Assessment of the knowledge, attitudes, and practices of pharmacy students toward self-medication using artificial intelligence: a cross-sectional study

Abstract

Background

The objectives of our study were to assess self-reported knowledge, attitudes and practices (KAP) among pharmacy students regarding self-medication (SM) using artificial intelligence (AI) for personal use, to determine the relationships amongst KAP parameters, and to examine the association between students’ demographics with their KAP.

Methods

A cross-sectional study was conducted. Four hundred and four pharmacy students completed the Arabic online survey. The survey consisted of four sections demographics, KAP. Mann-Whitney U tests, Kruskal-Wallis H test and Spearman correlation were used to analyze data.

Results

The majority of pharmacy students were female (77.5%), enrolled at public universities (70.8%) as bachelor’s degree (53.7%). Students showed moderate self-reported knowledge (36.64 ± 5.39, range: 9–45) and attitudes (46.02 ± 7.94, range: 12–60), while mean practice scores were low to moderate. Self-reported Knowledge and practices scores were significantly associated with educational level (p = 0.044 and p = 0.005, respectively), however gender, university type and chronic diseases status had no significant differences observed among KAP. Positive relationships were observed between age and attitudes. Moreover, strong correlations were demonstrated among KAP. Saving time and effort was the main motivation for SM using AI (52.0%). The most utilized AI tool was ChatGPT (77.48%) and painkillers (47.3%) were the most searched drug for SM.

Conclusion

Pharmacy students in this study exhibited relatively low-to-moderate levels of practice regarding SM using AI, despite having moderate self-reported knowledge and positive attitudes. Considering rising reliance on AI-based health platforms, integrating AI-related modules into pharmacy curricula will potentially promote safe, responsible, and effective utilization of AI tools in SM practices.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12909-026-09882-5.

Background

Self-medication (SM) refers to the practice of individuals taking medications to treat their self-identified signs, symptoms or diseases without professional consultation [1, 2]. It remains a global public health concern with a prevalence ranging from 11.7 to 92% across different parts of the world with higher percentages observed in developing countries [3, 4]. SM can lead to a number of issues, from misdiagnosis, delaying important diagnosis or treatment, getting inappropriate therapy, missing drug interactions or contraindications and developing abuse, side effects to antimicrobial drug resistance [1].

Simultaneously, AI is integrated into the health care system, and it is reshaping health care as well as medical research [5]. It shows promise in the interpretation of clinical images, supporting medical diagnosis and analyzing lab results [6]. In the pharmacy field, AI has shown potential to enhance self- and precision medicine, dose adjustment, prediction of therapeutic outcomes, and side effects mitigation [7].

However, the rapid growth of AI in diagnostic decision-making also raises important questions about its limitations, potential risks and impact on the professional role of pharmacists. This creates an emerging educational challenge, as pharmacy students must not only understand safe medication use, but also develop the ability to evaluate AI-generated health information and recognize the limitations of its use. While these tools can support faster and more accurate assessments, concerns remain about overreliance on technology and data privacy [7].

SM has been progressively prevalent among pharmacy students, prompted by their enhanced medical understanding, easy access to drugs, and confidence in self-diagnosis. Appropriate SM can enhance healthcare efficiency and alleviate pressure on healthcare systems. It is crucial to understand pharmacy students’ KAP regarding SM, as they are prospective healthcare professionals who will influence medication-use behaviors in the broader population. In recent years, the internet, particularly AI-driven platforms, has become a principal source of health-related information, with many people relying on online symptoms searches for self-diagnosis and treatment decisions. The efficacy and safety of SM are contingent upon its proper utilization, significantly shaped by individuals’ KAP, as emphasized by Nikitha et al. ¹

Students’ positive perceptions and their demand for AI integration are influenced by the rapid adoption of AI in healthcare. In addition, Pharmacy students represent the future healthcare workforce who should be prepared to utilize AI responsibly. Recommendations from worldwide professional pharmacy organizations, such as the International Pharmaceutical Federation, emphasize the importance of incorporating education of AI into pharmacy curriculum [7]. Still, there are no studies in the region addressing gaps in their attitudes and practices toward this issue, assessing these components is essential to detect misconception, guide curriculum development and design educational strategies for the safe and responsible use of AI in healthcare.

