Translation, cultural adaptation, and validation of the perceived Artificial Intelligence Literacy Questionnaire-6 (PAILQ-6) in Arabic-speaking pharmacists: A cross-sectional methodological study Authors: Ammar Abdulrahman Jairoun, Moyad Shahwan, Abeer M. Al-Ghananeem, Faris El-Dahiyat, Ammar Ali Saleh Jaber, Nasser M. Alorfi, Mansour M. Alourfi, Ammar Aldabbagh, Nasser M. Aldekhail, Faris A. Alsolami Published: 2026 ABSTRACT BackgroundThe growing use of artificial intelligence (AI) in healthcare and pharmacy has increased the need to determine whether pharmacists are prepared to understand, assess, and use AI-supported technologies responsibly. Arabic-speaking pharmacy settings currently lack a validated Arabic instrument specifically designed to assess pharmacists' perceived AI literacy. This study aimed to translate, culturally adapt, and validate the Arabic version of the Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) and to examine demographic and professional factors associated with perceived AI literacy among pharmacists.MethodsA cross-sectional methodological validation study was conducted among 500 pharmacists practicing in Arabic-speaking countries. Data were collected using an anonymous online self-administered questionnaire. Translation and cultural adaptation followed the seven-step framework described by Sousa and Rojjanasrirat. Psychometric evaluation included confirmatory factor analysis using maximum likelihood (ML) and robust maximum likelihood estimation, internal consistency assessment, convergent validity testing, and measurement invariance analysis across gender. Multivariable linear regression was used to examine independent factors associated with perceived AI literacy.ResultsThe Arabic PAILQ-6 retained the original one-factor structure and demonstrated strong psychometric performance. The robust model showed excellent fit (Satorra–Bentler χ2 = 11.95, df = 9, p = 0.216; CFI = 0.994; NNFI = 0.990; RMSEA = 0.026). Standardized factor loadings ranged from 0.635 to 0.782. Internal consistency was high, with Cronbach's α = 0.862 and composite reliability = 0.863, and convergent validity was supported. Configural, metric, and scalar invariance were established across gender, indicating equivalent measurement among male and female pharmacists. In the adjusted regression model, previous AI experience (β = 0.131, p = 0.004) and industrial pharmacy practice (β = 0.157, p = 0.007) were independently associated with higher perceived AI literacy. Gender, age, years of professional experience, and other pharmacy practice settings were not significant independent predictors.ConclusionsThe Arabic PAILQ-6 demonstrated satisfactory validity, reliability, and measurement invariance across gender for assessing perceived AI literacy among pharmacists in Arabic-speaking countries. The findings should be interpreted in light of the cross-sectional design, convenience sampling, and reliance on self-reported perceived AI literacy, which limit causal interpretation and may affect generalizability. The instrument may be useful for identifying training needs, supporting pharmacy curriculum development, evaluating educational initiatives, and assessing workforce preparedness for the responsible use of AI. Previous exposure to AI technologies was associated with higher perceived AI literacy. Introduction Artificial intelligence (AI) is now a major force in healthcare transformation. Its influence extends across clinical services, health system management, patient monitoring, and decision support. In pharmacy, AI has begun to reshape both technical and patient-facing aspects of professional practice. AI-enabled tools can support medication selection, flag potential adverse drug reactions, detect drug–drug interactions, automate prescription verification, and improve inventory control. Through these applications, AI has the potential to improve medication safety, reduce preventable errors, and support more efficient pharmaceutical care.1, 2, 3, 4 Clinical decision support systems that incorporate AI can also process large and complex clinical datasets in real time, helping pharmacists evaluate treatment options, reduce medication-related risks, and contribute to improved therapeutic outcomes.2, 4, 5 AI has also strengthened the movement toward precision medicine, particularly when combined with pharmacogenomics, because it can support individualized therapeutic decisions based on patients' genetic profiles and clinical characteristics.1, 2, 5 In addition, automated dispensing systems and pharmacy robotics have improved workflow efficiency, reduced dispensing-related errors, and created opportunities for pharmacists to spend more time on direct patient care.5, 6 AI-supported patient engagement tools have further expanded the role of pharmacists by enabling personalized medication counseling, adherence support, health behavior guidance, and ongoing monitoring of patients outside traditional care settings.3, 4, 7 AI is also contributing to remote patient monitoring, telepharmacy, wearable health technologies, and intelligent virtual assistants, with potential benefits for healthcare access and continuity of care. In the pharmaceutical sciences, AI is being applied to molecular modeling, drug repurposing, drug delivery optimization, polypharmacology, and other areas of drug discovery and development, with potential benefits for innovation and drug development timelines.8 The influence of AI also extends to pharmacy education and professional development. AI-enabled learning systems that use machine learning and natural language processing can provide adaptive learning pathways that respond to individual learner needs. These systems can support pharmacy students and practicing pharmacists in developing knowledge and skills across areas such as pharmacovigilance, clinical decision-making, precision therapeutics, and regulatory compliance.9 As AI becomes more integrated into healthcare, pharmacy education will need to prepare students and practicing professionals with the knowledge and skills required to engage appropriately with AI-supported technologies. Despite these opportunities, several barriers continue to limit the effective integration of AI into pharmacy practice and education. Pharmacy education has not yet fully adapted to the demands of AI-enabled healthcare. Challenges include limited faculty preparedness, insufficient AI content in pharmacy curricula, concerns about privacy and data protection, and limited institutional capacity to support AI-based learning.9 Healthcare organizations must also address ethical and professional concerns related to algorithmic transparency, explainability, accountability, privacy, and data security when implementing AI systems.10 In addition, the absence of comprehensive regulatory frameworks and consistent governance models creates uncertainty about how AI should be used safely, responsibly, and ethically in clinical environments.11 These challenges indicate that successful AI integration depends not only on access to technology but