Abstract
Background
Despite rapid advances in artificial intelligence (AI), its adoption in Danish general practice remains limited and decentralized, relying on individual general practitioners’ (GPs’) decisions. This study aimed to evaluate Danish GPs’ acceptance of AI and the importance of factors derived from established technology-acceptance models.
Methods
A cross-sectional web survey was conducted from mid-September 2024 to March 2025, using a culturally adapted version of a previously validated tool. The survey included 42 items: nine background items and 33 items covering 11 factors, including Medical and Non-Medical Performance Expectancy, Effort Expectancy, Social Influence (Medical and Patient), Facilitating Conditions, Perceived Trust, Anxiety, Professional Identity, Innovativeness and Behavioral Intention. Descriptive statistics, subgroup comparisons and Cronbach’s alpha were provided.
Results
A total of 109 GPs and residents responded, with 84.4% (n = 92) completing the survey. Attitudes toward AI were generally neutral to positive, with mean scores above neutral in seven of the 11 factors. The most positive factors measured were Medical and Non-Medical Performance Expectancy. Scores related to Behavioral Intention were also high. Lower scores appeared in Perceived Trust, Facilitating Conditions and Anxiety. Older GPs (60 years and above) reported lower scores on Non-Medical Performance Expectancy and Social Influence (Medical); GPs in smaller cities showed more positive attitudes across several factors. Cronbach’s alpha indicated good internal consistency on all but two scales.
Conclusion
In summary, Danish GPs demonstrate a strong intention to adopt AI. However, issues persist as significant barriers regarding Anxiety, Perceived Trust and Facilitating Conditions. Implementation strategies considering context and clinician demographics are recommended.
Keywords: General practice, general practitioner, AI, accept, survey
Introduction
Primary care worldwide faces increasing pressure from demographic shifts, including an aging population with multiple chronic conditions [1]. As the first point of contact in healthcare, general practice is experiencing rising demands on both its clinical and administrative capacities, causing stress and burnout among general practitioners (GPs) [2–4]. These challenges have heightened interest in technological solutions, especially artificial intelligence (AI), which is often discussed in policy circles as a way to boost efficiency, support clinical decision-making and improve patient outcomes [5]. In this study, AI is conceptualized broadly as an umbrella term encompassing various data-driven technologies that may support clinical and administrative tasks in general practice.
Despite rapid advances in AI and its anticipated benefits, actual systematic adoption of AI in Danish general practice remains limited. At the same time, there are substantial political ambitions, particularly in light of the ongoing Danish healthcare reform, to accelerate the digitalization of the healthcare system, including general practice [6]. However, concerns related to data security, professional liability, the impact on GP–patient relationships, and ethical issues continue to hinder implementation efforts [7,8]. Furthermore, most GPs in Denmark are self-employed and run independently owned clinics [9], making them not only the primary users but also key decision-makers in the potential adoption of new technologies, such as AI. Therefore, implementation in Danish general practice does not depend on centralized decisions, as in hospital settings, but relies on individual GPs’ motivation, perceived usefulness and willingness to invest. This decentralized structure may slow widespread adoption even when technologies are available and potentially beneficial.
While there is growing international interest in understanding health professionals’ perspectives on and use of AI [10–12], empirical data specific to general practice in a Danish context remains limited [8,13]. Gaining insight into the factors that influence GP acceptance of AI is essential for the feedback loop guiding design, development and implementation strategies [14]. Additionally, these insights can help ensure that technological innovations align with the realities of daily clinical practice, preventing wasted efforts, as seen with e.g. the English National Programme for Information Technology (NPfIT) [15].
Research on technology acceptance in healthcare has grown significantly in recent years, particularly in the context of AI [16]. Some studies have employed theoretical frameworks, such as the Unified Theory of Acceptance and Use of Technology (UTAUT) [17], e.g. the study by Cornelissen et al. [18] applied this model, with modifications, in a Dutch healthcare setting to analyze the factors influencing medical professionals’ acceptance of AI. In a sample of 67 hospital-based medical professionals, the study found that Medical Performance Expectancy, defined as the belief that AI would improve clinical performance, was the strongest predictor of intention to use AI. This was followed by Non-Medical Performance Expectancy, Effort Expectancy, Perceived Trust and Professional Identity. Other factors, such as Social Influence (Patients), Anxiety and Innovativeness, were not found to be significant in that context.