Therefore, this study aims to evaluate the KAP among pharmacy students in Jordan regarding SM using AI. It additionally determines the relationships among KAP components. Moreover, the research examines the association between students’ demographic data and their levels of KAP related to SM using AI, providing insights into factors that may influence views and behaviors within this evolving field.

Methodology

Study design and setting

A cross-sectional, survey-based study was conducted to evaluate pharmacy students’ KAP regarding SM facilitated by AI technologies for personal use. The study targeted pharmacy students enrolled at universities across Jordan between August 2025 and January 2026.

Study population and sample size

Sample size was calculated using OpenEpi^®, assuming a population of approximately 30,000 pharmacy students [8], 95% confidence level, 5% margin of errors of error, and expected proportion of 50%. Based on these assumptions, a sample size of 380 participants was required. A total of 404 completed responses were obtained, exceeding the minimum required sample size and indicating adequacy for the purposes of our study.

Sampling technique and data collection

A convenience sampling method was used. The survey was developed using Google Forms and distributed electronically. Participants were recruited through various channels, sharing the survey link during university lectures, dissemination by academic staff to their students, and peer-to-peer sharing through social media and messaging platforms, such as Facebook/LinkedIn professional groups, and WhatsApp. This approach enabled broad participation across different universities in Jordan. To minimize duplicate responses, participants were instructed to complete the questionnaire only once.

Inclusion and exclusion criteria

Eligible participants included pharmacy students currently enrolled in diploma, bachelor’s, master’s, or doctoral programs at Jordanian universities.

Study instrument

Data collection was performed via a structured, self-administered online questionnaire administered in Arabic, designed to evaluate pharmacy students’ KAP concerning AI-enabled SM. The development of questionnaires was based on previously validated tools and relevant literature [1]. The instrument comprised four sections, with a total of 35 items which assessed AI use in SM for personal use. It consists of 32 Likert-scale items and three descriptive multiple-choice questions. Likert-scale items were involved in the calculation of self-reported knowledge, attitudes and practice, while descriptive multiple-choice questions were analyzed independently and were not involved in the calculation of KAP score (Supplementary file).

Section one included five items capturing demographic variables: gender, age, educational attainment (diploma, bachelor’s, master’s, doctoral), university type (public or private), and the presence of chronic health conditions.

Section two assessed participants’ self-reported knowledge of AI-based SM through nine items. Respondents rated their agreement with statements utilizing a 5-point self-reported Likert scale (1 = Strongly Disagree to 5 = Strongly Agree). The content addressed familiarity with AI concepts in SM, awareness of prevalent health-related AI applications, and specific AI functionalities such as symptom-based medication recommendations, pharmaceutical identification via visual cues, prediction of drug–drug and drug–food interactions, provision of adverse effect information, disease recognition from reported symptoms, and medication adherence reminders. This section also included an item evaluating awareness of the necessity for consulting healthcare professionals prior to following AI-driven advice.

Section three evaluated self-reported practices related to AI-assisted SM, consisting of six scored items plus three additional multiple-choice questions. Participants indicated frequency or agreement using the same 5-point Likert scale. Items explored prior utilization of AI tools for drug recommendations, frequency of SM behaviors, use of AI to determine proper dosages and administration routes, reliance on AI over professional medical consultation, validation of AI-generated information against authoritative scientific sources, and discussions with licensed healthcare providers regarding AI recommendations. The multiple-choice questions identified primary motivators for AI use in SM, the most frequently accessed AI platforms, and common medication classes sought self-treatment.

Section four investigated attitudes towards AI-supported SM through twelve items, rated on the 5-point self-reported Likert scale. Items assessed perceptions of AI’s usability, assistance necessity, perceived effectiveness in expediting decision-making, safety and accuracy considerations, role in determining when to seek medical advice, and potential to minimize medication errors. Further items addressed perceptions of AI’s impact on pharmacy employment opportunities, the necessity for AI literacy within pharmacy education, the importance of ethical AI use, concerns regarding data privacy, and the influence of limited access to electronic devices or the internet on AI application for SM.

Reliability and validity assessment

A pilot study involving 30 pharmacy students was performed to evaluate the reliability and validity of the study instrument. Responses from these participants were excluded from the final analysis.