also on adequate education, organizational preparedness, appropriate governance, and pharmacists' ability to critically evaluate and use AI-generated information. The growing presence of AI in healthcare means that pharmacists need competencies that go beyond conventional pharmaceutical knowledge. AI literacy has become an important professional capability because it encompasses the knowledge, skills, attitudes, and ethical awareness required to use AI technologies safely and responsibly. Pharmacists need to understand core AI concepts, including machine learning, predictive analytics, natural language processing, and generative AI. They also need to recognize the limitations and risks of AI systems, including algorithmic bias, automation bias, confounding, hallucinations, and uncertainty in model outputs.12 The safe use of AI in pharmacy also requires the ability to interpret AI-generated recommendations critically and integrate them into evidence-based practice without replacing professional judgment or compromising patient safety.13, 14 Ethical competence is equally important because pharmacists must be able to consider privacy, cybersecurity, fairness, accountability, transparency, professional responsibility, and the effects of AI on patient trust and professional roles.14, 15 Developing AI literacy requires both theoretical understanding and practical exposure. Evidence suggests that many pharmacy students and healthcare professionals still have limited foundational knowledge of AI despite increasing awareness of its relevance to healthcare.16, 17 This indicates a need for structured education covering machine learning, natural language processing, generative AI, and other emerging digital technologies. Pharmacists also need to understand the specific ways in which AI is applied within pharmacy, including medication management, pharmacovigilance, drug discovery, telepharmacy, automated dispensing, and pharmaceutical distribution systems.18, 19 Technical knowledge alone, however, is not sufficient. Pharmacists must also develop critical thinking skills that allow them to evaluate AI outputs, confirm recommendations using evidence-based resources, and communicate AI-supported decisions to patients and members of multidisciplinary healthcare teams.20 Although many pharmacists recognize that AI may improve healthcare delivery, concerns persist regarding the accuracy of AI tools, ethical risks, and the need to preserve human-centered patient care.21, 22 At the same time, studies suggest that pharmacists and pharmacy students are generally willing to learn about AI and support the inclusion of AI education in undergraduate programs and continuing professional development.23, 24 The persistence of barriers such as limited AI curricula, inadequate faculty expertise, financial constraints, and unresolved ethical issues shows that more comprehensive educational initiatives are needed. These initiatives should integrate technical, clinical, and ethical dimensions of AI into pharmacy education and professional training.25 As AI becomes more relevant to pharmacy, measuring pharmacists' AI literacy has become increasingly important. Valid and culturally appropriate assessment tools are needed to identify educational gaps, evaluate training programs, monitor workforce readiness, guide curriculum development, and inform policy. A number of AI literacy instruments have been developed for different educational and professional groups. The Artificial Intelligence Literacy Questionnaire (AILQ), for example, assesses affective, behavioral, cognitive, and ethical dimensions of AI literacy among secondary school students and has shown acceptable psychometric performance in different cultural settings.26, 27 The Generative AI Literacy Assessment Test (GLAT) was developed to assess generative AI competencies among higher education students and has demonstrated good reliability and external validity.28 The Artificial Intelligence Competence and Literacy Test (CAIA) evaluates conceptual understanding, ethical awareness, and applied AI competencies among university students using a multidimensional structure.29 In healthcare education, the AI in Healthcare Scale (AIHS) has been validated among nursing students in Arab countries,30 while the Arabic Artificial Intelligence Literacy Scale (AILS) measures AI literacy among Arabic-speaking university students.31 Although these instruments have advanced AI literacy research, most were not designed for pharmacists. Many target school students, university students, or healthcare groups outside the pharmacy profession. Existing Arabic instruments also tend to assess general AI literacy rather than perceived AI literacy in practicing pharmacists. This distinction is relevant because pharmacists work across medication management, pharmaceutical care, clinical decision-making, patient counseling, pharmacovigilance, and digital health services. Pharmacy educators, healthcare organizations, and regulatory authorities in Arabic-speaking countries therefore need a brief, culturally appropriate, and psychometrically evaluated Arabic questionnaire that can assess perceived AI literacy among pharmacists. Such an instrument could help identify competency gaps, evaluate AI-related training, assess workforce preparedness, and inform responsible AI implementation in pharmacy settings.32, 33 The Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) is a concise six-item instrument developed to measure perceived AI literacy rather than objective technical knowledge or performance.34 Its short format reduces respondent burden and is well suited to survey-based assessment among practicing professionals. Its focus on individuals' perceived understanding of and ability to engage with AI technologies also corresponds closely with the construct examined in the present study. For these reasons, the PAILQ-6 was selected for translation, cultural adaptation, and psychometric evaluation in an Arabic-speaking pharmacy population. Accordingly, the present study aimed to translate, culturally adapt, and psychometrically validate the Arabic version of the PAILQ-6 among pharmacists practicing in Arabic-speaking countries. Methods and materials Study design and setting This study adopted a cross-sectional methodological validation design. Data were collected between March 2026 and May 2026 through an anonymous online self-administered questionnaire distributed using electronic survey platforms. The study was conducted in accordance with internationally recognized guidance for the translation and cross-cultural adaptation of psychometric instruments. It also followed established methodological standards for questionnaire validation, reliability assessment, and construct validation. Study participants The study population consisted of licensed pharmacists currently practicing in Arabic-speaking countries. Participants were eligible if they were able to read and understand Arabic and had completed at least 3 months of professional practice after registration or licensure. This criterion