Development and Validation of an Artificial Intelligence Awareness, Knowledge, Attitude, and Confidence Survey Among Physiotherapy Students

INTRODUCTION

Artificial intelligence (AI) has created transformative impacts in healthcare, as in many other fields, in recent years. First defined in 1956 by John McCarthy as “the science and engineering of making intelligent machines,” AI today refers to systems capable of performing tasks that require human intelligence, such as visual perception, speech recognition, decision-making, and natural language processing.^[1,2] In healthcare, AI has a wide range of applications, from diagnostic processes and patient monitoring to personalized treatment approaches and reducing error rates in medical imaging. Particularly in fields such as radiology and diagnostic imaging, AI algorithms with high sensitivity have been reported to produce meaningful improvements in clinical outcomes.^[3,4]

The impact of AI is not limited to medicine alone; it has increasingly been felt in physiotherapy practices, including motion analysis, monitoring rehabilitation processes, the use of robotic rehabilitation devices, and the design of individualized treatment plans.^[5-8] These systems, operating with layered algorithmic architectures, go beyond pattern recognition and association functions to assist healthcare professionals in processes such as test guidance, patient screening, and treatment planning.^[9]

The effective use of AI technologies largely depends on healthcare professionals’ levels of knowledge, awareness, and attitudes toward technology.

Global studies indicate that health sciences students generally hold positive attitudes toward the use of AI in healthcare, although their technical knowledge often remains limited.^[10,11] Some students perceive AI as a potential threat to their professions; nevertheless, they consider the inclusion of basic AI education in curricula beneficial.^[3] Cross-national comparative research further shows that students question the potential impact of AI on professional roles, yet remain willing to integrate this technology into practice when appropriate training is provided.^[12]

Although the role of AI in disciplines such as physiotherapy and rehabilitation, where patient-centred care and manual practices are central, remains debated, its potential contributions in these areas are becoming increasingly evident. The literature highlights a scarcity of studies focusing specifically on physiotherapy students’ knowledge, awareness, attitudes, and confidence regarding AI. However, these factors have the potential to shape the future direction of clinical practice and directly influence the integration of technology into healthcare delivery.^[3,13,14]

This study aims to evaluate physiotherapy and rehabilitation students’ levels of knowledge, awareness, attitudes, and confidence regarding AI and to examine the relationships among these variables. The findings are expected to contribute to the development of physiotherapy education programs in terms of AI integration and to inform strategies for enhancing students’ confidence toward technology. Based on this rationale, we hypothesized that physiotherapy and rehabilitation students’ levels of knowledge, awareness, and attitude-confidence regarding artificial intelligence would be significantly interrelated; that these levels would vary according to demographic factors such as age, gender, grade level, and technological experience; and that students who had received AI-related training would demonstrate significantly higher levels compared to their peers without such training.

MATERIAL AND METHODS

This research was designed as a cross-sectional and descriptive survey study. The study methodology was developed in accordance with the “Checklist for Reporting of Survey Studies (CROSS)” and the “Checklist for Reporting Results of Internet E-Surveys (CHERRIES)” guidelines, and it was conducted in Istanbul between February 2024 and May 2025. Ethical approval for this study was obtained from the Non-Interventional Scientific Research Ethics Committee of Istanbul Atlas University (decision no: 05/24, dated 26 May 2025). The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Informed consent was obtained digitally via Google Forms from all participants prior to data collection.

Structural equation modelling (SEM)

Model fit indices

Structural equation modelling was conducted, and the goodness-of-fit indices for the proposed model are presented below. The Chi-square statistic was found to be significant (χ² (202) = 495.99, p <0.001); however, given the sensitivity of this test to sample size, other fit indices were considered. The model demonstrated an overall acceptable fit to the data (CFI = 0.916, TLI = 0.904, RMSEA = 0.070). Although the GFI (0.865) and AGFI (0.830) values were slightly below the conventional threshold of 0.90, they were close to the acceptable lower limit of 0.85. Thus, the model was considered marginally but acceptably fitting in these indices. Taken together with the other indices, the results suggest that the model demonstrated an overall valid and adequate fit [Table 1].

Table 1: Model fit indices

Conformity criterion	Value
χ2/DF	2.455
CFI	0.916
TLI	0.904
RMSEA	0.070
Model	CMIN	DF	p	CMIN/DF
Default model	495,991	202	<0.001	2,455
Saturated model	,000	0
Independence model	3718,147	231	<0.001	16,096

p <0.05 statistically significant. χ²/DF: Chi-square divided by degrees of freedom, CFI: Comparative fit index, TLI: Tucker-Lewis Index, RMSEA: Root Mean Square Error of Approximation, CMIN: Minimum drug concentration, DF: Degrees of freedom.