Cronbach alpha coefficients were used to test internal consistency of study tools’ items. As shown in Table 1 the highest reliability coefficient was noted for attitude scale (α = 0.875), Followed by the knowledge scale (α = 0.804) then practice scale (α = 0.766). However, all reliability coefficients exceeded minimum acceptable level of (0.70) is recommended by Nunnally (1978) [9].

Table 1.

Item-total correlation analysis of KAP tools’

Knowledge tool	Pearson r	Attitude tool	Pearson r	Practice tool	Pearson r
Knowledge 1	0.452*	Attitude 1	0.810**	Practice 1	0.516**
Knowledge 2	0.518**	Attitude 2	0.555**	Practice 2	0.548**
Knowledge 3	0.632**	Attitude 3	0.747**	Practice 3	0.734**
Knowledge 4	0.739**	Attitude 4	0.739**	Practice 4	0.662**
Knowledge 5	0.682**	Attitude 5	0.802**	Practice 5	0.612**
Knowledge 6	0.716**	Attitude 6	0.711**	Practice 6	0.612**
Knowledge 7	0.723**	Attitude 7	0.792**
Knowledge 8	0.735**	Attitude 8	0.745**
Knowledge 9	0.410*	Attitude 9	0.481**
		Attitude 10	0.600**
		Attitude 11	0.609**
		Attitude 12	0.585**

Two forms of validity were conducted including face validity and item-total correlation analysis.

A panel of five expert pharmacists reviewed the individual items for three study instruments for clarity, relevancy, suitability and appropriateness. In addition, a small sample of pharmacy students (N = 7) further assessed face validity at respondent level. They were asked to give their comments on clarity and understandability for three instruments as well as the time needed to complete the questionnaires. Finally, appropriate revisions were made based on their feedback.

Item-total correlation analysis was done to assess the association between each item and the relevant scale. Though, Item-total correlations refer to extent to which individual items are related to the overall construct they are supposed to measure. Pearson product moment correlation coefficients were computed for each item with its corresponding construct.

For the knowledge instrument, item-total correlation coefficients ranged (0.410–0.739) showing moderate to strong, positive and significant correlations. Attitude items exhibited even stronger, positive and significance correlation coefficients and ranged (0.481 to 0.810). Lastly, item-total correlation coefficients for practice tool ranged 0.516-to 0.734 again demonstrating moderate to strong positive and significant correlations (Table 1).

All item-total correlation coefficients were above 0.30, indicating satisfactory item discrimination and internal consistency of the scales. Moreover, this item-total correlation, alongside acceptable reliability coefficients, reinforces the internal coherence and suitability of KAP instruments for use in clinical practice to collect data from the target population.

Data analysis

The data was analyzed using IBM Statistical Package for the Social Sciences (SPSS) Statistics software (Version 28.0). All statistical tests were two tailed and p < 0.05 assumed significant difference. The KAP sections’ scores were converted into percentage and categorized based on Bloom’s cut-off points: low (< 60%), moderate (60–80%), and high (> 80%) (observed value-minimum possible score)/ (Max-Min possible score). Because KAP scores were not normally distributed, non-parametric tests were employed. Mann-Whitney U tests were used to compare KAP median scores between two independent groups such as gender. Whereas Kruskal-Wallis H test was used to compare KAP scores across more than two groups such as educational levels. Spearman correlation was employed to detect the correlation between age and KAP median scores.

Ethical consideration

The study adhered to the principles outlined in the Declaration of Helsinki. Participation was voluntary, and informed consent was obtained electronically prior to data collection. All responses were kept anonymous, and all data were securely stored by the principal investigator. Ethical approval was granted by Institutional Review Board (IRB) of Al-Balqa Applied University (Protocol No (2026/2025/4/56)). No funding was received from any public, commercial, or nonprofit organizations. The authors declare no conflicts of interest. De-identified data and analysis code are available upon request from the corresponding author.