was used to ensure that participants had experienced routine pharmacy work, clinical duties, patient care responsibilities, and exposure to emerging digital technologies, including AI-enabled healthcare applications. Pharmacists still completing probation, pharmacy students, pharmacy interns, pharmacy technicians, and other healthcare professionals were excluded. These groups were not included because they may not yet have sufficient independent professional experience to judge their perceived AI literacy within routine pharmacy practice. Participation was voluntary, and all participants provided electronic informed consent before taking part in the study. Sample size calculation Sample size adequacy was considered in relation to the psychometric purpose of the study and the need to obtain stable parameter estimates for confirmatory factor analysis (CFA). Established methodological recommendations suggest approximately 5–10 participants per questionnaire item, while a total sample of at least 200 participants is generally considered adequate for factor analysis and structural equation modeling.35, 36 For the six-item PAILQ-6, the item-based criterion would correspond to a minimum of 30–60 participants. However, because larger samples provide more stable CFA estimates and greater precision in psychometric evaluation, the study aimed to recruit a substantially larger sample. Sample adequacy was also considered in accordance with the methodological principles of the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN), which provide a framework for evaluating the methodological quality of studies examining measurement properties.37 Translation and cultural adaptation Translation and cultural adaptation of the original English-language PAILQ-6 followed the seven-step framework described by Sousa and Rojjanasrirat.38 This approach was used to maintain linguistic accuracy, conceptual equivalence, and cultural appropriateness throughout the adaptation process. Before translation and cultural adaptation, the corresponding author contacted Professor Simone Grassini, the developer of the original PAILQ-6, to request permission to use, translate, culturally adapt, and validate the instrument among Arabic-speaking pharmacists. Written permission was granted by the instrument developer, who confirmed that the authors were free to use the PAILQ-6 for the present study. The original PAILQ-6 publication also identified cross-cultural validation and evaluation of the questionnaire in diverse populations as important directions for future research.34 The adaptation process comprised seven sequential steps: forward translation, reconciliation, backward translation, comparison with the source instrument, pilot testing, preliminary psychometric evaluation, and full psychometric testing. Two certified bilingual translators independently translated the questionnaire from English into Arabic. The research team compared and reconciled the two translations to produce a single Arabic version. Two additional bilingual translators who were blinded to the original questionnaire independently translated the reconciled Arabic version back into English. The back-translated versions were compared with the original instrument to assess semantic and conceptual equivalence. Any discrepancies were discussed by the translators and research team and resolved by consensus. Cultural adaptation focused on maintaining the conceptual meaning of the original items while ensuring that their wording was linguistically and contextually appropriate for Arabic-speaking populations. A pilot study was conducted among 25 pharmacists with characteristics similar to those of the target population to assess item clarity and comprehensibility. Participants rated each item using a dichotomous response of “clear” or “unclear.” All items were considered clear and comprehensible, and no further modifications were required. Research instrument The final Arabic questionnaire was organised in two sections. Sociodemographic information — age, gender, Pharmacist type, years of experience and previous experience with AI — was gathered in the first section. The Arabic version of the PAILQ-6 occupied the second section. The PAILQ-6 is a concise six-item instrument developed to measure perceived artificial intelligence literacy.34 Its design favours simple presentation and online administration through digital devices and web-based platforms. Item responses were recorded on a seven-point Likert scale anchored at 1 (“strongly disagree”) and 7 (“strongly agree”), with higher totals reflecting greater perceived AI literacy. Seven-point Likert formats are widely adopted because they yield greater response variability and improved measurement precision.39 Questionnaire scoring Scoring procedures replicated those specified for the original instrument. Item scores were summed to yield a total score, with higher values reflecting stronger perceived understanding of AI concepts, greater confidence in using AI technologies, and broader awareness of AI-related applications. Sampling technique and data collection Participants were recruited using a non-probability convenience sampling strategy. The aim was to reach pharmacists across multiple Arabic-speaking countries and practice settings. The survey link was distributed through professional pharmacy associations, pharmacist social media communities, academic pharmacy networks, professional mailing lists, and widely used messaging applications, including WhatsApp and Telegram. Using several recruitment channels helped increase geographical reach and improve the diversity of the sample. Data were collected through an anonymous online self-administered questionnaire hosted on a secure electronic survey platform. Before accessing the questionnaire, prospective participants were shown an electronic information sheet. This sheet described the study aims, eligibility requirements, voluntary nature of participation, confidentiality procedures, and the right to withdraw at any point without penalty. Only those who provided electronic informed consent were able to continue to the questionnaire. Several steps were taken to improve data quality. Where technically possible, the survey platform was configured to restrict responses to one submission per participant. After data collection, all responses were reviewed for completeness, logical consistency, and duplication. Incomplete questionnaires, duplicate entries, responses showing obvious response patterns, and submissions with implausible completion times were removed before analysis. Strategies to minimize bias and enhance validity Several procedures were incorporated to reduce potential bias and strengthen the methodological quality of the study. Potential selection bias was mitigated by recruiting pharmacists through multiple professional and academic channels and by seeking participation from different pharmacy practice settings and Arabic-speaking countries. Although this approach increased sample heterogeneity, the use of non-probability convenience