Results of the measurement model

The analysis revealed that the factor loadings of all observed variables on the latent constructs were statistically significant (p <0.001), with most values exceeding 0.50. These findings indicate that the measurement model was adequate [Figure 1].

Close

Figure 1: Confirmatory factor analysis results for the three-factor model.

Export to PPT

The model included three latent variables (knowledge, awareness, and attitude/confidence), each represented by their respective observed indicators. Items associated with the Knowledge construct demonstrated strong factor loadings, whereas some items under the Awareness construct showed relatively lower loadings. Significant positive correlations were identified among the latent variables. In particular, a very strong relationship was observed between Knowledge and Awareness (r = 0.93).

RESULTS

A total of 300 students participated in the study, the majority of whom were female (n = 218, 72.7%). Of the participants, 96% (n = 288) reported that they had not received any education on artificial intelligence either before or during their physiotherapy training, and 80.3% (n = 241) stated that they had not previously received any education in computer science, statistics, or mathematics. Only 7.3% (n = 22) of the participants reported having a family member (parent or sibling) who had received education related to artificial intelligence [Table 2].

Among the artificial intelligence tools most frequently recognized by the participants, ChatGPT ranked first with 31.09%, followed by Apple’s Siri application with 23.42% [Figure 2]. In addition to ChatGPT and Apple’s Siri, students reported familiarity with several other AI tools, including Google Gemini (10.19%), Amazon Alexa (9.66%), Microsoft Azure AI (6.93%), and Grammarly (3.89%). Less frequently recognized tools included DALL·E (1.37%), while 4.05% of participants reported awareness of other AI-based applications. In terms of frequency of use, 33% of the students (n = 99) reported using AI frequently in daily life, whereas 37.3% (n = 112) stated that they used it rarely.

Close

Figure 2: Distribution of artificial intelligence tools recognized by participants.

Export to PPT

In Table 3, the mean, standard deviation, median, and interquartile range values of the total scores obtained from the 5-point Likert-type questions measuring participants’ knowledge levels regarding artificial intelligence in the field of physiotherapy are presented according to sociodemographic variables. The analysis revealed statistically significant differences in the median knowledge scores according to age group (<20 vs. ≥20 years), year of study, and prior education or background in computer science, mathematics, or statistics.

Specifically, participants aged ≥20 years had higher mean knowledge scores (2.14 ± 0.70) compared to those aged <20 years (1.86 ± 0.63). When evaluated by year of study, knowledge scores increased progressively with higher class levels (1st year: 1.95 ± 0.68, 2nd year: 1.89 ± 0.66, 3rd year: 2.31 ± 0.64, 4th year: 2.49 ± 0.63).

In addition, participants with prior education/background in computer science, mathematics, or statistics (2.35 ± 0.82) scored higher than those without such background (2.01 ± 0.65). No significant differences were observed in relation to other demographic variables.

Table 4 presents the mean and standard deviation values of participants’ total awareness scores regarding artificial intelligence in the field of physiotherapy, according to demographic variables. The analysis indicated statistically significant differences in awareness scores based on age group (<20 vs. ≥20 years), year of study, and prior education/background in computer science, mathematics, or statistics. By age group, participants aged ≥20 years had higher mean awareness scores (2.65 ± 0.72) compared to those aged <20 years (2.43 ± 0.67). When analysed by year of study, awareness scores showed a progressive increase with advancing class levels (1st year: 2.56 ± 0.75, 2nd year: 2.48 ± 0.74, 3rd year: 2.64 ± 0.67, 4th year: 2.88 ± 0.49). First-year physiotherapy students demonstrated one of the lowest knowledge and awareness scores compared to upper-year students, suggesting limited familiarity with AI tools during the early stages of physiotherapy education. In addition, participants with prior education/background in computer science, mathematics, or statistics (2.82 ± 0.71) scored higher than those without such background (2.54 ± 0.70).

These findings suggest that age, class level, and academic background may have an influence on artificial intelligence awareness.

Table 5 presents the distribution of participants’ total attitude and confidence scores toward artificial intelligence in the field of physiotherapy according to demographic variables. The analyses revealed no statistically significant differences among the variables.

