Truace
Subscribe
50/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Survey of 395 chatbot users in China links heavier reliance to lower resilience

Study title: AI Chatbot Dependence as an Indirect Link Between Social Anxiety and Psychological Resilience: The Moderating Role of Perceived Social Support

Evidence type
Survey
Sample
n = 395
Sources
1 sourcePeer-reviewed
Published
3 min read
Truvace record
TRV-2026-2553cite or check it

Study title: AI Chatbot Dependence as an Indirect Link Between Social Anxiety and Psychological Resilience: The Moderating Role of Perceived Social Support

A cross-sectional survey of 395 adult AI chatbot users in China found that people with higher social anxiety reported more reliance on chatbots and lower psychological resilience. Chatbot dependence was itself linked to lower resilience after social anxiety was accounted for (b = -0.117, p = 0.005), producing a small negative indirect association of -0.070.

Read the study abstract

Generative artificial intelligence has transformed AI chatbots into systems that can provide information, emotional feedback, and social interaction. Although prior research has examined AI adoption, trust, and continuance intention, less is known about sustained reliance on AI chatbots and its relationship with psychological adaptation. Drawing on an integrated compensation-resource framework, this study examined whether social anxiety was indirectly associated with psychological resilience through AI chatbot d…

The problem

In a survey of 395 adult chatbot users in China, greater chatbot dependence was associated with lower psychological resilience even after accounting for social anxiety, a small indirect effect of -0.070.

In brief

Social anxiety was strongly tied to both outcomes: it correlated with chatbot dependence at r = 0.623 and with resilience at r = -0.718. The unique link between dependence and resilience was small (f2 = 0.021), and the direct anxiety-resilience link remained large, so most of the pattern did not run through chatbots.

Perceived social support softened the pattern but did not remove it. The interaction added only 0.7% of explained variance in dependence, and the indirect association stayed significantly different from zero at low, average and high support. Because all variables were measured at one time point, the study cannot say whether reliance lowers resilience, lower resilience drives reliance, or both reinforce each other.

Main points

  1. The sample was 395 adults who had used a chatbot such as ChatGPT, Claude, Gemini or Character.AI in the previous six months, recruited through the Wenjuanxing platform in China; 10 of 405 responses were excluded.
  2. Social anxiety predicted chatbot dependence (β = 0.623) and lower resilience (β = -0.640); social anxiety alone explained 38.8% of the variance in dependence.
  3. After accounting for social anxiety, chatbot dependence was negatively linked to resilience (β = -0.126), a small unique effect (f2 = 0.021).
  4. The indirect association through dependence was -0.070 (95% CI -0.120 to -0.022), while the direct anxiety-resilience link stayed at b = -0.573.
  5. At low, average and high social support the indirect estimates were -0.066, -0.057 and -0.049, all with confidence intervals excluding zero; the index of moderated mediation was 0.008.
  6. Dependence was measured with an eight-item scale covering reliance for information, emotional support and everyday decisions, and the authors say a high score is not evidence of addiction or problematic use.
  7. The authors say the data do not justify a specific regulatory measure or clinical treatment, and they call for longitudinal and experimental follow-up.

The full story

Participants completed validated scales: SIAS-6 for social anxiety, the eight-item AICDS for chatbot dependence, the 12-item MSPSS for perceived social support and the CD-RISC-10 for resilience. Responses were analysed with regression-based path analysis and PROCESS Model 7 using 5,000 bootstrap resamples. Cronbach's alpha ranged from 0.905 to 0.952.

The largest effects involved social anxiety rather than chatbots. Social anxiety and resilience correlated at r = -0.718, and the resilience model explained 52.6% of variance. Chatbot dependence correlated with resilience at r = -0.525, but its unique contribution shrank once anxiety was included.

Perceived social support was negatively associated with dependence (b = -0.171), and the Social Anxiety × Perceived Social Support interaction was b = -0.066 (p = 0.033). The authors describe this buffering as statistically significant but modest and say replication in independent samples is needed.

The authors suggest universities, organisations and mental-health services might ask what functions chatbots serve in daily life, not only how often people use them. They also say features such as usage reminders or offline-interaction prompts are design hypotheses that have not been tested.

Sources

1 on file
  1. Behavioral Sciencesdoi:10.3390/bs16091505
    Peer-reviewed2026-08-27

Talk it over

Four questions to start onWho carries this cost, and what would have to change to prevent it?

Start with a question

Survey of 395 chatbot users in China links heavier reliance to lower resiliencehttps://truvace.com/item/ai-chatbot-dependence-as-an-indirect-link-between-social-anxiety-and-psychologic

  1. Who carries this cost, and what would have to change to prevent it?

    The problem In a survey of 395 adult chatbot users in China, greater chatbot dependence was associated with lower psychological resilience even after accounting for social anxiety, a small indirect effect of -0.070.

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a survey with a sample of 395, from Behavioral Sciences.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat All variables were self-reported at a single time point in one Chinese online sample, so the study cannot establish causal or temporal direction, and reverse or reciprocal relationships remain plausible.

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Health