Results

Participants were pharmacy students with varying levels of education, recruited via convenience sampling. A total of 404 out of 411 participants filled out the questionnaire completely (completion rate = 98.3%). The age ranged from 19 to 50 with a mean of 22.33 ± 3.97 years. The majority identified as female (77.5%). About 53.7% of participants had Bachelor of Pharmacy. Most students attended public universities (70.8%). Among the respondents, only 4.7% reported the existence of chronic illness. A summary of socio-demographic data is provided in Table 2.

Table 2.

Sociodemographic characteristics of study’s participants as n (%) or mean (SD)

Variables	Categories	Frequency	Mean (SD)
Age (years)			22.33 (3.97)
Gender	Male Female	91 (22.5) 313 (77.5)
Education level	Diploma in Pharmacy Bachelor of Pharmacy Doctor of Pharmacy (PharmD) Higher degrees in Pharmacy	130 (32.2) 217 (53.7) 37 (9.2) 20 (5.0)
Type of University	Public University Private University	286 (70.8) 118 (29.2)
Presence of chronic illness	No Yes	385 (95.3) 19 (4.7)

The assessment of self-reported knowledge score regarding SM using AI is demonstrated in Table 3. The findings showed a moderate level of self-reported knowledge. (36.64 ± 5.39). The item “I recognize the importance of specialized medical consultation before adopting any medical recommendation from AI” received the highest mean score of 4.53(± 0.75). The lowest mean score was related to this item “I have the knowledge of the popular types of health-related AI applications” (3.79 ± 1.00). Around half of the participants (52.5%, n = 212) had moderate scores of self-reported knowledges, while 158 participants (39.1%) had high scores of self-reported knowledge regarding SM using AI. Only 34 participants (8.4%) had low scores of self-reported knowledges.

Table 3.

Mean score of self-reported knowledge toward self-medication using AI

No	Self-reported knowledge toward self-medication using AI	Mean	SD
1	I am aware of how AI can be used to support self-management of health conditions without direct medical supervision.	4.05	0.95
2	I have the knowledge of the popular types of health-related AI applications	3.79	1.00
3	I know that certain AI applications can suggest medications depending on diagnosis and symptoms	4.13	0.84
4	AI can determine the type of drug based on its shape or external appearance.	3.92	0.98
5	AI can predict the potential drug-drug and drug-food interactions	4.14	0.85
6	AI can remind the patient to take their medications on time.	3.96	1.06
7	AI can realize the probable disease by being provided with the patient’s symptoms.	3.86	0.99
8	AI provides information on the possible side effects caused with medications use.	4.27	0.73
9	I recognize the importance of specialized medical consultation before adopting any medical recommendation from AI.	4.53	0.75
	Total self-reported knowledge toward self-medication using AI	36.64	5.39
	# Levels of self-reported knowledge toward self-medication using AI Low a score ≤ 30 Moderate a score 31–37 High a score ≥ 38	N 34 212 158	% 8.4 52.5 39.1

# Score converted into percentage and categorized based on Bloom’s cut-off points: low (< 60%), moderate (60–80%), and high (> 80%) (observed value-minimum possible score)/ (Max-Min possible score)

Furthermore, the participants demonstrated a moderate level of positive attitudes regarding SM using AI (54.2%, n = 219). The highest mean score item was “I recognize that the importance of providing pharmacy students with training on the ethical use on AI applications” (4.29 ± 0.82). On the other hand, the item “I need help while using AI in SM” received the lowest mean attitude score (3.20 ± 1.2) as detailed in Table 4.

Table 4.

Mean attitudes score toward self-medication using AI

No	Attitudes toward self-medication using AI	Mean	SD
1	I find using AI tools for self-medication to be easy.	4.05	0.91
2R	I need help while using AI in self-medication.	3.20	1.20
3	AI tools help me make faster decisions regarding self-medication.	3.83	0.98
4	I believe that self-medication using AI is safe if used within certain limits.	3.84	0.98
5	I believe that AI provides accurate information that helps in making correct pharmaceutical decisions	3.47	1.06
6	AI tools help me determine when to consult the physician.	3.83	0.96
7	I believe that AI tools can assist in reducing medication errors.	3.72	1.03
8 R	I believe that AI will impact job opportunities in the pharmacy field.	3.74	1.12
9	I believe that AI training should be incorporated in pharmacy curriculum.	4.09	0.97
10	I recognize that the importance of providing pharmacy students with training on the ethical use on AI applications.	4.29	0.82
11 R	I think that using AI tools in self-medication creates concerns for me related to data privacy.	3.91	1.03
12 R	I acknowledge that limited access to electronic devices or the internet may affect the use of AI in self-medication.	4.05	0.87
	Total attitudes toward self-medication using AI	46.02	7.94
	# Levels of attitudes toward self-medication using AI Low a score ≤ 40 Moderate a score 41–51 High a score ≥ 51	N 93 219 92	% 23.0 54.2 22.8

