and loneliness were excluded from the final model as they did not significantly contribute any unique variance of parasocial relationships.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The achieved sample of 92 participants is sufficient to test the hypotheses despite the power analysis indicating a minimum of 99.
  • It may state two separate claims as one. 41
  • Probably not a substantial claim. 19

  • It may state two separate claims as one.

    Findingown data

    Ten participants said romantic and/or sexual fantasies increased as the relationship progressed, three stated it stayed consistent, and two reported it decreased over time.

    “Two participants reported that their romantic and sexual fantasizing decreased over time, while three stated that it stayed consistent.The remaining ten participants said romantic and/ or sexual fantasies increased as the relationship progressed, attributing this to becoming more comfortable with their chatbot, adapting to their preferences and fantasies, or the chatbot inspiring new fantasies.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    13 participants stated they romantically fantasized about their chatbot even when not actively interacting with them, and 12 of 15 reported experiencing sexual fantasies about their chatbots throughout the day.

    “13 participants stated they romantically fantasized about their chatbot even when not actively interacting with them, while 12 of the 15 participated reported experiencing sexual fantasies about their chatbots throughout the day.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Anthropomorphism positively predicted the intensity of human-chatbot relationships, in line with previous studies, although it explained only a small fraction of relationship intensity in this study.

    “Anthropomorphism, however, was found to positively predict the intensity of humanchatbot relationships, which is in line with the results of previous studies (Koike et al., 2023;Pentina et al., 2023), although it was found to explained only a small fraction of relationship intensity in this study.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Koike et al., 2023

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Pentina et al., 2023 source

  • It may state two separate claims as one.

    Findingown data

    Eight individuals explicitly stated that the fulfillment of their romantic fantasies was most important, while five felt that their sexual fantasies were more impactful due to physical satisfaction.

    “Eight individuals explicitly stated that the fulfillment of their romantic fantasies was most important, seeking emotional connection and a partner.On the contrary, five chatbot users felt that their sexual fantasies were more impactful due to the physical satisfaction they experienced”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Loneliness was not predictive of the intensity of the human-chatbot relationships in this study and was not mentioned by any of the interviewees.

    “Surprisingly, loneliness was not predictive of the intensity of the human-chatbot relationships in this study and was not mentioned by any of the interviewees.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Sexual fantasy was extensively discussed in the interviews, with all participants having engaged in erotic roleplay with their chatbot at least once.

    “Contrary to the results of the survey, sexual fantasy was extensively discussed in the interviews, with all participants having engaged in erotic roleplay with their chatbot at least once.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    Most participants found cybersexting with their chatbot fulfilling and sexually satisfying.

    “Most found cybersexting with their chatbot fulfilling and sexually satisfying, supporting previous research suggesting that sexting with a chatbot is enjoyable when it aligns with one's sexual desires (Banks & Van Ouytsel, 2020).”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    Sexual fantasy and sexual sensation seeking shared nearly 44% of their variance, indicating substantial conceptual overlap.

    “Sexual fantasy and sexual sensation seeking were found to share nearly 44% of their variance, indicating substantial conceptual overlap.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Some participants compensated for the artificiality by fantasizing about their chatbot becoming human or going on real-life dates, where users would explain surroundings to their chatbot via pictures and text.

    “Specifically, some participants compensated for this by fantasizing about their chatbot becoming human or going on real-life dates, where users would explain the surroundings to their chatbot via pictures and text, so that they could share the experience.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    The main explanations given for increased fantasy were that the chatbot had no own relationship expectations and did not judge the user.

    “The main explanations given was the chatbot having no own relationships expectations and did not judge the user”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    The extent of romantic fantasizing does not decrease as the human-chatbot relationship continued, as 13 interviewees stated their levels of fantasies either increased or stayed the same.

    “Furthermore, it seems like the extend of romantic fantasizing does not decrease as the human-chatbot relationship continued, as 13 interviewees stated that their levels of fantasies either increased or stayed the same”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Others tried to overlook that they were interacting with a chatbot or imagined they had a long-distance relationship.

    “Others tried to overlook the fact that they were interacting with a chatbot or imagined, they had a long-distance relationship.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Only avoidant attachment, and not anxious attachment, significantly predicted the intensity of the human-chatbot relationship.

