Truace
Subscribe
63/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

In a study of 1,200 Chinese students, AI chatbot support eased loneliness only for those with strong real-life support

Study title: Perceived support from AI chatbots and loneliness among higher vocational college students: a short-term three-wave study of social interaction anxiety and perceived social support

Evidence type
Observational study
Sample
n = 1,200
Sources
1 sourcePeer-reviewed
Published
3 min read
Truvace record
TRV-2026-2540cite or check it

Study title: Perceived support from AI chatbots and loneliness among higher vocational college students: a short-term three-wave study of social interaction anxiety and perceived social support

A three-wave survey of 1,200 students at four higher vocational colleges in Jiangxi and Guangdong, China, tracked how supported students felt when using AI chat tools such as ChatGPT, Doubao, ERNIE Bot, Kimi, Tongyi Qianwen or DeepSeek. Waves were about 8 weeks apart during the 2025-2026 spring semester. Higher perceived AI support at the second wave predicted lower loneliness at the third wave.

Read the study abstract

Generative artificial intelligence and AI companion tools are increasingly entering higher vocational college students' learning, emotion regulation, and everyday social communication contexts. This study defines perceived support from AI chatbots as students' subjective experience of receiving support while using AI chat tools, rather than objective social support from real interpersonal sources. A short-term three-wave design was used. Students were recruited through class WeChat groups from four higher vocati…

56Gain side

How the two sides scoreClose call

59Problem side

The gain

In a three-wave study of 1,200 vocational college students, those who felt more supported by AI chatbots at the second wave reported less loneliness about 8 weeks later, mainly if they had strong real-life support.

The problem

In the same study, students who felt more supported by AI chatbots reported higher social interaction anxiety at the next wave, which in turn predicted more loneliness later.

A fixed formula sets both scores from the evidence and its sources. Reader votes don’t move them.

What the scores mean

G is the gain side, out of 100. It goes up when the gain is specific, when it measurably helped people and when the sources behind it are strong.

P is the problem side, out of 100. It goes up when the problem is specific, when it measurably hurt people and when the sources behind it are strong.

In brief

The same perceived support also predicted higher social interaction anxiety at the following wave, and that anxiety predicted more loneliness. The indirect path from first-wave AI support through second-wave anxiety to third-wave loneliness had a standardized effect of 0.038 (95% CI 0.026 to 0.050).

Real-life support changed the picture. Among students with high baseline perceived social support, AI support predicted less later loneliness (slope -0.287); among those with low support the slope was -0.023 and not significant. The authors say the low-support group was not harmed but showed no compensatory benefit. The design is observational, so the results show associations, not cause and effect.

Main points

  1. The valid first-wave sample was 1,500 students; 1,200 completed all three waves, an 80.0% retention rate, and 60.2% of them were female with a mean age of 19.43.
  2. Students used AI chat tools an average of 38.5 minutes a day; 5.2% used them more than 2 hours daily.
  3. Second-wave AI support predicted lower third-wave loneliness (beta = -0.166, p < 0.001), but first-wave support did not significantly predict second-wave loneliness (beta = -0.043, p = 0.064).
  4. Each one-standard-deviation rise in second-wave AI support corresponded to roughly 0.92 points lower loneliness on the 8 to 32 scale.
  5. AI support predicted higher social interaction anxiety at the next wave (beta = 0.161 and 0.158), and each one-standard-deviation rise in anxiety corresponded to about 1.30 points more loneliness.
  6. Baseline real-life support moderated the link to loneliness (beta = -0.132, p < 0.001) and to social anxiety (beta = -0.078, p < 0.001).
  7. Between 55.0% and 72.9% of students scored at or above the PHQ-9 threshold for moderate depressive symptoms across the waves.

The full story

Students were recruited through class WeChat groups by course instructors and had to report using at least one generative AI chat tool in the prior 3 months. Of 1,639 who opened the survey, 1,500 passed screening and validity checks. Loneliness was measured with the 8-item UCLA scale, social anxiety with the Social Interaction Anxiety Scale, and real-life support with the MSPSS at first wave only.

AI support was measured with a 20-item online social support scale adapted to AI chat tools. The authors treat it as students' subjective sense of being supported, not real interpersonal support. An initial cross-lagged model fit poorly (RMSEA 0.089, TLI 0.880), and a revised model with added second-order paths fit better (RMSEA 0.077, TLI 0.910).

Attrition checks found no significant differences between students who completed all waves and those who dropped out, and the direction of the main paths held across three covariate specifications. Baseline real-life support also strongly predicted lower third-wave loneliness (beta = -0.449). The authors conclude that AI literacy and mental health education should connect AI-based support with real social support networks.

What this doesn’t fix

Because intact classes were recruited and no class identifiers were kept, clustering could not be adjusted for, so p values may be overly optimistic; the revised model was changed after poor initial fit, measures were self-reported, and the results are associations, not proof of cause.

Sources

1 on file
  1. Frontiers in Psychologydoi:10.3389/fpsyg.2026.1928350
    Peer-reviewed2026-09-15

Talk it over

Four questions to start onThis record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?

Start with a question

In a study of 1,200 Chinese students, AI chatbot support eased loneliness only for those with strong real-life supporthttps://truvace.com/item/perceived-support-from-ai-chatbots-and-loneliness-among-higher-vocational-colleg

  1. This record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?

    The gain In a three-wave study of 1,200 vocational college students, those who felt more supported by AI chatbots at the second wave reported less loneliness about 8 weeks later, mainly if they had strong real-life support.

    The problem In the same study, students who felt more supported by AI chatbots reported higher social interaction anxiety at the next wave, which in turn predicted more loneliness later.

    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 an observational study with a sample of 1,200, from Frontiers in Psychology.

    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 Because intact classes were recruited and no class identifiers were kept, clustering could not be adjusted for, so p values may be overly optimistic; the revised model was changed after poor initial fit, measures were self-reported, and the results are…

    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 Education