sampling did not eliminate the possibility of selection bias. Information bias was reduced through the use of a standardized anonymous online questionnaire, which minimized interviewer influence and allowed participants to complete the survey privately. Internal validity was supported through systematic translation and cultural adaptation, pilot testing of the Arabic questionnaire, standardized data collection procedures, and established psychometric methods. Data integrity was further supported by screening responses for duplication, incompleteness, inconsistency, and implausible response patterns before analysis. Recruitment across varied demographic and professional groups was intended to increase sample diversity and the applicability of the findings across different pharmacy practice contexts. Statistical analysis Data were analyzed using IBM SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA), EQS version 6.4 (Multivariate Software Inc., Encino, CA, USA), and R (lavaan package) for measurement invariance analyses. Descriptive statistics were used to summarize participants' demographic and professional characteristics, with categorical variables presented as frequencies and percentages. The factorial validity of the Arabic version of the Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) was evaluated using confirmatory factor analysis (CFA). Model fit was assessed using the chi-square statistic (χ2), Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), also referred to as the Non-Normed Fit Index (NNFI), Standardized Root Mean Square Residual (SRMR), and Root Mean Square Error of Approximation (RMSEA) with its 90% confidence interval. The Incremental Fit Index (IFI) and Normed Fit Index (NFI) were also examined as supplementary fit indices. Model fit was considered acceptable when CFI and TLI/NNFI were ≥ 0.90 (preferably ≥0.95), SRMR was ≤0.08, and RMSEA was ≤0.08 (preferably ≤0.06), consistent with established recommendations.40 Because multivariate normality could not be assumed, robust maximum likelihood (Robust ML) estimation with the Satorra–Bentler scaled chi-square statistic was also performed, and the robust model was considered the primary basis for model evaluation. Internal consistency reliability was assessed using Cronbach's alpha (α) and composite reliability (ρ), with values ≥0.70 considered satisfactory.35 Measurement invariance across gender was evaluated sequentially at the configural, metric, and scalar levels. Invariance was assessed using changes in model fit between increasingly constrained models. It was considered supported when |ΔCFI| was ≤0.010 and ΔRMSEA was ≤0.015, together with ΔSRMR ≤0.030 for metric invariance and ≤ 0.010 for scalar invariance.41, 42 Because RMSEA may overestimate model misfit in models with relatively few degrees of freedom, greater emphasis was placed on changes in CFI and SRMR when interpreting measurement invariance.43 To identify factors associated with perceived AI literacy, multivariable linear regression analysis was performed with the total PAILQ-6 score as the dependent variable. Independent variables included previous AI experience, gender, pharmacist practice setting, age group, and years of professional experience. Regression coefficients (B), standardized coefficients (β), 95% confidence intervals (CIs), and p-values were reported. Statistical significance was defined as a two-sided p < 0.05. Results Demographic characteristics A total of 500 pharmacists participated in the study. Most participants were male (62.8%, n = 314), while 37.2% (n = 186) were female. Community pharmacists constituted the largest professional group (36.4%, n = 182), followed by hospital pharmacists (27.8%, n = 139), industrial pharmacists (18.2%, n = 91), and academic/research pharmacists (17.6%, n = 88). The largest age group was 30–39 years (42.0%, n = 210), followed by 25–29 years (30.2%, n = 151), 40–49 years (19.4%, n = 97), and 50 years or older (8.4%, n = 42). More than half of the participants had 1–5 years of professional experience (55.0%, n = 275), whereas 40.2% (n = 201) had 6–10 years of experience and 4.8% (n = 24) had 11 years or more. Regarding artificial intelligence, 37.0% (n = 185) reported having previous AI experience, while the majority (63.0%, n = 315) reported no prior AI experience (Table 1).Table 1Demographic and professional characteristics of the study participants (N = 500).CharacteristicCategoryn%GenderMale31462.8Female18637.2Pharmacist typeCommunity pharmacists18236.4Hospital pharmacists13927.8Industrial pharmacists9118.2Academic/Research pharmacists8817.6Age group (years)25–2915130.230–392104240–499719.4≥50428.4Years of professional experience1–5 years275556–10 years20140.2≥11 years244.8Previous experience with AINo31563Yes18537 Table 1Demographic and professional characteristics of the study participants (N = 500).CharacteristicCategoryn%GenderMale31462.8Female18637.2Pharmacist typeCommunity pharmacists18236.4Hospital pharmacists13927.8Industrial pharmacists9118.2Academic/Research pharmacists8817.6Age group (years)25–2915130.230–392104240–499719.4≥50428.4Years of professional experience1–5 years275556–10 years20140.2≥11 years244.8Previous experience with AINo31563Yes18537 Demographic and professional characteristics of the study participants (N = 500). Confirmatory factor analysis Confirmatory factor analysis was conducted to evaluate the hypothesised unidimensional structure of the Arabic version of the Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) among 500 pharmacists practicing in Arabic-speaking countries. The model specified the six questionnaire items as indicators of a single latent factor and was estimated using both conventional maximum likelihood (ML) and robust maximum likelihood procedures. Assessment of distributional assumptions indicated substantial multivariate non-normality, with a Mardia multivariate kurtosis coefficient of 54.175 and a normalized estimate of 61.819. Therefore, the Satorra–Bentler robust estimates were considered the primary basis for evaluating model fit, while conventional ML estimates were also reported for comparison. Under conventional ML estimation, the one-factor model produced a statistically significant chi-square value, χ2 (9) = 28.382, p < 0.001. Nevertheless, the remaining fit indices indicated an acceptable-to-good model fit: comparative fit index (CFI) = 0.984, incremental fit index (IFI) = 0.984, non-normed fit index (NNFI) = 0.973, normed fit index (NFI) = 0.977, standardized root mean square residual (SRMR) = 0.027, and root mean square error of approximation (RMSEA) = 0.066, 90% CI [0.039, 0.094]. (Table 2; Fig. 1).Table 2CFA model fit under ML and robust ML.Fit statisticML estimateRobust ML estimateChi-square28.38211.952aDegrees of freedom99p-value<0.0010.216NFI0.9770.976NNFI/TLI0.9730.99CFI0.9840.994IFI0.9840.994MFI0.9810.997GFI0.981–AGFI0.956–SRMR0.027–RMSEA0.0660.026RMSEA 90% CI0.039–0.0940.000–0.060Notes: ML = maximum likelihood; NFI = normed fit index; NNFI = non-normed fit index; TLI = Tucker–Lewis index; CFI = comparative fit index; IFI = incremental fit index; MFI = McDonald's fit index; GFI = goodness-of-fit index; AGFI = adjusted goodness-of-fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.aSatorra–Bentler scaled chi-square.Fig. 1Standardized confirmatory factor analysis model of the Arabic PAILQ-6.Fig. 1 Table 2CFA model fit under ML and robust ML.Fit statisticML estimateRobust ML estimateChi-square28.38211.952aDegrees of freedom99p-value<0.0010.216NFI0.9770.976NNFI/TLI0.9730.99CFI0.9840.994IFI0.9840.994MFI0.9810.997GFI0.981–AGFI0.956–SRMR0.027–RMSEA0.0660.026RMSEA 90% CI0.039–0.0940.000–0.060Notes: ML = maximum likelihood; NFI = normed fit index; NNFI = non-normed fit index; TLI = Tucker–Lewis index; CFI = comparative fit index; IFI = incremental fit index; MFI = McDonald's fit index; GFI = goodness-of-fit index; AGFI = adjusted goodness-of-fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation.aSatorra–Bentler scaled chi-square. CFA model fit under ML and robust ML. Notes: ML = maximum likelihood; NFI = normed fit index; NNFI = non-normed fit index; TLI = Tucker–Lewis index; CFI = comparative fit index; IFI = incremental fit index; MFI = McDonald's fit index; GFI = goodness-of-fit index; AGFI = adjusted goodness-of-fit index; SRMR = standardized root mean square residual; RMSEA = root mean square error of approximation. Satorra–Bentler scaled chi-square. Standardized confirmatory factor analysis model of the Arabic PAILQ-6. The robust ML results demonstrated a better-fitting model. The Satorra–Bentler scaled chi-square was non-significant, SBχ2 (9) = 11.952, p = 0.216, with a scaling correction factor of 0.421. Robust fit indices indicated excellent model fit: CFI = 0.994, IFI = 0.994, NNFI = 0.990, NFI = 0.976, and RMSEA = 0.026, 90% CI [0.000, 0.060]. Collectively, these results supported the proposed unidimensional structure of the Arabic PAILQ-6 among pharmacists. All six items loaded positively and substantially on the latent perceived AI literacy construct. Standardized factor loadings ranged from 0.635 to 0.782. Item 5 demonstrated the strongest loading (λ = 0.782), followed by Item 6 (λ = 0.751), Item 2 (λ = 0.748), Item 4 (λ = 0.723), Item 1 (λ = 0.652), and Item 3 (λ = 0.635). All freely estimated factor loadings were statistically significant under both conventional and robust estimation (all p < 0.001). The proportion of variance explained by the latent factor ranged from 40.4% for Item 3 to 61.1% for Item 5 (Table 3).Table 3Standardized factor loadings and item reliability.ItemsStandardized loading, λML SEᵃML zRobust SEaRobust zR2I understand the basic concepts of artificial intelligence0.652––––0.424I believe I can contribute to AI projects0.7480.08713.8670.1438.4920.56I can judge the pros and cons of AI.0.6350.08712.1610.1517.0340.404I keep up with the latest AI trends.0.7230.08513.5050.1219.5180.523I'm comfortable talking about AI with others0.7820.08514.3210.1349.0690.611“I can think of new ways to use existing AI tools.0.7510.08813.9090.1448.5530.564Notes: λ = standardized factor loading; SE = standard error; R2 = proportion of item variance explained by the latent construct. All freely estimated loadings were significant at p < 0.001.aStandard errors and z-statistics relate to the corresponding unstandardized loading estimates. Item 1 was fixed at 1.00 to identify the model and therefore has no standard error or z-statistic. Table 3Standardized factor loadings and item reliability.ItemsStandardized loading, λML SEᵃML zRobust SEaRobust zR2I understand the basic concepts of artificial intelligence0.652––––0.424I believe I can contribute to AI projects0.7480.08713.8670.1438.4920.56I can judge the pros and cons of AI.0.6350.08712.1610.1517.0340.404I keep up with the latest AI trends.0.7230.08513.5050.1219.5180.523I'm comfortable talking about AI with others0.7820.08514.3210.1349.0690.611“I can think of new ways to use existing AI tools.0.7510.08813.9090.1448.5530.564Notes: λ = standardized factor loading; SE = standard error; R2 = proportion of item variance explained by the latent construct. All freely estimated loadings were significant at p < 0.001.aStandard errors and z-statistics relate to the corresponding unstandardized loading estimates. Item 1 was fixed at 1.00 to identify the model and therefore has no standard error or z-statistic. Standardized factor loadings and item reliability. Notes: λ = standardized factor loading; SE = standard error; R2 = proportion of item variance explained by the latent construct. All freely estimated loadings were significant at p < 0.001. Standard errors and z-statistics relate to the corresponding unstandardized loading estimates. Item 1 was fixed at 1.00 to identify the model and therefore has no standard error or z-statistic. Reliability and convergent validity The Arabic version of the Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) demonstrated satisfactory reliability and convergent validity (Table 4). Internal consistency reliability was supported by a Cronbach's alpha coefficient of 0.862, exceeding the recommended threshold of 0.70. Similarly, the model-based reliability coefficient (ρ = 0.863) indicated good construct reliability, while the maximal weighted internal consistency reliability (H = 0.869) further confirmed the stability and consistency of the latent construct. Standardized factor loadings ranged from 0.635 to 0.782, with all items exhibiting statistically significant loadings on the underlying perceived AI literacy factor (all p < 0.001). The Average Variance Extracted (AVE = 0.514) exceeded the recommended criterion of 0.50, indicating that the latent construct explained more than half of the variance in its observed indicators.Table 4Reliability and convergent validity.MeasureValueRecommended thresholdCronbach's alpha0.862≥0.70Reliability coefficient rho0.863≥0.70Maximal weighted internal consistency reliability0.869≥0.70Composite reliability0.863≥0.70Average variance extracted0.514≥0.50Abbreviations: AVE = Average Variance Extracted; α = Cronbach's alpha; ρ = Reliability coefficient (McDonald's omega equivalent reported by EQS); H = Maximal weighted internal consistency reliability. Note: Values of α, ρ, and H ≥ 0.70 indicate satisfactory internal consistency reliability, whereas AVE ≥ 0.50 supports convergent validity. Table 4Reliability and convergent validity.MeasureValueRecommended thresholdCronbach's alpha0.862≥0.70Reliability coefficient rho0.863≥0.70Maximal weighted internal consistency reliability0.869≥0.70Composite reliability0.863≥0.70Average variance extracted0.514≥0.50Abbreviations: AVE = Average Variance Extracted; α = Cronbach's alpha; ρ = Reliability coefficient (McDonald's omega equivalent reported by EQS); H = Maximal weighted internal consistency reliability. Note: Values of α, ρ, and H ≥ 0.70 indicate satisfactory internal consistency reliability, whereas AVE ≥ 0.50 supports convergent validity. Reliability and convergent validity. Abbreviations: AVE = Average Variance Extracted; α = Cronbach's alpha; ρ = Reliability coefficient (McDonald's omega equivalent