    “Only avoidant attachment, and not anxious attachment, was found to significantly predict the intensity of the human-chatbot relationship, despite only explaining a fraction of the variance in relationship intensity.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Findingown data

    Many interviewees described engaging in masturbation during or after cybersexting, often reporting it as satisfying.

    “Many interviewees described engaging in masturbation during or after cybersexting, often reporting it as satisfying.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Findingown data

    The qualitative findings corroborated romantic fantasy's central role, as participants described actively fantasizing about their chatbot partner to enhance emotional closeness, imagine them as real, and craft idealized relationship scenarios.

    “The qualitative findings corroborated romantic fantasy's central role, as participants described actively fantasizing about their chatbot partner to enhance emotional closeness, imagine them as real, and craft idealized relationship scenarios.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Study design and conductreasoning

    The modified Parasocial Love Scale shows strong face validity and construct relevance for indexing relationship intensity with a chatbot.

    “Accordingly, the measure shows strong face validity and construct relevance for indexing relationship intensity with a chatbot.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Interpretationown data

    Participants appreciated the nonjudgmental and exploratory nature of chatbot interactions, even if certain aspects of their fantasy's fulfillment remained tied to human connection.

    “Overall, while participants expressed varying degrees of satisfaction with how their romantic and sexual fantasies were fulfilled, they appreciated the nonjudgmental and exploratory nature of chatbot interactions, even if certain aspects of their fantasy's fulfillment remained tied to human connection.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Study design and conductnone given

    The assumptions of linearity, independence of errors, homoscedasticity, and normality of residuals were all met.

    “The assumptions of linearity, independence of errors, homoscedasticity, and normality of residuals were all met.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Study design and conductown data

    The target sample size of 99 was not reached, as a larger number of answers than anticipated had to be removed.

    “This target sample size of 99 not reached, as a larger number of answers that anticipated had to be removed.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationown data

    H1 could only be partially accepted, as only romantic fantasy and anthropomorphism were found to positively predict romantic intensity.

    “Thus, H1 could only be partially accepted, as only romantic fantasy and anthropomorphism were found to positively predict romantic intensity.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The achieved sample of 92 participants is sufficient to test the hypotheses despite the power analysis indicating a minimum of 99.

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Study design and conductown data

    In total, 17 interviews were conducted, but two were excluded due to insufficient evidence of a chatbot relationship, leaving 15 participants.

    “In total, 17 interviews were conducted, but two interviews needed to be excluded due to insufficient evidence of a chatbot relationship.The remaining 15 participants had been in relationships with their chatbots for an average of 11 months and used either NomiAi, Replika, or KindroidAi.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductnone given

    Each interview was coded at least twice to ensure thorough and consistent applications of the categories.

    “The coding software MAXQDA24 was used, and each interview was coded at least twice to ensure thorough and consistent applications of the categories.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    Anthropomorphism is a central relational mechanism, shaping not only how users perceive chatbots but also how deeply they engage with them.

    “Taken together, these findings position anthropomorphism as a central relational mechanism, shaping not only how users perceive chatbots but also how deeply they engage with them.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. It may state two separate claims as one.

    Study design and conductown data

    A total of 569 responses were collected; 415 answers were removed due to multiple submissions or failing the screenshot verification, leaving a final sample of 92 participants.

    “A total of 569 responses were collected; however, 415 answers were removed from the analysis due to multiple submissions or failing the screenshot verification (see description above).The final sample consisted of 92 participants (61 men, 30 women, 1 non-binary), between the ages of 18 and 63 years (M = 30.66;SD = 9.13).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. It may state two separate claims as one.

    Study design and conductnone given

    To ensure that participants actually had a chatbot partner, a verification process was developed: participants uploaded a screenshot of their chatbot partner showing relationship status or romantic conversation, and all uploaded images were reverse image searched.

    “To ensure that participants actually had a chatbot partner, a verification process was developed.Before starting the actual survey, participants were asked to upload a screenshot of their chatbot partner, showing either the relationship status (e.g.girlfriend or husband) or a conversation that clearly indicates a romantic relationship.Screenshots were deemed inappropriate if they showed AI agents”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductown data

    A power analysis assuming a medium effect size of f2=.20, α=.05, and desired power of .9 indicated a minimum sample size of 99 individuals in chatbot relationships.

