This analytical cross-sectional study was conducted in Saudi Arabia to compare perceptions of ChatGPT, including perceived usefulness, perceived risk, anxiety, and social influence, on health attitudes among ChatGPT users and nonusers.

In total, 337 participants aged 18 years or older, residing in Saudi Arabia and having internet access, were recruited through an online survey. The single-population proportion formula was used to determine the required sample size, with a 95% CI (Z) and a 5% margin of error (E). After reviewing the existing literature to determine the smallest sample size (p), we found a study that reported a positive health care–related view of ChatGPT prevalence of approximately 70% [20]. Using the following formula, the minimum required sample size was calculated to be 323:

The sample size was chosen to estimate the overall prevalence and characteristics of ChatGPT use within the target population, rather than to aim for an equal number of users and nonusers. In this study, the participants were classified into two groups: (1) ChatGPT users, defined as respondents who reported having used ChatGPT, and (2) ChatGPT nonusers, defined as respondents who had not used ChatGPT but were aware of and familiar with it (eg, through media exposure, discussions in academic classes, or peers). This approach ensured that all participants included in the analysis had sufficient awareness of ChatGPT to provide meaningful responses.

The inclusion criteria were participants aged 18 years or older, living in Saudi Arabia, and having access to the internet. Participants who reported no awareness or familiarity with ChatGPT were excluded from the analysis. This classification resulted in an analytic sample of 255 participants (n=194, 76.1% users and n=61, 23.9% nonusers).

The questionnaire used in this study was adapted from a previously validated instrument based on a modified TAME-ChatGPT (Technology Acceptance Model Edited to Assess ChatGPT Adoption), which demonstrated acceptable reliability in prior studies (Cronbach α≥0.71) [21]. In this study, the adapted questionnaire was pilot-tested to assess clarity and comprehension before deployment, as shown in Multimedia Appendix 1.

The survey measured ChatGPT perception on four constructs that corresponded to the research aim: (1) perceived usefulness, which reflects participants’ beliefs that ChatGPT is useful for accessing and supporting health-related information; (2) perceived risk, which includes threats related to privacy, security, and reliability when accessing and supporting health-related information; (3) anxiety, which refers to fear or worry about using ChatGPT for health-related purposes; and (4) social influence, which captures the perceived influence of peers or family on participants’ readiness or recommendation to use the tool.

All items were evaluated using a 5-point Likert scale (from “strongly disagree” to “strongly agree”), with higher summed scores of relevant items indicating greater support for the construct.

Sociodemographic information and health-related characteristics were also collected in the questionnaire, considering the status of ChatGPT users and nonusers who were aware of ChatGPT but had not used it. This method enabled comparison of perceptions and attitudes toward ChatGPT between user and nonuser groups in the context of health-related decision-making.

Data collection was conducted from mid-November 2023 to February 2024 using a convenience sampling technique. Participants were recruited via official university email systems and social media platforms. Email invitations were distributed across Imam Abdulrahman bin Faisal University (Eastern province), Umm Al-Qura University (Makkah province), Shaqra University, and Prince Sultan University (Riyadh province) to capture diverse geographic locations and perspectives. Additional invitations were shared via Telegram (12.6.2) and WhatsApp (version ; Meta Platforms Inc) groups.

In this study, SPSS (version 23.0; IBM Corp) was used for all data analysis. Descriptive statistics, including frequencies, percentages, means, and SD, were used to summarize the characteristics of participants and the overall scores of perception. Additionally, the χ² test was used to evaluate relationships between ChatGPT use status (users vs nonusers) and sociodemographic variables. Furthermore, factor scores were summarized using descriptive statistics, including means (SDs), medians (IQRs), and ranges, to describe the central tendency and dispersion of responses, as the primary goal was to characterize the distribution of perceptions across factors. Both mean (SD) and median (IQR) were reported for all perception factor scores to consistently summarize central tendency and dispersion and to provide a complete description of their distribution, regardless of normal distribution. The level of statistical significance was set at P≤.05.