reported by EQS); H = Maximal weighted internal consistency reliability. Note: Values of α, ρ, and H ≥ 0.70 indicate satisfactory internal consistency reliability, whereas AVE ≥ 0.50 supports convergent validity. Measurement invariance across gender The measurement invariance results across gender are presented in Table 5. The Arabic PAILQ-6 demonstrated configural, metric, and scalar invariance between male and female pharmacists. Changes in model fit indices between successive models remained within the recommended thresholds (|ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.030 for metric invariance and ≤ 0.010 for scalar invariance), indicating that the questionnaire measured perceived artificial intelligence literacy equivalently across genders. These findings support meaningful comparisons of both structural relationships and latent mean scores between male and female pharmacists. It is important to note that RMSEA has recognized limitations when applied to models with relatively few degrees of freedom, as it may overestimate model misfit even when the underlying model fits the data adequately.43 Accordingly, greater emphasis was placed on changes in CFI and SRMR, which are less sensitive to model size and complexity, and these indices consistently supported gender measurement invariance.Table 5Measurement invariance across gender.Modelχ2 (df)CFITLIRMSEA (90% CI)SRMRΔCFIΔRMSEAΔSRMRConfigural56.158 (18)0.9690.9480.092 (0.065–0.120)0.032–––Metric65.214 (23)0.9660.9550.086 (0.062–0.111)0.049−0.003−0.0060.017Scalar67.653 (28)0.9680.9650.075 (0.053–0.098)0.050.002−0.0110.001Abbreviations: CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual. Note: Changes were calculated relative to the immediately preceding model. Metric invariance was considered supported when |ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.030; scalar invariance was considered supported when |ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.010. Because RMSEA may overestimate model misfit in models with relatively few degrees of freedom, greater emphasis was placed on changes in CFI and SRMR when evaluating invariance (Kenny et al., 2015). Table 5Measurement invariance across gender.Modelχ2 (df)CFITLIRMSEA (90% CI)SRMRΔCFIΔRMSEAΔSRMRConfigural56.158 (18)0.9690.9480.092 (0.065–0.120)0.032–––Metric65.214 (23)0.9660.9550.086 (0.062–0.111)0.049−0.003−0.0060.017Scalar67.653 (28)0.9680.9650.075 (0.053–0.098)0.050.002−0.0110.001Abbreviations: CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual. Note: Changes were calculated relative to the immediately preceding model. Metric invariance was considered supported when |ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.030; scalar invariance was considered supported when |ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.010. Because RMSEA may overestimate model misfit in models with relatively few degrees of freedom, greater emphasis was placed on changes in CFI and SRMR when evaluating invariance (Kenny et al., 2015). Measurement invariance across gender. Abbreviations: CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual. Note: Changes were calculated relative to the immediately preceding model. Metric invariance was considered supported when |ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.030; scalar invariance was considered supported when |ΔCFI| ≤ 0.010, ΔRMSEA ≤0.015, and ΔSRMR ≤0.010. Because RMSEA may overestimate model misfit in models with relatively few degrees of freedom, greater emphasis was placed on changes in CFI and SRMR when evaluating invariance (Kenny et al., 2015). Factors associated with pharmacists' perceived artificial intelligence literacy A multivariable linear regression analysis was conducted to identify factors independently associated with pharmacists perceived artificial intelligence literacy. The overall model was statistically significant, F (10, 489) = 3.043, p = 0.001, and explained 5.9% of the variance in PAILQ-6 scores (R2 = 0.059; adjusted R2 = 0.039). After adjustment for gender, pharmacist type, age, and professional experience, pharmacists with previous AI experience had PAILQ-6 scores that were, on average, 1.62 points higher than those without previous AI experience (B = 1.621, β = 0.131, 95% CI: 0.515–2.726, p = 0.004). Industrial pharmacists also reported significantly higher perceived AI literacy than community pharmacists, with an adjusted mean difference of 2.43 points (B = 2.425, β = 0.157, 95% CI: 0.670–4.180, p = 0.007). Academic/research pharmacists showed a borderline higher score than community pharmacists (B = 1.664, p = 0.052), but this association did not reach statistical significance. Gender, hospital pharmacy practice, age group, and years of professional experience were not independently associated with PAILQ-6 scores (all p > 0.05) (Table 6).Table 6Multivariable linear regression of factors associated with pharmacists' perceived AI literacy.PredictorBβtp-value95% CI for BPrevious AI experience1.6210.1312.880.0040.515 to 2.726Gender (Ref. Female)Male pharmacist vs0.8080.0651.460.145−0.279 to 1.895Pharmacist (Ref. Community pharmacist)Industrial pharmacist2.4250.1572.7160.0070.670 to 4.180Academic/research pharmacist1.6640.1061.9470.052−0.015 to 3.343Hospital pharmacist−0.011−0.001−0.0130.99−1.583 to 1.562Age Group (Ref. ≥50 years)Age 25–29 years1.8540.1431.6570.098−0.345 to 4.052Age 30–39 years1.6630.1381.6620.097−0.303 to 3.630Age 40–49 years1.8510.1231.6960.091−0.293 to 3.995Professional experience (Ref. ≥ 11 years)Professional experience 1–5 years0.6150.0510.4910.624−1.846 to 3.076Professional experience 6–10 years0.9210.0760.7250.469−1.573 to 3.414Abbreviations: B = unstandardized regression coefficient; β = standardized regression coefficient; CI = confidence interval. Bold values indicate statistical significance at p < 0.05. Table 6Multivariable linear regression of factors associated with pharmacists' perceived AI literacy.PredictorBβtp-value95% CI for BPrevious AI experience1.6210.1312.880.0040.515 to 2.726Gender (Ref. Female)Male pharmacist vs0.8080.0651.460.145−0.279 to 1.895Pharmacist (Ref. Community pharmacist)Industrial pharmacist2.4250.1572.7160.0070.670 to 4.180Academic/research pharmacist1.6640.1061.9470.052−0.015 to 3.343Hospital pharmacist−0.011−0.001−0.0130.99−1.583 to 1.562Age Group (Ref. ≥50 years)Age 25–29 years1.8540.1431.6570.098−0.345 to 4.052Age 30–39 years1.6630.1381.6620.097−0.303 to 3.630Age 40–49 years1.8510.1231.6960.091−0.293 to 3.995Professional experience (Ref. ≥ 11 years)Professional experience 1–5 years0.6150.0510.4910.624−1.846 to 3.076Professional experience 6–10 years0.9210.0760.7250.469−1.573 to 3.414Abbreviations: B = unstandardized regression coefficient; β = standardized regression coefficient; CI = confidence interval. Bold values indicate statistical significance at p < 0.05. Multivariable linear regression of factors associated with pharmacists' perceived AI literacy. Abbreviations: B = unstandardized regression coefficient; β = standardized regression coefficient; CI = confidence interval. Bold values