# Score converted into percentage and categorized based on Bloom’s cut-off points: low (< 60%), moderate (60–80%), and high (> 80%) (observed value-minimum possible score)/ (Max-Min possible score)

R Reversed

Moreover, students demonstrated comparable low-to-moderate practice levels at (37.9% (n = 153) and 39.9% (n = 161), respectively. The lowest mean score for using AI to get information about the appropriate dosage or the administration method of the medication 3.40 (± 1.2). While the highest one was for verifying AI- provided drug information with other medical references. (4.08 ± 1.03) (Table 5).

Table 5.

Mean practice scores toward self-medication using AI

No	Practice toward self-medication using AI	Mean	SD
1	I have previously used AI applications and tools to get drug recommendations.	3.54	1.13
2	I frequently use self-medication.	3.57	1.06
3	I use AI to know the appropriate dosage or the administration method of the medication.	3.40	1.20
4	I rely on AI applications rather than consulting the doctor for managing minor illnesses.	3.48	1.29
5	I verify the drug information provided by AI by comparing it with other medical references.	4.08	1.03
6	I discussed the AI recommendations with my licensed professional teachers or pharmacist to ensure their accuracy	3.68	1.21
	Total practice toward self-medication using AI	21.75	5.01
	# Levels of practice toward self-medication using AI Low a score ≤ 20 Moderate a score 21–25 High a score ≥ 26	N 153 161 90	% 37.9 39.9 22.3

# Score converted into percentage and categorized based on Bloom’s cut-off points: low (< 60%), moderate (60–80%), and high (> 80%) (observed value-minimum possible score)/ (Max-Min possible score)

As shown in Table 6, both male and female had similar median scores for self-reported knowledge (36.0), practice (22.0) and attitudes (47.0). There were no significant differences between male and female in the median score of self-reported knowledge (p = 0.226), Practice (p = 0.779) and attitudes (p = 0.911) using Mann-Whitney U tests. Regarding education levels, there were significant differences in self-reported knowledge score (p = 0.044) and Practice scores (p = 0.005). The effect size for education level on self-reported knowledge, as measured by Epsilon-squared, was 0.020. Furthermore, the effect size of education level on practices was 0.0314. The 95% Confidence interval (CI) comparing self-reported knowledge between diploma and bachelor’s degree ranges from 0.21 to 0.79, while for practices, it ranges from 0.76 to 4.24. Students with diploma of pharmacy had the highest median self-reported knowledge score (37.0) followed by students holding bachelor’s degree in pharmacy (36.0), Doctor of Pharmacy (36.0) and higher degrees in pharmacy (35.5). Additionally, students with diploma of pharmacy had the highest median practice scores (23.5). On the other hand, there were no significant differences in the median scores of attitudes among the different education levels (p = 0.285). Regarding the type of university, there were no significant differences between public and private universities in the median scores of self-reported knowledge (p = 0.21), practice (p = 0.387) and attitudes (p = 0.895). They had similar median scores of self-reported knowledge (36.0) and attitudes (47.0).

Table 6.

Association between self-reported knowledge, attitudes, and practices of pharmacy students toward self-medication using AI and their demographics