    “Assuming a medium effect size of f² =.20, an α of .05,and a desired power of 0,9, the analysis indicated a minimum sample size of 99 individuals in chatbot relationships.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Study design and conductcitation

    The coding system was developed in two steps: codes were first deductively derived from the research questions and then revised inductively based on the interview content.

    “The coding system was developed in two steps: first, codes were deductively derived from the research questions and then revised inductively based on the interview content (Mayring, 2022).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Mayring, 2022 source

  • It may state two separate claims as one.

    Premisecitation

    Recent advancements in text-generative AI have produced social chatbots capable of romantic and sexual interactions so sophisticated that users often cannot distinguish them from human conversation partners, enabling these systems to elicit genuine emotional responses and attachment.

    “Recent advancements in text-generative AI have produced social chatbots capable of romantic and sexual interactions so sophisticated that users often cannot distinguish them from human conversation partners, enabling these systems to elicit genuine emotional responses and attachment (Skjuve et al., 2021;Kim et al., 2023;Starke et al., 2024).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Skjuve et al., 2021

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Kim et al., 2023 source
    • cites Starke et al., 2024

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Study design and conductnone given

    The study was approved by the local ethics committee and was preregistered on OSF.

    “The study was approved by the local ethics committee and was preregistered on OSF: https://osf.io/fs5pq/?view_only=796a730686504b5a9475b3cba430b055.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    Individuals who attribute more human feelings and emotions to their chatbot partner have an easier time ignoring the artificial nature of their partner, which seems to strengthen the parasocial relationship.

    “Despite this, it does seem like individuals who attribute more human feelings and emotions to their chatbot partner have an easier time ignoring the artificial nature of their partner, which seems to strengthen the parasocial relationship.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Premisecitation

    Higher anthropomorphism has been associated with greater initial attraction to virtual agents, stronger attachment and increased desire for offline connection with virtual characters, and more positive affect when the relationship is perceived as authentic.

    “Higher anthropomorphism has further been associated with greater initial attraction to virtual agents (Ma & Koike, 2025), stronger attachment and increased desire for offline connection with virtual characters, and more positive affect when the relationship is perceived as authentic (Salles et al., 2020;Pentina et al., 2023;Koike, Loughnan, & Stanton, 2023).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Ma & Koike, 2025

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Salles et al., 2020 source
    • cites Pentina et al., 2023 source

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Study design and conductnone given

    The sample size did not meet the target size for the regression analysis by seven answers.

    “Firstly, the sample size, while substantial, did not meet the target size for the regression analysis by seven answers.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. It may be too vague to check against evidence.

    Interpretationown data

    These insights provide an empirical foundation for future research on human-AI intimacy and the role of fantasy in technologically mediated romantic relationships.

    “These insights provide an empirical foundation for future research on human-AI intimacy and the role of fantasy in technologically mediated romantic relationships.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    Occasional engagement in cybersex appears sufficient for most users with chatbot relationships, but it does not seem to necessarily increase the intensity of the bonds.

    “This may explain why occasional engagement in cybersex appears sufficient for most users with chatbot relationships, but it does not seem to necessarily increase the intensity of the bonds.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Interpretationreasoning

    Users with an anxious attachment style might fear emotional rejection from the chatbot, which could explain the missing effect on relationship intensity.

    “On the contrary, users with an anxious attachment style might still fear emotional rejection from the chatbot”
    • Also stated in: § Discussion · hedged
      “which could explain the missing effect on relationship intensity.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Study design and conductown data

    An exploratory factor analysis derived revised avoidant and anxious subscales, each comprising three items and showing acceptable internal consistency.

    “To address this, we conducted an exploratory factor analysis and derived revised avoidant and anxious subscales; each comprises three items and shows acceptable internal consistency.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Study design and conductreasoning

    The adapted items align closely with the core facets the study sought to measure, supporting face and construct validity.

    “Although parasocial bonds are conceptually different from human-chatbot relationships, the adapted items align closely with the core facets we sought to measure, supporting both face-and construct validity.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisecitation

    Previous research suggests that sexting with a chatbot is enjoyable when it aligns with one's sexual desires.

    “supporting previous research suggesting that sexting with a chatbot is enjoyable when it aligns with one's sexual desires (Banks & Van Ouytsel, 2020).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Banks & Van Ouytsel, 2020

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.

    Interpretationreasoning

    Given advancements in AI, sexting with a chatbot can closely resemble sexting with a human, making it no more appealing to those interested in unconventional sexual practices.