DISCUSSION

In this study, the levels of knowledge, awareness, attitudes, and confidence regarding artificial intelligence (AI) among physiotherapy and rehabilitation students in Türkiye were examined, and the relationships between these parameters were explored. The findings indicated that the majority of students had no prior educational background related to AI. However, demographic variables such as age, year of study, and history of computer/mathematics-based courses were found to have significant effects on students’ knowledge and awareness levels. Notably, 51% of students reported being unfamiliar with AI-supported software systems for patient assessment, monitoring, or remote follow-up, and only 4% had received any AI-related education.

While students’ knowledge levels were relatively limited, their awareness and confidence levels were found to be high. Another important finding was the increase in knowledge and awareness with advancing age and academic year. These findings point to a gap in physiotherapy curricula, as the lack of AI training may reflect a structural educational shortcoming. Considering that rehabilitation outcomes are increasingly linked to technology-based systems such as long-term monitoring, patient adherence, and continuous follow-up, insufficient training may pose risks for future clinical practice. The strong relationship observed between awareness and confidence, despite limited knowledge, may be attributed to students’ perception that AI could strengthen their profession and should be integrated into education. Differences across demographic groups, particularly age and class level, suggest that exposure to clinical practice and employment concerns may enhance students’ recognition of AI’s potential benefits, including improved diagnostic accuracy, better access to healthcare, and reduced workload. The fact that most knowledge acquisition seems to occur through self-directed learning and digital resources, rather than formal education, highlights a need for structured integration of AI into physiotherapy training.

The results of this study are consistent with previous research reporting limited AI knowledge among health sciences students. Amiri et al.^[23] demonstrated that students across disciplines had low levels of AI-related knowledge, while Dashti et al.^[24] reported similar findings among dental students. Our results also align with Bozdemir-Özel and Yakut-Özdemir,^[15] who emphasized that clinicians’ uncertainty and anxiety toward AI often stem from insufficient foundational knowledge. Likewise, Topol et al.^[10] and Bisdas et al.^[25] argued that without basic AI literacy, healthcare professionals may adopt a passive role in clinical decision-making processes.

Despite limited knowledge, the relatively high levels of awareness and confidence observed in our study are consistent with reports indicating that students generally perceive AI as a supportive and augmentative tool rather than a replacement for healthcare professionals.^[17,26-28] This apparent discrepancy between knowledge and confidence has also been reported in imaging-related disciplines. For example, a quantitative review by Bhandari et al.^[28] showed that while students and clinicians in radiology often view AI optimistically as a diagnostic aid, students simultaneously express anxiety regarding future job prospects, suggesting that confidence may coexist with uncertainty when formal training is lacking.

Recent large-scale studies among medical and radiography students further support these findings. Allam et al.^[29] and Jebreen et al.^[30] reported substantial gaps in AI knowledge and training among medical students, alongside strong support for the integration of AI into undergraduate curricula. Interestingly, students with higher AI knowledge in some studies were more likely to express concerns regarding workforce displacement, particularly in radiology-focused specialties.^[29] Similar patterns have been observed among radiography students, who generally report positive attitudes toward AI’s assistive role but remain uncertain about its implications for job security and professional identity.^[31-33]

These contrasting findings suggest that attitudes toward AI vary not only by discipline but also by the extent to which AI has been formally integrated into education and clinical training. In fields such as radiology, where AI applications are already visible in practice, students may develop heightened awareness of both opportunities and threats. In contrast, in physiotherapy education, where structured AI exposure remains limited, students may demonstrate high confidence and positive expectations without a corresponding depth of knowledge. This interpretation is supported by our findings and underscores the importance of curriculum-based AI literacy in aligning confidence, awareness, and actual competence in future physiotherapy practice.

The increase in knowledge and awareness observed in this study during the academic year is also consistent with Pucchio et al.^[16] who found that exposure to AI and awareness rise with training duration, though knowledge gains remain limited.

CONCLUSION

This study highlights physiotherapy students’ levels of knowledge, attitudes, and confidence toward artificial intelligence, underscoring its growing importance in professional education. AI may contribute not only to clinical decision support and patient monitoring but also to the development of professional identity and technological adaptation among students. Systematic integration of AI into curricula by educational institutions and policymakers could advance professional standards and support more effective healthcare delivery.

Future research should go beyond measuring knowledge and attitudes to examine the direct impact of AI-based education on students’ clinical decision-making, professional competence, and patient outcomes, thereby providing a more comprehensive understanding of AI’s role in physiotherapy education.