Variables	Categories	Frequency	Self-reported Knowledge	Attitudes	Practices
			Median	Median	Median
Gender	Male Female	91 313	36.0 36.0	47.0 47.0	22.0 22.0
	Z-value (Up-value)		1.21 (0.226)	0.11 (0.911)	0.28 (0.779)
Education level	Diploma in Pharmacy Bachelor of Pharmacy Doctor of Pharmacy (PharmD) Higher degrees in Pharmacy	130 217 37 20	37.0 a 36.0 b 36.0 35.5	48.0 46.0 46.0 47.5	23.5 a 21.0 b 21.0 20.5
	H-value (H p-value)		8.11 (0.044)	3.79 (0.285)	12.64 (0.005)
Type of University	Public University Private University	286 118	36.0 36.0	47.0 47.0	22.0 21.0
	Z-value (U p-value)		1.24 (0.21)	0.13 (0.895)	0.87 (0.387)
Do you have any chronic illness	No Yes	385 19	36.0 37.0	47.0 47.0	22.0 22.0
	Z-value (Up-value)		0.45 (0.655)	0.62 (0.536)	0.39 (0.698)
Age (years)	rs (p-value)		0.02 (0.646)	0.12 (0.015)	0.03 (0.509)
Self-reported Knowledge	rs (p-value)			0.61 (< 0.001)	0.56 (< 0.001)
Attitudes	rs (p-value)				0.71 (< 0.001)

Bold values indicate statistical significance (p-value < 0.05)

U Mann-Whitney U tests, H Kruskal-Wallis H test, rs Spearman correlation. Bonferroni correction was used to adjust multiple comparisons., a: Diploma in Pharmacy, b: Bachelor of Pharmacy

Students with or without chronic illness had the same KAP median scores. Accordingly, there were no statistical differences between self-reported knowledge (p = 0.655), practice (p = 0.698) and attitude (0.536) median scores (Table 6).

In addition, there is a significant positive association between attitude median scores and age (effect size = 0.12, 95% CI: 0.02 to 0.21; p = 0.015). On the other hand, self-reported knowledge (p = 0.646) and practice (p = 0.509) had no significant correlations with age. There was statistically significant positive correlation amongst KAP variables. Practice (p ≤ 0.001) and attitudes (p ≤ 0.001) demonstrated strong correlation with self-reported knowledge. Moreover, Attitudes displayed a strong association with practice (p ≤ 0.001). The effect size for the relationship between self-reported knowledge and attitudes was 0.61 with a 95% CI (0.543–0.669). Similarly, the effect size for the relationship between self-reported knowledge and practices was 0.56, with 95% CI of 0.487–0.625. While the effect size between Attitude and practices was 0.71, with 95.0% CI of 0.656–0.757 (Table 6).

The most common motivators of AI-related SM use were saving time and effort (52.0%; n = 210) and curiosity and experimentation (31.7%; n = 128). Less common reasons are shown in Fig. 1.

Fig. 1.

Motivators for using AI applications related to SM

The most commonly used AI tool was ChatGPT (77.48%; n = 313) followed by drugs.com (10.46%; n = 43) as illustrated in Fig. 2.

Fig. 2.

AI tools/ digital health information resources are often used for self-medication, diagnosis, consultation, or viewing test results/images

The most frequently searched medications using AI tools for SM were pain killers (47.3%; n = 191) followed by antibiotics (27.7% ;n = 112) and antihistamines (11.6%; n = 47) (Fig. 3).

Fig. 3.

Types of medications searched using AI for self-medication purposes

Discussion

This study investigated KAP related to AI-assisted SM among pharmacy students, emphasizing the growing impact of AI and online health information on SM practices [10]. In our study, the results indicated that female students accounted for the majority of responses, highlighting the prevalence of women in the pharmaceutical field. This observation supports other studies and aligns with the wider demographic trends seen in pharmacy education and practice [11].

The mean self-reported knowledge score was (36.64 ± 5.39), where more than 90% of students were classified within the moderate (52.5%) or high (39.1%) knowledge categories, indicating a generally good level of understanding about AI-assisted SM. The students demonstrated a high level of awareness of AI capabilities, including suggesting medications based on symptoms and predicting drug-drug and drug-food interactions and providing information about potential side effects as well as recognizing the necessity of consulting healthcare professionals before acting on AI recommendations. These results are consistent with recent multi-country survey studies indicating a growing understanding of pharmacy students and faculty members of basic concepts of AI and their recognition of its value as a supportive tool in clinical and pharmaceutical practice despite the variability in the availability of formal training in AI tools [11].