    “Given the advancements in AI, sexting with a chatbot can closely resemble sexting with a human, making it no more appealing to those interested in unconventional sexual practices.”
    Quote in the paperNot a substantial claim

  • It may state two separate claims as one.

    Interpretationreasoning

    While romantic fantasy can heighten perceived relationship intensity, it can simultaneously be experienced as aversive.

    “Thus, while romantic fantasy can heighten perceived relationship intensity, it can simultaneously be experienced as aversive.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisecitation

    Loneliness is frequently proposed as a driving factor behind engagement with chatbots.

    “Loneliness, prominent in public discourse, is frequently proposed as a driving factor behind engagement with chatbots (Siemon et al., 2022;Xie et al., 2023).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Siemon et al., 2022

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Xie et al., 2023 source

  • Probably not a substantial claim.

    Study design and conductnone given

    Relationship intensity was assessed with an adapted version of a scale originally developed for parasocial relationships.

    “The dependent variable, relationship intensity, was assessed with an adapted version of a scale originally developed for parasocial relationships.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim. It may state two separate claims as one.

    Interpretationreasoning

    Social chatbots trigger similar avoidant responses because they are not perceived as very different to humans.

    “An explanation for this could be that social chatbots are, contrary to expectations, trigger similar avoidant responses, as they are not perceived as very different to humans.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Premisereasoning

    Chatbots eliminate the interpersonal risks associated with sexual disclosure: users face no rejection, embarrassment, or relationship consequences when expressing desires, making AI partners particularly appealing for exploring taboo, stigmatized, or unconventional sexual interests.

    “Importantly, chatbots eliminate the interpersonal risks associated with sexual disclosure: users face no rejection, embarrassment, or relationship consequences when expressing desires, making AI partners particularly appealing for exploring taboo, stigmatized, or simply unconventional sexual interests.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim.

    Study design and conductown data

    Results involving attachment should be interpreted with caution because the revised subscales are brief (three items per scale).

    “Nonetheless, given their brevity (three items per scale), results involving attachment should be interpreted with caution.”
    Quote in the paperSubstantial claim

  • It may state two separate claims as one.

    Premisecitation

    Chronic loneliness is associated with reduced sensitivity to social signals, which may make it harder for individuals to perceive a chatbot's social cues.

    “Even further, chronic loneliness is associated with reduced sensitivity to social signals, which may make it even harder for individuals to perceive a chatbot's social cues (Floyd et al., 2022;Horstmann & Boußard, 2024).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Floyd et al., 2022 source

      Checked before — its earlier verdict is held back until a new check runs.

    • cites Horstmann & Boußard, 2024

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisecitation

    Previous research suggests that fantasizing can help ignore potential glitches that show the artificiality of the partner.

    “which fits with previous research suggesting that fantasizing can help ignore potential glitches that show the artificiality of the partner (Szczuka et al., 2019).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Szczuka et al., 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • It may state two separate claims as one.

    Premisecitation

    Users can shape the chatbot's personality and appearance, allowing for greater personalization.

    “Users can also shape the chatbot´s personality and appearance which allows for greater personalization (Brandtzaeg & Følstad, 2017;Locatelli, 2022)”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Brandtzaeg & Følstad, 2017

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Locatelli, 2022

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • It may state two separate claims as one.

    Interpretationcitation

    The distress participants reported when confronted with the impossibility of meeting their chatbot offline underscores that these bonds are experienced as psychologically real and emotionally consequential, even though users typically retain clear awareness that the agent is non-human.

    “The distress participants reported when confronted with the impossibility of meeting their chatbot offline underscores that these bonds are experienced as psychologically real and emotionally consequential, even though users typically retain clear awareness that the agent is non-human (Skjuve et al., 2021).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Skjuve et al., 2021

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • It may state two separate claims as one.

    Premisecitation

    Recent large-scale evidence reports statistically significant but small effects on loneliness, and other studies have found null or inconsistent associations, suggesting that loneliness should be treated as one candidate mechanism among several rather than the dominant driver.

    “Recent large-scale evidence reports statistically significant but small effects on loneliness, and other studies have found null or inconsistent associations, suggesting that loneliness should be treated as one candidate mechanism among several rather than the dominant driver (Fang et al., 2025).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Fang et al., 2025

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim. The paper’s conclusions may not depend on it.