indicate statistical significance at p < 0.05. Discussion This study translated, culturally adapted, and validated the Arabic version of the Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) for use among pharmacists practicing in Arabic-speaking countries. The findings provide strong evidence supporting the psychometric performance of the Arabic PAILQ-6. The instrument retained a stable one-factor structure, demonstrated satisfactory reliability and convergent validity, and met the criteria for measurement invariance across gender. The regression analysis also identified two factors independently associated with higher perceived AI literacy: previous AI experience and employment as an industrial pharmacist. These findings suggest that the Arabic PAILQ-6 can be used to assess perceived AI literacy among pharmacists and may also help identify professional characteristics associated with differences in perceived AI literacy. The confirmatory factor analysis supported the proposed unidimensional structure of the PAILQ-6. All standardized factor loadings were above recommended thresholds, indicating that each item contributed meaningfully to the underlying construct. The robust maximum likelihood model demonstrated excellent fit, with CFI = 0.994 and RMSEA = 0.026. Reliability was also strong, with Cronbach's α = 0.862 and composite reliability = 0.863. These results indicate that the Arabic version preserved the conceptual structure of the original questionnaire and performed consistently among Arabic-speaking pharmacists. The findings are also consistent with previous research describing AI literacy as a broad but coherent construct that incorporates technical understanding, critical evaluation, ethical awareness, and practical engagement with AI technologies.44, 45, 46, 47 More broadly, the results support the view that AI literacy can be assessed reliably across professional contexts when translation, cultural adaptation, and psychometric evaluation are conducted systematically. The assessment of measurement invariance across gender is an important methodological contribution of this study. Configural, metric, and scalar invariance were supported, indicating that the Arabic PAILQ-6 measures perceived AI literacy in a comparable manner among male and female pharmacists. This provides a stronger basis for comparisons between gender groups because observed differences are less likely to result from measurement bias. This issue is particularly relevant because previous studies have reported gender-related differences in AI literacy. Some studies suggest that male participants report greater technical confidence, whereas female participants may demonstrate stronger ethical awareness or affective engagement with AI.44, 45, 46, 47, 48, 49 However, such comparisons are difficult to interpret when measurement equivalence has not first been established. By confirming gender measurement invariance, the present study provides a stronger methodological basis for future comparisons of perceived AI literacy between male and female pharmacists in Arabic-speaking populations. Previous AI experience was independently associated with higher perceived AI literacy. Pharmacists who reported prior experience with AI had higher PAILQ-6 scores than those without such experience. This finding is consistent with previous studies reporting associations between exposure to AI technologies and greater AI literacy, more favorable attitudes toward AI, and higher confidence in using these technologies.44, 50, 51 Practical engagement may allow individuals to move beyond an abstract understanding of AI and become more familiar with how AI systems operate, their potential applications, and their limitations. Evidence from healthcare students and professionals similarly suggests that interaction with AI applications may contribute to greater familiarity and perceived competence, even when formal AI education is limited.47, 52, 53, 54 These findings support the potential value of practical exposure as part of AI literacy education for pharmacists. Employment as an industrial pharmacist was also independently associated with higher perceived AI literacy. This association is plausible because industrial pharmacy increasingly involves AI applications across drug discovery, pharmaceutical manufacturing, quality assurance, pharmacovigilance, formulation optimization, regulatory science, and supply chain management. Pharmacists working in industrial settings may therefore have more opportunities to encounter AI-enabled technologies than pharmacists in settings where such tools are less integrated into routine workflows. Previous studies have also described expanding AI adoption within the pharmaceutical industry and greater engagement with AI in digitally advanced professional environments.4, 55, 56, 57 The observed association suggests that workplace context may contribute to differences in perceived AI literacy. It also highlights the potential value of providing appropriate AI-related training across pharmacy sectors where opportunities for direct exposure may differ. In contrast, gender, age, years of professional experience, hospital pharmacy practice, and academic/research employment were not independently associated with perceived AI literacy after adjustment for the other variables in the model. Earlier studies have sometimes reported demographic differences in AI literacy, particularly according to age and gender.44, 45, 46, 47, 48, 49, 55, 58 Such differences, however, may be less apparent when practical exposure, educational opportunities, and access to AI technologies are considered. The present findings suggest that professional exposure to AI may be more relevant to perceived AI literacy than demographic characteristics alone. This interpretation is consistent with research emphasizing the role of education and practical engagement in developing AI-related competence.47, 52, 59 The educational implications of these findings are important. Previous studies have reported that formal AI education remains limited despite growing interest in AI among healthcare students and professionals. Many individuals obtain information about AI through self-directed learning, social media, informal online resources, or workplace exposure rather than through structured educational programs.47, 52, 53, 54, 60, 61. Such variation in learning opportunities may contribute to uneven levels of perceived competence and leave ethical, legal, clinical, and regulatory issues insufficiently addressed. The present findings support the inclusion of structured AI literacy content in undergraduate pharmacy curricula, postgraduate education, and continuing professional development. Education should address technical concepts alongside ethical decision-making, responsible AI use, data governance, regulatory considerations, and pharmacy-specific applications.47, 55, 62 Training may also need to reflect differences in professional roles because community, hospital, industrial, academic, and research pharmacists encounter AI technologies