Educational level was significantly associated with self-reported knowledge scores. Diploma students demonstrated slightly higher median knowledge scores than bachelor’s students. However, the reasons for these differences cannot be determined from the present cross-sectional study. Variables such as employment status, prior work experience, academic workload, and exposure to AI-related training were not assessed. Therefore, these findings should be interpreted cautiously and explored further in future studies. Studies show that diploma pharmacy students usually work while studying and are exposed to a heavier academic load, including more specialized courses, than bachelor’s students. In contrast, bachelor’s students typically start working in pharmacies only after completing their full degree, and the pressure from specialized courses during the first two years is lower than that experienced by diploma students [12–15]. But we need to assess effect of work and academic load in future studies.

Students’ attitudes toward SM using AI were generally positive but centered at a moderate level, as reflected in the mean total attitude score of 46.02 ± 7.94 and the largest proportion of students (54.2%) in the moderate category [16]. Items showing strong agreement with incorporating AI training into the pharmacy curriculum and emphasizing ethical use indicate openness to AI as a supportive professional tool, rather than as a threat, which mirrors global findings that pharmacy students are willing to embrace AI when it is appropriately integrated and regulated [17].

At the same time, the answers to questions with reverse scores show concerns about data privacy and its potential impact on job opportunities, which shows that the acceptance of AI is conditionally cautious [18]. Furthermore, A slight positive correlation between age and attitudes indicates that it is statistically significant, suggesting that age was associated with more positive opinions, but that the various demographic groups in this cohort generally have similar attitudes.

Although students showed awareness have positive attitude toward use of AI in SM, their actual usage was limited. They showed a mean practice score of 21.75 ± 5.01 with most students were classified in the low (37.9%) or moderate (39.9%) practice levels, suggesting a gap between self-reported knowledge and actual practice [19]. Although many students reported using AI applications to obtain drug recommendations, dosing information, or to manage minor illnesses, the item means around the mid‑scale suggests that such use is not yet routine or intensive [20].

Importantly, students reported relatively high engagement in verifying AI‑generated drug information with other scientific sources and discussing AI recommendations with licensed professionals, indicating a cautious and responsible pattern of SM behavior [21]. Educational level was also associated with practice scores, with diploma students reporting higher practice scores than other groups. Because factors such as work experience, employment status, and academic workload were not measured, no explanation can be established for this observation. Future studies should examine these potential contributors directly [22–24].

In this study, a strong, statistically significant positive correlation was observed between self-reported knowledge and attitudes, self-reported knowledge and practice, and attitudes and practice, reflecting a coherent KAP pattern. These results suggest that higher self-reported knowledge was associated with more positive attitudes and greater reported use of AI in self-medication. However, the cross-sectional design does not permit conclusions regarding directionality or causality [25]. This structure supports the classic assumption of KAP models: self-reported knowledge, attitudes, and behavior have positive association with one another, provided that enabling conditions (training, access, and guidance) are present [26]. In practical terms, this suggests that educational interventions that simultaneously build knowledge (e.g., about AI capabilities and limitations), shape attitudes (e.g., addressing privacy and job‑related concerns), and offer supervised practice using AI tools may be the most effective way to promote safe, evidence‑based use of AI in SM among pharmacy students [11].

According to the motivations for using AI applications related to SM, Pharmacy students most used it to save time and effort (52.0%) followed by curiosity and experimentation (31.7%), Other motivations such as confidence in AI’s accuracy and capabilities (12.1%), avoid consulting a physician (2.7%) and based on the recommendations of others (1.5%) were reported less frequently. This revealed that students appreciate the availability and ease of AI utilization in SM. This is supported by research showing that pharmacy students and other health students use AI applications such as ChatGPT for gaining information and academic purposes, such as drug information and exam preparation, reflecting the applicability of AI, convenience, usefulness for decision-making, efficiency perceptions, and quick access to health information. However, there were concerns about AI accuracy and its effects on ethics [27–29].