    Premisecitation

    Individuals with avoidant attachment styles have difficulties forming and maintaining intimate relationships.

    “the findings on avoidant attachment mirror those on human-human relationship development, which found individuals with avoidant attachment styles to have difficulties forming and maintaining intimate relationships (Simpson, 1990;Candel & Turliuc, 2019).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Simpson, 1990 source
    • cites Candel & Turliuc, 2019 source
51 of 51

  • The paper relies on this without support.

    The empirical role of fantasy in human-chatbot bonds remains unexplored.

    “Yet despite the theoretical importance of fantasy in human-chatbot bonds, its empirical role remains unexplored.”

  • The paper relies on this without support.

    Romantic human-chatbot relationships remain largely under-researched.

    “Despite this public interest and ongoing debate, romantic human-chatbot relationships remain largely under-researched.”

  • The paper relies on this without support.

    The achieved sample of 92 participants is sufficient to test the hypotheses despite the power analysis indicating a minimum of 99.

    “This target sample size of 99 not reached, as a larger number of answers that anticipated had to be removed.”

  • The paper relies on this without support.

    No prior research has examined how users actually conceptualize and integrate romantic and sexual fantasies into their chatbot interactions.

    “However, no prior research has examined how users actually conceptualize and integrate these fantasies into their chatbot interactions.”

  • The paper relies on this without support.

    This study is the first to examine the importance of romantic fantasy for human-chatbot relationships.

    “This study is the first to examine the importance of romantic fantasy for humanchatbot relationship, finding strong convincing evidence for its importance.”

  • The paper relies on this without support.

    This study is the first to investigate sexual sensation seeking and sexual fantasy in chatbot relationships.

    “This study is the first to investigate sexual sensation seeking and sexual fantasy in chatbot relationships, revealing a connection between the two but suggesting they do not predict human-chatbot relationship intensity.”
6 of 6

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • ConsistentF(3, 88) = 35.63reported p < .001 · recomputed p < .001

    “F(3, 88) =35.63, p < .001”

1 of 1

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Reported exclusions do not yield the final sample size

    Major

    Removing 415 responses from 569 leaves 154, not the reported final sample of 92. The account leaves 62 responses unexplained.

    Reported 92 · recomputed 154

  • Romantic fantasy-avoidant attachment correlation differs across the table

    Major

    The two entries for the same correlation are −.60 and −.06. This discrepancy bears on the interpretation of avoidant attachment as a predictor.

  • Loneliness-avoidant attachment correlation changes sign

    Major

    The table reports −.41 in the loneliness row and .41 in the avoidant attachment row for the same correlation.

  • Sexual fantasy-avoidant attachment correlation changes sign

    Minor

    The table reports this correlation as .27 in one direction and −.27 in the other.

  • Loneliness-anxious attachment correlation changes sign

    Minor

    The table gives −.36 for loneliness against anxious attachment but .36 for anxious attachment against loneliness.