in different contexts. The validated Arabic PAILQ-6 can support these educational and workforce development activities. As an Arabic AI literacy instrument validated specifically among pharmacists, it provides a brief psychometrically supported measure that may be useful to educators, researchers, healthcare organizations, and policymakers. The instrument can be used to identify perceived training needs, evaluate AI education programs, assess workforce preparedness, and support the development of AI-related initiatives in Arabic-speaking pharmacy settings. As AI becomes increasingly integrated into pharmaceutical care and healthcare delivery, appropriate assessment of pharmacists' perceived AI literacy may help inform efforts to support safe, ethical, and effective engagement with these technologies. Several limitations should be considered when interpreting the findings. First, the cross-sectional design did not permit assessment of temporal stability or causal relationships between participant characteristics and perceived AI literacy. Although the Arabic PAILQ-6 demonstrated strong internal consistency and construct validity, test-retest reliability was not evaluated. Longitudinal studies are needed to determine whether PAILQ-6 scores remain stable over time. Second, convenience sampling through an online survey may have introduced selection bias. Pharmacists who were more digitally engaged, interested in AI, or active in professional online networks may have been more likely to participate. This limits the extent to which the findings can be generalized to all pharmacists in Arabic-speaking countries, particularly those with limited access to digital platforms or lower participation in online professional networks. Third, perceived AI literacy was assessed using self-report. Self-reported measures may be influenced by social desirability, confidence, self-perception, or limited awareness of actual competence. Perceived AI literacy should therefore not be interpreted as equivalent to objectively demonstrated AI competence. Future research could combine the PAILQ-6 with objective knowledge tests, scenario-based assessments, or performance-based measures to provide a broader assessment of pharmacists' AI literacy. Fourth, convergent validity was supported at the construct level in the present study through the CFA results and Average Variance Extracted. However, concurrent validity and external convergent or discriminant validity against other validated instruments were not examined because no established Arabic-language instrument assessing a directly comparable construct of perceived AI literacy was available for pharmacists at the time of the study. Future research should evaluate these additional forms of validity when suitable comparator instruments become available. Finally, the study established measurement invariance across gender, but equivalence across other relevant groups remains to be examined. Future studies should assess whether the Arabic PAILQ-6 functions similarly across countries, age groups, pharmacy practice sectors, levels of professional experience, and educational backgrounds. Further research should also examine predictive validity and responsiveness to AI-related educational or training interventions. Despite these limitations, the present study provides a rigorously evaluated Arabic version of the PAILQ-6 for assessing perceived AI literacy among pharmacists and offers a useful instrument for future research, educational evaluation, and workforce development in Arabic-speaking settings. Conclusion This study translated, culturally adapted, and psychometrically validated the Arabic version of the Perceived Artificial Intelligence Literacy Questionnaire (PAILQ-6) among pharmacists practicing in Arabic-speaking countries. The Arabic PAILQ-6 retained a stable one-factor structure and demonstrated excellent model fit under robust maximum likelihood estimation. It also showed high internal consistency, satisfactory convergent validity, and measurement invariance across gender. These findings indicate that the instrument can be used to compare perceived AI literacy between male and female pharmacists without evidence of gender-related measurement bias. Previous AI experience and employment as an industrial pharmacist were independently associated with higher perceived AI literacy. These findings suggest that opportunities for practical exposure to AI technologies may be relevant to pharmacists' perceived understanding, confidence, and readiness to engage with AI. Overall, the Arabic PAILQ-6 provides a reliable and valid tool for assessing perceived AI literacy in pharmacy settings. It may support researchers, educators, healthcare institutions, and policymakers in identifying educational needs, evaluating AI training initiatives, assessing workforce preparedness, and informing the responsible integration of AI into pharmacy practice across Arabic-speaking countries. Future studies should evaluate the Arabic PAILQ-6 across different Arabic-speaking countries and pharmacy practice settings and examine test-retest reliability, additional forms of validity, and responsiveness to AI-focused educational interventions. The instrument may also be used to identify areas of unmet learning need and to guide the development and evaluation of targeted AI literacy programs for pharmacists. Ethics and consent to participate declarations The study received ethical approval from the Institutional Ethical Review Committee of Ajman University (P-H-S-2025-2-14). This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. No identifiable personal information was collected, and all data was handled confidentially and used solely for research purposes. Consent to publish declaration Not applicable. CRediT authorship contribution statement Ammar Abdulrahman Jairoun: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Formal analysis, Data curation, Conceptualization. Moyad Shahwan: Writing – original draft, Supervision, Resources, Project administration, Formal analysis, Conceptualization. Abeer M. Al-Ghananeem: Writing – review & editing, Validation, Supervision, Project administration, Investigation, Funding acquisition. Faris El-Dahiyat: Writing – review & editing, Supervision, Software, Methodology, Formal analysis. Ammar Ali Saleh Jaber: Writing – original draft, Visualization, Resources, Data curation. Nasser M. Alorfi: Writing – review & editing, Validation, Funding acquisition, Data curation. Mansour M. Alourfi: Writing – review & editing, Resources, Methodology. Ammar Aldabbagh: Writing – original draft, Software, Resources. Nasser M. Aldekhail: Writing – review & editing, Resources, Project administration. Faris A. Alsolami: Writing – review & editing, Software, Funding acquisition, Data curation. Declaration of generative AI and AI-assisted technologies in the writing process Not applicable. Funding This research work was funded by Umm Al-Qura University, Saudi Arabia under grant number: 26UQU4420038GSSR06. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.