CoChat

Investigating the Impact of User Trust on the Adoption and Use of ChatGPT: Survey Analysis

Open access
Choudhury, Avishek, Shamszare, Hamid · 2023 · Journal of Medical Internet Research

Published version of a preprint

The DOI matches its Crossref record; 52 of the 52 references it cites were found in a registry or an index; 3 of the 32 sources its claims cite were checked for support.

Abstract

Background: ChatGPT (Chat Generative Pre-trained Transformer) has gained popularity for its ability to generate human-like responses. It is essential to note that overreliance or blind trust in ChatGPT, especially in high-stakes decision-making contexts, can have severe consequences. Similarly, lacking trust in the technology can lead to underuse, resulting in missed opportunities. Objective: This study investigated the impact of users’ trust in ChatGPT on their intent and actual use of the technology. Four hypotheses were tested: (1) users’ intent to use ChatGPT increases with their trust in the technology; (2) the actual use of ChatGPT increases with users’ intent to use the technology; (3) the actual use of ChatGPT increases with users’ trust in the technology; and (4) users’ intent to use ChatGPT can partially mediate the effect of trust in the technology on its actual use. Methods: This study distributed a web-based survey to adults in the United States who actively use ChatGPT (version 3.5) at least once a month between February 2023 through March 2023. The survey responses were used to develop 2 latent constructs: Trust and Intent to Use, with Actual Use being the outcome variable. The study used partial least squares structural equation modeling to evaluate and test the structural model and hypotheses. Results: In the study, 607 respondents completed the survey. The primary uses of ChatGPT were for information gathering (n=219, 36.1%), entertainment (n=203, 33.4%), and problem-solving (n=135, 22.2%), with a smaller number using it for health-related queries (n=44, 7.2%) and other activities (n=6, 1%). Our model explained 50.5% and 9.8% of the variance in Intent to Use and Actual Use, respectively, with path coefficients of 0.711 and 0.221 for Trust on Intent to Use and Actual Use, respectively. The bootstrapped results failed to reject all 4 null hypotheses, with Trust having a significant direct effect on both Intent to Use (β=0.711, 95% CI 0.656-0.764) and Actual Use (β=0.302, 95% CI 0.229-0.374). The indirect effect of Trust on Actual Use, partially mediated by Intent to Use, was also significant (β=0.113, 95% CI 0.001-0.227). Conclusions: Our results suggest that trust is critical to users’ adoption of ChatGPT. It remains crucial to highlight that ChatGPT was not initially designed for health care applications. Therefore, an overreliance on it for health-related advice could potentially lead to misinformation and subsequent health risks. Efforts must be focused on improving the ChatGPT’s ability to distinguish between queries that it can safely handle and those that should be redirected to human experts (health care professionals). Although risks are associated with excessive trust in artificial intelligence–driven chatbots such as ChatGPT, the potential risks can be reduced by advocating for shared accountability and fostering collaboration between developers, subject matter experts, and human factors researchers.

Overview

CoChat Summary

TL;DR

This survey of 607 US adults using ChatGPT 3.5 found trust significantly predicted both intent to use and actual use, with intent partially mediating trust’s effect on actual use. The authors argue trust is critical for adoption but caution against overreliance, especially for health-related queries.

Key findings

  • Trust had significant direct effects on Intent to Use (β=0.711, 95% CI 0.656-0.764) and Actual Use (β=0.302, 95% CI 0.229-0.374); the model explained 50.5% and 9.8% of the variance in Intent to Use and Actual Use, respectively.
  • Intent to Use significantly predicted Actual Use (β=0.114, 95% CI 0.001-0.229), and the indirect effect of Trust on Actual Use partially mediated by Intent was significant (β=0.113, 95% CI 0.003-0.227).
  • Primary uses were information gathering (n=219, 36.1%), entertainment (n=203, 33.4%), problem-solving (n=135, 22.2%), health-related queries (n=44, 7.2%), and other activities (n=6, 1%).
  • Frequency of use was at least once a month for 182 (30%), once per week for 158 (26%), more than once per week for 149 (24.5%), and almost every day for 118 (19.4%).