While these findings provide valuable insights, the general practice setting differs significantly from hospital environments in terms of organizational structure, access to resources, clinical independence and patient continuity. Thus, despite AI becoming more relevant in general practice, there is a notable lack of empirical research specifically on GPs’ acceptance of AI. Building on the work of Cornelissen et al. [18], this study employs a structured web-based survey to assess the relative importance of factors in determining GPs’ acceptance of AI in Danish general practice. The focus of this study is therefore not on specific AI methods or tools, but on GPs’ perceptions of AI as a general concept. By focusing on perceptions in the Danish general practice setting, the study aims to generate context-specific insights relevant to future design, development and implementation strategies in Denmark.
Materials and methods
Study design
To address the study’s aim, we designed a cross-sectional study. The study report is inspired by A Consensus-Based Checklist for Reporting of Survey Studies (CROSS) [19]. The CROSS checklist is found in Appendix 1.
Data collection methods
The study survey [18] included 42 items: nine single-choice background items and 33 battery items with ordinal response options, addressing factors that influence the acceptability of AI. The complete survey is provided in Appendix 2.
Due to the target population being GPs, the background items included in the survey were chosen based on knowledge and the composition of Danish general practice [20]. In Denmark, a universal healthcare system exists in which general practices provide primary healthcare to patients registered with them. General practice functions as a gatekeeper to secondary healthcare through a referral system. Consultations and treatments are funded by taxes and are provided free of charge to patients. GPs are compensated through a combination of capitation and fee-for-service, negotiated as a collective agreement between the Danish Organization of General Practitioners (PLO) and the Danish regions [9].
The background items covered the GP’s gender (male, female), age (≤40, 41–59, ≥60), seniority (residency or years of practice as a GP) and employment status (practice owner, employed or other). It also addresses the organization of the GP’s affiliated clinic (solo practice, partnership, etc.), the number of GPs working in the clinic, the number of patients registered with the clinic, the location (urban, rural, mixed) and the region (North Denmark Region, Central Denmark Region, Region of Southern Denmark, Region Zealand or Capital Region of Denmark).
The survey items addressed the 11 factors influencing AI acceptance, as detailed in Table 1, inspired by Cornelissen et al. [18].
Table 1.
The 11 factors that are investigated for influential significance in relation to AI in Danish general practice, inspired by Cornelissen et al. [18].
| Factor | Definition |
|---|---|
| Medical performance expectancy | How much an individual believes that using AI solutions will improve the quality of care. |
| Non-medical performance expectancy | How much an individual believes that using AI solutions can improve their productivity, efficiency and communication. |
| Effort expectancy | The degree to which an individual believes in the ease of using AI solutions. |
| Social influence (medical) | The degree to which an individual perceives it essential that other medical organizations or colleagues use AI solutions. |
| Social influence (patients) | The degree to which an individual perceives it essential that patients believe they should use AI solutions. |
| Facilitating conditions | The degree to which an individual believes that the necessary organizational and technical infrastructure exists to support the use of AI solutions. |
| Perceived trust | The degree to which an individual trusts that AI solutions have the ability, integrity and benevolence to provide a service. |
| Anxiety | The degree to which an individual feels comfortable using AI solutions and is confident during their interaction with AI solutions. |
| Professional identity | The degree to which an individual believes that AI solutions will affect their professional identity regarding income, status and career growth. |
| Innovativeness | The degree to which an individual adopts an innovation relatively earlier compared to other members of their social system. |
| Behavioral intention | The level of intent to use AI solutions in general practice. |
The original survey by Cornelissen et al. [18] used a predefined case called ‘AI-powered care pathways.’ In our study, we adapted this survey for a Danish general practice setting and broadly examined the acceptability of AI solutions in this context. As a result, ‘AI-powered care pathways’ was translated as ‘AI solutions.’ In the survey, ‘AI solutions’ were presented and defined to respondents as tools integrated into the electronic health record (EHR) system that use algorithms to support various clinical tasks. Examples of possible implementations of ‘AI solutions’ were attached to the definition, including summarizing clinical records, generating diagnosis codes and triaging blood test results. The term and definition used in the survey were approved upon consensus among the authors of this paper.
Study preparation
A cross-cultural adaptation of the survey by Cornelissen et al. [18] was accomplished with inspiration from Beaton et al. [21] to fit the survey into a Danish general practice setting. The cross-cultural adaptation for this study followed the following stages:
First author NLJ was appointed as translator 1 due to their understanding of the concept being examined. NLJ independently adapted the survey questions and produced a report highlighting the most distinctive and challenging translations. ChatGPT was selected as the second ‘naïve’ translator and was instructed to translate the survey questions into Danish. ChatGPT was explicitly instructed to translate ‘AI-powered care pathways’ into the Danish term for ‘AI solutions.’ However, no written report of distinctive and challenging translations was produced in this case.