The institutional review board of Imam Abdulrahman bin Faisal University granted ethics approval (IRB-HAP-05-D-003) for this study. All participants provided informal consent before completing the survey. Participant privacy and confidentiality were maintained, with all data anonymized and access restricted to the research team. No financial compensation was provided to participants.

This study analyzed views on ChatGPT as a health information–seeking tool among users and nonusers in Saudi Arabia. The findings indicate that most participants reported traditional health authorities as their most frequently cited sources of health information. The use of ChatGPT differed across participants, with some reporting its use for health education and others for physical activity. Fewer participants reported using it for psychological counseling, sleep-related problems, and medical consultation. Although this has recently become a topic of discussion, research indicates progress in this area [22]. However, nutrition advice and social counseling were found to be more effective. This aligns with studies suggesting that AI-powered chatbots can significantly improve dietary guidance [23]. These results indicate that ChatGPT might not yet be a substitute for traditional health information sources but rather an additional source of health-related information.

It has been noted that the use of ChatGPT was common across different sociodemographic groups, and differences between users and nonusers were not found to be significant. Younger participants (age ≤24 years) and those with doctoral degrees were slightly more likely to be using ChatGPT; however, these differences were not statistically significant. According to previous research, higher educational and socioeconomic levels may also affect the use of AI-based applications such as ChatGPT [24-26]. Although age has been reported to influence willingness to embrace AI technologies [27], this effect was not observed in this study. ChatGPT was used for health education and instructions on physical activity; however, the study did not evaluate whether the information prompted actual behavior change.

Prior research investigating perceptions of usefulness, ease of use, risk, and psychosocial issues has indicated a generally positive attitude toward ChatGPT [28]. Other studies involving health care professionals and students have yielded similar results, with perceived usefulness moderated by reliability and perceived risks [12,29]. These results are consistent with this study, in which respondents reported moderate perceived usefulness of ChatGPT and concerns about privacy, security, and misinformation. Although prior studies indicated that social and age-related variables could contribute to the adoption of AI technologies [30,31], the data in this study did not show statistically significant relationships between age and ChatGPT use.

In this study, a large number of ChatGPT users found it useful and relatively easy to use, but very few said that they used it when making health-related decisions. There were also concerns regarding the accuracy, privacy, and security of information, especially among nonusers. These issues align with previous studies showing that users remain skeptical about trusting AI-generated information and about using it in medical practice [32,33].

Even though numerous participants were interested in technologies such as ChatGPT and reported positive expectations of its usefulness, the findings should be interpreted in light of the descriptive nature of the data. Perceived usefulness was more likely to be accentuated among users, whereas perceived risk was more relevant among nonusers.

This study has a number of limitations. First, the cross-sectional design does not permit any causal inference. Second, the use of convenience sampling can introduce selection bias and reduce the representativeness of the sample. Third, the study used self-reported data, which could be affected by recall bias, reporting bias, and social desirability bias.

Despite these limitations, this study provides insight into people’s perception of ChatGPT as a health information–seeking tool. Although the participants were aware of its potential usefulness, concerns regarding privacy, security, and information reliability remained important considerations. Future studies should consider using more diverse and representative samples and investigating how perceptions of AI-based health information tools change over time.

The study investigated perceptions of ChatGPT as a health information–seeking tool among users and nonusers in Saudi Arabia. ChatGPT was viewed as a convenient and accessible source of health-related information, especially for health education and assistance with physical activity. However, concerns regarding privacy, security, and reliability of information were raised by both users and nonusers. These findings indicate that interest in AI-based tools such as ChatGPT exists, but issues related to trust, data privacy, and information quality can arise when they are used to seek health-related information.

There is a need for further investigation into AI-driven health care decision-making, particularly regarding privacy, trust, and long-term adoption patterns. Policymakers should focus on enhancing security measures, ensuring accuracy, and fostering trust to facilitate the responsible use of AI tools in the health care system.