Methods

  • Web-based semistructured survey distributed to US adults who actively used ChatGPT version 3.5 at least once a month; data collected February through March 2023, with 607 respondents completing the survey.
  • Two latent constructs, Trust and Intent to Use, were developed from Likert-scale items (1=strongly disagree to 4=strongly agree); Actual Use was the outcome, captured with a single-item frequency question (1=once a month to 4=almost every day).
  • Analyses used partial least squares structural equation modeling in R with the seminr package, including convergent and discriminant validity checks and bootstrapped hypothesis tests.

Limitations

  • The authors state the study used a cross-sectional survey and self-report measures, which may introduce biases; the US-focused sample may limit generalizability to other cultural contexts. They recommend future longitudinal data, cross-cultural exploration, and controlling for confounders such as AI familiarity, prior chatbot experience, and demographics.
  • Evident from the text: Actual Use was measured with a single item, and the reported Trust→Actual Use coefficient varies across the abstract/Table 2/results (0.221 vs 0.302), indicating some inconsistency in reporting.

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

Integrity signals

Registry

Crossref · 10.2196/47184

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 The conceptual structural framework. H1 through H4 indicate the hypotheses. The dashed line connecting trust and actual use indicates the indirect effect, whereas solid lines indicate the direct paths.
Figure 1 The conceptual structural framework. H1 through H4 indicate the hypotheses. The dashed line connecting trust and actual use indicates the indirect effect, whereas solid lines indicate the direct paths.CC BY · PubMed Central
Figure 2 Conceptual framework illustrating the significant paths connecting trust in ChatGPT (Chat Generative Pre-trained Transformer), users' intent to use ChatGPT, and its actual use (AU). T1 through T7: factors for trust; U1 through U3: factors for intent to use. *P<.05 and ***P<.001.
Figure 2 Conceptual framework illustrating the significant paths connecting trust in ChatGPT (Chat Generative Pre-trained Transformer), users' intent to use ChatGPT, and its actual use (AU). T1 through T7: factors for trust; U1 through U3: factors for intent to use. *P<.05 and ***P<.001.CC BY · PubMed Central

Verification

Claims the paper makes

2 Valid · 0 Not valid · 16 Not checked · 37 Own results and interpretations

19 more claims

Create a free account to see all 55 claims and to verify the 29 references still open.

See all claims — free account
  • It may state two separate claims as one. 9
  • Probably not a substantial claim. 25

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

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.

4 of 4

Citations

Cites

Show all 52 references

Research works printed but not linked: 3

  1. Whalen, J, 1995, Cambridge Series on Human Computer Interaction, 161
  2. Kuberkar, S, 2020, International Journal of Emerging Technologies in Learning, 11, 948
  3. 10.1056/cat.23.0043 · doi:10.1056/cat.23.0043

Cited by

Items 401–500 of 569

Metadata

Authors
Choudhury, A., & Shamszare, H.
Affiliations
West Virginia University (US)
Corresponding author
Choudhury, Avishek
Venue
Journal of Medical Internet Research
Published
2023-06-14
Volume
25
Pages
e47184
ISSN
1438-8871 (ISSN-L)
OpenAlex
W4378214315
Cited by
470
References
51
Citations by year
2026: 143 · 2025: 174 · 2024: 115 · 2023: 37
FWCI
120.63
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • Journal of Medical Internet Research (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Social Sciences › Decision Sciences › Information Systems and Management › Technology Adoption and User Behaviour
  • Social Sciences › Business, Management and Accounting › Marketing › AI in Service Interactions
  • Social Sciences › Human Factors and Ergonomics › Human-Automation Interaction and Safety
Keywords
ChatGPT, user trust, technology adoption, technology use, intention to use, actual use, large language models, generative AI, information gathering, problem solving, entertainment use, overreliance on AI, high-stakes decision making, AI safety, AI governance, AI regulation, AI literacy
MeSH terms
Adult, Artificial Intelligence, Health Personnel, Humans, Surveys and Questionnaires, Intention, Trust

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash).

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

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.