Moreover, our results showed that ChatGPT was the most widely used tool among participants for activities such as SM, medical consultation, diagnosis, and interpretation of test results, with more than three-quarters of respondents reporting its use. In comparison, conventional medical websites, including Drugs.com and Medscape, were used less frequently, and other tools accounted for only a minimal proportion of use. This pattern reflects a substantial shift in health information-seeking behavior toward interactive AI-based platforms, underscoring the scientific importance of examining this trend in the present study [30, 31]. Strong reliance on AI for health decisions raises important concerns, particularly about the accuracy and clinical reasoning of the information and whether it truly reflects evidence-based medical practice. There is a high risk when people rely on AI to manage their own medications, as they may follow inaccurate or inappropriate advice [32]. This can lead them to overlook important symptoms that require a doctor’s attention, which, in turn, can raise the risk of unexpected harm to patients. Accordingly, these issues could increase the chance of unintended harm to patients. The growing tendency to turn to AI rather than traditional medical sources suggests that people’s health‑seeking habits are changing, which can shape how they communicate with doctors, follow treatment plans, and decide when to seek professional care. These trends highlight how important it is to carefully assess how well AI tools work in medical settings, so it is clear whether they truly support patient decisions or contribute to the dissemination of misleading, or potentially unsafe health information [33].

Our study demonstrated that pain killers (47.3%, n = 191), followed by antibiotics (27.7%, n = 112) and antihistamines (11.6%, n = 47), were the most frequently searched medications using AI tools for SM. In accordance with previous studies, painkillers were the most utilized drug for SM among Jordanian students for minor diseases such as tooth pain and headache [34]. Regarding antibiotics, the percentages of SM using antibiotics remain elevated in Jordan, and this is a major contributing factor leading to antibiotic resistance and drug treatment issues [35]. Although the lower prevalence of use of antihistamines for SM, research demonstrated that antihistamines continue to be used for SM without consultation and prescription, especially for allergy and cold [34]. These findings highlight the continuous trend of SM with nonprescription drugs and the importance of regulations that target public health to educate the public about the risks of using these medications, especially with prescribed drugs such as antibiotics.

Beyond healthcare-specific applications, foundational computational concepts and algorithmic thinking remain important for understanding the principles that underpin modern AI systems. Recent work has highlighted the educational value of computational models and innovative approaches to teaching computing concepts, which may contribute to stronger AI literacy and critical evaluation skills among future healthcare professionals [36–40].

Our study has employed cross-sectional questionnaires that limit the ability to establish causal relationship between pharmacy students and the outcomes of our study. Additionally, the self-reported questions in the survey may introduce potential biases, such as recall bias or social desirability bias, where pharmacy students may adjust their responses to align perceived expectations and lack of objective verification of AI use. Our attempt to overcome these biases, included the use of an anonymous survey and ensured students of confidentiality, thereby improving the validity and reliability of the self-reported data. Convenience sampling can potentially increase the risk of selection bias and limited generalizability, as students with high interest in AI maybe have been more likely to participate in this study. Even though the efforts were made to minimize duplicate responses, multiple online submissions could not be assured. Finally, the small subgroup of participants with chronic illness may limit the interpretation of outcomes related to this subgroup.

Conclusions

This study evaluated pharmacy students’ self-reported knowledge, attitudes, and practices regarding self-medication using artificial intelligence. The findings indicate that students generally demonstrated moderate levels of perceived knowledge and favorable attitudes toward AI-assisted self-medication; however, their actual use of these tools remains relatively limited. The positive associations observed among knowledge, attitudes, and practices emphasize the need for educational interventions to foster the safe, ethical, and evidence-based use of AI technologies. Considering issues related to privacy, ethics, and the reliability of AI-generated health information, incorporating AI-focused training into pharmacy curricula may enhance students’ ability to utilize these technologies responsibly and effectively in future professional practice.

Abbreviations

AI

Artificial intelligence

CI

Confidence interval

KAP

Knowledge, Attitudes and Practices

SM

Self-medication

SPSS

The Statistical Package for the Social Sciences

Authors’ contributions

NB: Conceptualization, Writing-original draft, Methodology, Formal analysis AA: Writing-review and editing, MethodologyOA: Writing-original draft, Investigation, Methodology HB: Writing-original draft, InvestigationMB: Writing-review and editing.

Funding

The authors declare that they did not receive any funding while preparing on this manuscript.

Data availability

De-identified data and analysis code are available upon request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board (IRB) of Al-Balqa Applied University, Jordan (Protocol No (2026/2025/4/56)). Electronic informed consent was obtained from all participants prior to data collection.”

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.