NLJ and JLT collaborated to synthesize the results from stage 1, creating a unified survey translation suitable for Danish general practice. A written report was generated to document the synthesis process and identify the discrepancies between the original translations.
Two translators translated the survey into English using the synthesized translation from stage 2. These translators were native English speakers fluent in Danish. Neither translator was familiar with the concepts explored, but received specific instructions regarding their translation tasks. This stage served as a validity check to ensure the translated survey adequately reflected the original version. Discrepancies between the original survey and the back-translated version were highlighted and documented.
The study’s authors established an expert committee to consolidate all versions of the survey and create what is regarded as the pre-final version for pilot testing. The author group discussed and resolved any discrepancies.
The pre-final version of the survey was pilot tested by five medical doctors involved in research in general practice. The participants conducted a think-aloud retrospective pilot test. They first completed the survey, during which the time required was measured (approximately six minutes), followed by an item-by-item assessment. All pilot tests were audio-recorded for the iterative development of the survey.
All reports were gathered by NLJ and kept as documentation of the process.
Sample characteristics
The study population comprised GPs and residents in training working in general practices across the five Danish regions. Prospectively, both groups will be collectively referred to as GPs.
The study applied convenience sampling [22], including all GPs who volunteered to participate. This sampling technique was chosen due to the previously observed challenges associated with recruiting GPs for survey research [23,24]. Based on Cornelissen et al. [18], it was aimed to collect responses from a similar sample size (i.e. approximately 60 respondents). Due to the use of convenience sampling, it is not possible to assess the sample’s representativeness of the study population. However, efforts were made to reach a broad, diverse group of GPs through the dissemination activities.
Survey administration
Data were collected and managed using the Research Electronic Data Capture (REDCap) tool hosted by Aalborg University [25,26]. REDCap is a secure, web-based software platform for data capture in research studies. It provides (1) an intuitive interface for validated data capture; (2) audit trails to track data manipulation and export procedures; (3) automated export processes for seamless downloads to standard statistical packages; and (4) procedures for data integration and interoperability with external sources.
Due to restrictions on access to personal contact information, direct outreach to Danish GPs was not feasible, and a pragmatic recruitment strategy was applied. The survey was disseminated through professional networks, social media platforms (e.g. LinkedIn), and at relevant conferences and union meetings. As part of the recruitment strategy, a financial incentive for participation, equivalent to a 10-minute consultation fee, was offered in accordance with the GP collective agreement on involvement in research studies.
The survey lacked mechanisms to prevent multiple submissions. However, the distribution strategy and target population likely limited the risk of duplicate responses.
Data collection was conducted pragmatically from mid-September 2024 through the end of March 2025. The survey was continuously shared across the platforms listed above throughout the specified period to encourage data collection.
Ethical considerations
The study has been registered with the Aalborg University (AAU) Research Ethics Committee (AAU084-1074333). The research activity has been assessed as falling outside the committee’s criteria for ethical approval and, consequently, is exempt from the committee’s review. The AAU Research Ethics Committee considers the research activity to be ethically justifiable.
Informed consent was obtained implicitly from all respondents. Consent was indicated through voluntary access to the survey link and completion of the survey. Furthermore, the survey was conducted anonymously, and no personally identifiable information was collected. Data were stored securely in the described REDCap system. Only the first author (NLJ) of the research team had access to the data.
Statistical analysis
Baseline characteristics are presented as means and standard deviations for continuous variables and as frequencies and percentages for categorical variables. Each Likert-scale item is described by the frequency and percentage of each response category, along with the item’s mean and standard deviation. We present composite scores as averages of their items, excluding missing values from these calculations.
Reliability is assessed using Cronbach’s alpha, which measures the internal consistency of the scales. Cronbach’s alpha ranges from 0 to 1, with higher values indicating greater internal consistency. A value of 0.70 or above generally reflects acceptable reliability, 0.80 or above indicates good reliability, and 0.90 or above signifies excellent reliability.
We report differences in average composite scores, along with 95% confidence intervals, stratified by the selected variables.
Data management and statistical analysis were performed using Stata 18 (College Station, TX).
Results
A total of 109 GPs or residents in training responded to the survey. Of all responders, 84.4% (n = 92) completed the survey, while 15.6% (n = 17) only responded partially.
Respondent characteristics
Descriptive results are presented in Table 2.
Table 2.