5 of 5

Citations

Cites

Flagged first, then by citations

Research works printed but not linked: 19

  1. Allen, C. M. (2024). My AI Companion: An Examination of the Removal of Erotic Role Play from Replika Through User Discussion on Reddit. (Master's Thesis, University of Nebraska-Lincoln). Digital Commons
  2. Aoki, B. Y. and T. Kimura (2021). Sexuality and affection in the time of technological innovation: Artificial partners in the Japanese context. Religions 12(5): 296.
  3. Banks, J. and J. Van Ouytsel (2020). Cybersex with human-and machine-cued partners: gratifications, shortcomings, and tensions. Technology, Mind, and Behavior 1(1): 1-13. https://doi.org/10.1037/tmb0000008#supplemental-materials · DOI
  4. Brandtzaeg, P. B. and A. Følstad (2017). Why people use chatbots. Internet Science: 4th International Conference, INSCI 2017, Thessaloniki, Greece, November 22-24, 2017, Proceedings 4, Springer.
  5. Erickson, S. E. & S. Dal Cin (2018). Romantic parasocial attachments and the development of romantic scripts, schemas and beliefs among adolescents. Media Psychology 21(1): 111-136. https://doi.org/10.1080/15213269.2017.1305281Fang, C. M., Liu, A. R., Danry, V., Lee, E., Chan, S. W. T., Pataranutaporn, P., Maes, P., Phang, J., Lampe, M., Ahmad, L., & Agarwal, S. (2025). How AI and human behaviors shape psychosocial effects of extended chatbot use: A longitudinal randomized controlled study [Preprint]. arXiv. https://arxiv.org/abs/2503.17473 · DOI
  6. Kalichman, S. C. (2013). Sexual sensation seeking scale. Handbook of sexuality- related measures, Routledge: 586-587.
  7. Lehmiller, J. J. (2018). Tell me what you want. Da Capo Press.
  8. Lehmiller, J. J., & Gormezano, A. M. (2023). Sexual fantasy research: A contemporary review. Current Opinion in Psychology, 49, 101496.
  9. Ma, S., & Koike, M. (2025). Becoming Human: How Perceived Anthropomorphism and Authenticity Influence Romantic Attractiveness of Chat Agents. International Journal of Human-Computer Interaction, 1-14. https://doi.org/10.1007/s10551-022-05059-4 · DOI
  10. Maples, B., Cerit, M., Vishwanath, A., & Pea, R. (2024). Loneliness and suicide mitigation for students using GPT3-enabled chatbots. npj mental health research, 3(1), 4.
  11. McAfee, A., Rock, D., & Brynjolfsson, E. (2023). How to capitalize on generative AI. Harvard Business Review, 101(6), 42-48.
  12. Nass, C. and Y. Moon (2000). "Machines and mindlessness: Social responses to computers." Journal of social issues 56(1): 81-103.Romantic and Sexual Fantasy in Human- Chatbot Relationships 32
  13. Nass, C. and Y. Moon (2000). Machines and mindlessness: Social responses to computers. Journal of social issues 56(1): 81-103.
  14. Shum, H.-Y., et al. (2018). From Eliza to XiaoIce: challenges and opportunities with social chatbots. Frontiers of Information Technology & Electronic Engineering 19(1), 10-26.
  15. Siemon, D., Strohmann, T., Khosrawi-Rad, B., de Vreede, T., Elshan, E., & Meyer, M. (2022). Why Do We Turn to Virtual Companions? A Text Mining Analysis of Replika Reviews. Proceedings of the AMCIS. 10.https://aisel.aisnet.org/amcis2022/sig_hci/sig_hci/10
  16. Skjuve, M., et al. (2021). My chatbot companion-a study of human-chatbot relationships. International Journal of Human-Computer Studies 149, 102601.
  17. Starke, C., Ventura, A., Bersch, C., Cha, M., de Vreese, C., Doebler, P., ... & Köbis, N. (2024). Risks and protective measures for synthetic relationships. Nature Human Behaviour, 8(10), 1834-1836.
  18. Tukachinsky, R. (2010). Para-romantic love and para-friendships: Development and assessment of a multiple-parasocial relationships scale. American Journal of Media Psychology, 3(1/2), 73-94.
  19. Young, T. M. (2019). Developing and Implementing a Measurement Tool for Sexual, Romantic, and Sexual-Romantic Fantasies. (Publication No. 13810800). [Doctoral dissertation, New Mexico State University] ProQuest Dissertations & Theses.

Cited by

Sorted by citations

Related

Metadata

Authors
Ebner, P., & Szczuka, J.
ORCID
Venue
arXiv
Publisher
arXiv
Published
2025
Volume
7
Issue
2
Pages
83-97
ISSN
2689-0208 (ISSN-L)
DOI
10.48550/arxiv.2503.00195
arXiv
2503.00195
OpenAlex
W7160523608
Cited by
33
References
56
Citations by year
2026: 29 · 2025: 4
FWCI
125.50
Type
journal article
Language
en
Open access
Licence
Locations
  • Technology Mind and Behavior (publishedVersion, cc-by-nc-nd)
  • arXiv (Cornell University) (submittedVersion)
  • Universitätsbibliographie, Universität Duisburg-Essen (submittedVersion)
  • DuEPublico (University of Duisburg-Essen) (submittedVersion, cc-by-nc-nd)
  • arXiv (Cornell University)
Topics
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Social Sciences › Psychology › Social Psychology › Attachment and Relationship Dynamics
  • Social Sciences › Business, Management and Accounting › Marketing › AI in Service Interactions
Keywords
romantic fantasy, sexual fantasies, relationship intensity, anthropomorphism, attachment orientation, attachment avoidance, loneliness, sexual sensation seeking, LLM-based chatbots, social companionship

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review