Respondent characteristics.
| Background variable | Incomplete | Complete | Total | p Value |
|---|---|---|---|---|
| n (%) | 17 (15.6) | 92 (84.4) | 109 (100.0) | |
| Gender, n (%) | ||||
| Female | 3 (30.0) | 41 (44.6) | 44 (43.1) | |
| Male | 7 (70.0) | 51 (55.4) | 58 (56.9) | 0.38 |
| Age, n (%) | ||||
| ≤40 years | 5 (50.0) | 29 (31.5) | 34 (33.3) | |
| 41–59 years | 4 (40.0) | 56 (60.9) | 60 (58.8) | |
| ≥60 years | 1 (10) | 7 (7.6) | 8 (7.8) | 0.44 |
| Specialist years, n (%) | ||||
| Resident | 2 (20.0) | 25 (27.2) | 27 (26.5) | |
| ≤10 years | 5 (50.0) | 30 (32.6) | 35 (34.3) | |
| 11–20 years | 1 (10.0) | 28 (30.4) | 29 (28.4) | |
| ≥21 years | 2 (20.0) | 9 (9.8) | 11 (10.8) | 0.37 |
| Employment status, n (%) | ||||
| Employed in a practice | 3 (30.0) | 27 (29.3) | 30 (29.4) | |
| Owner of own practice | 5 (50.0) | 54 (58.7) | 59 (57.8) | |
| Other | 2 (20.0) | 11 (12.0) | 13 (12.7) | 0.75 |
| Practice type, n (%) | ||||
| Solo practice | 2 (20.0) | 11 (12.0) | 13 (12.7) | |
| Partnership practice | 6 (60.0) | 76 (82.6) | 82 (80.4) | |
| Other | 2 (20.0) | 5 (5.4) | 7 (6.9) | 0.15 |
| Specialist count, n (%) | ||||
| 1 | 2 (20.0) | 8 (8.7) | 10 (9.8) | |
| 2–3 | 6 (60.0) | 46 (50.0) | 52 (51.0) | |
| >3 | 2 (20.0) | 38 (41.3) | 40 (39.2) | 0.30 |
| Patient count, n (%) | ||||
| <2000 | 3 (30.0) | 7 (7.6) | 10 (9.8) | |
| 2000–5000 | 6 (60.0) | 40 (43.5) | 46 (45.1) | |
| >5000 | 1 (10.0) | 45 (48.9) | 46 (45.1) | 0.02 |
| Practice location, n (%) | ||||
| Large city (>100,000 inhabitants) | 5 (50.0) | 35 (38.0) | 40 (39.2) | |
| Medium-sized city (20,000–100,000 inhabitants) | 1 (10.0) | 28 (30.4) | 29 (28.4) | |
| Small city (<20,000 inhabitants) | 1 (10.0) | 23 (25.0) | 24 (23.5) | |
| Rural area | 3 (30.0) | 6 (6.5) | 9 (8.8) | 0.04 |
| Region, n (%) | ||||
| North Jutland | 1 (10.0) | 17 (18.5) | 18 (17.6) | |
| Central Jutland | 4 (40.0) | 23 (25.0) | 27 (26.5) | |
| South | 1 (10.0) | 20 (21.7) | 21 (20.6) | |
| Zealand | 2 (20.0) | 13 (14.1) | 15 (14.7) | |
| Capital | 2 (20.0) | 19 (20.7) | 21 (20.6) | 0.74 |
The respondents were evenly distributed by gender, with a slight majority of men (56.9%) compared to women (43.1%). Most respondents were aged 41–59 years (58.8%), followed by those aged under 40 (33.3%). Only 7.8% of respondents were over 60.
The years of experience among respondents were distributed as follows: 26.5% were residents in training, 34.3% had ≤10 years of experience, 28.4% had 11–20 years of experience, and 10.8% had ≥21 years of experience.
Among respondents, 57.8% owned their practice, 29.4% were employed in a practice, and 12.7% had other forms of employment. The majority of respondents (80.4%) worked in partnership practices, while the remainder worked in either solo practices (12.7%) or other types of practices (6.9%). Most respondents worked in practices with 2–3 GPs (51.0%), while 39.2% were in practices with more than three GPs.
Most (39.2%) of the respondents’ clinics were located in a large city (with more than 100,000 inhabitants), while only 8.8% were in a rural area.
Respondents were evenly distributed across Danish regions, with the highest representation in the Central Denmark Region (26.5%) and the lowest in the Zealand Region (14.7%).
Non-completers did not differ significantly from completers regarding gender, age, specialist years, employment status, specialist count and region; however, significant differences (p < 0.05) were observed for patient count and practice location.
General attitudes toward AI.
Main findings are presented in Table 3.
Table 3.
Summary of respondents’ responses.
| Strongly disagree, n (%) | Disagree, n (%) | Neither, n (%) | Agree, n (%) | Strongly agree, n (%) | Mean ± SD | ||
|---|---|---|---|---|---|---|---|
| Medical performance expectancy | AI improves patient care | 0 (0.00) | 3 (3.26) | 17 (18.48) | 54 (58.70) | 18 (19.57) | 3.95 ± 0.72 |
| AI improves communication | 0 (0.00) | 14 (15.22) | 40 (43.48) | 26 (28.26) | 12 (13.04) | 3.39 ± 0.90 | |
| AI supports diverse treatments | 0 (0.00) | 2 (2.17) | 13 (14.13) | 57 (61.96) | 20 (21.74) | 4.03 ± 0.67 | |
| Non-medical performance expectancy | AI improves efficiency | 0 (0.00) | 1 (1.09) | 10 (10.87) | 49 (53.26) | 32 (34.78) | 4.22 ± 0.68 |
| AI reduces administrative work | 2 (2.17) | 5 (5.43) | 19 (20.65) | 41 (44.57) | 25 (27.17) | 3.89 ± 0.94 | |
| AI improves professional communication | 1 (1.09) | 18 (19.57) | 42 (45.65) | 23 (25.00) | 8 (8.70) | 3.21 ± 0.90 | |
| Effort expectancy | Competent to use AI | 0 (0.00) | 12 (13.04) | 12 (13.04) | 51 (55.43) | 17 (18.48) | 3.79 ± 0.90 |
| AI easy to understand | 2 (2.17) | 14 (15.22) | 24 (26.09) | 43 (46.74) | 9 (9.78) | 3.47 ± 0.94 | |
| Quick to learn AI | 2 (2.17) | 12 (13.04) | 11 (11.96) | 47 (51.09) | 20 (21.74) | 3.77 ± 1.01 | |
| Social influence (medical) | Other specialties improved | 1 (1.09) | 4 (4.35) | 40 (43.48) | 42 (45.65) | 5 (5.43) | 3.50 ± 0.72 |
| Colleagues have positive experiences | 4 (4.35) | 13 (14.13) | 21 (22.83) | 45 (48.91) | 9 (9.78) | 3.46 ± 1.00 | |
| Colleagues recommend AI | 8 (8.70) | 21 (22.83) | 31 (33.70) | 26 (28.26) | 6 (6.52) | 3.01 ± 1.06 | |
| Social influence (patients) | Patients excited for AI | 0 (0.00) | 20 (21.74) | 42 (45.65) | 28 (30.43) | 2 (2.17) | 3.13 ± 0.77 |
| Patients want AI inclusion | 3 (3.26) | 24 (26.09) | 39 (42.39) | 23 (25.00) | 3 (3.26) | 2.99 ± 0.88 | |
| Patients positive on AI | 1 (1.09) | 11 (11.96) | 41 (44.57) | 31 (33.70) | 8 (8.70) | 3.37 ± 0.85 | |
| Facilitating conditions | Technical requirements met | 7 (7.61) | 34 (36.96) | 32 (34.78) | 16 (17.39) | 3 (3.26) | 2.72 ± 0.95 |
| Organizations provide AI training | 2 (2.17) | 16 (17.39) | 21 (22.83) | 46 (50.00) | 7 (7.61) | 3.43 ± 0.94 | |
| AI compatible with systems | 4 (4.35) | 20 (21.74) | 33 (35.87) | 27 (29.35) | 8 (8.70) | 3.16 ± 1.01 | |
| Perceived trust | Sensitive data secure | 11 (11.96) | 25 (27.17) | 30 (32.61) | 22 (23.91) | 4 (4.35) | 2.82 ± 1.07 |
| AI recommendations reliable | 4 (4.35) | 26 (28.26) | 33 (35.87) | 27 (29.35) | 2 (2.17) | 2.97 ± 0.92 | |
| AI respects patient integrity | 11 (11.96) | 23 (25.00) | 37 (40.22) | 17 (18.48) | 4 (4.35) | 2.78 ± 1.03 | |
| Anxiety | Feel safe using AI | 0 (0.00) | 7 (7.61) | 33 (35.87) | 41 (44.57) | 11 (11.96) | 3.61 ± 0.80 |
| Understand AI recommendations | 12 (13.04) | 34 (36.96) | 26 (28.26) | 17 (18.48) | 3 (3.26) | 2.62 ± 1.04 | |
| AI won’t affect ethics | 1 (1.09) | 11 (11.96) | 30 (32.61) | 32 (34.78) | 18 (19.57) | 3.60 ± 0.97 | |
| Professional identity | AI affects income | 7 (7.61) | 39 (42.39) | 40 (43.48) | 5 (5.43) | 1 (1.09) | 2.50 ± 0.76 |
| AI affects status | 9 (9.78) | 42 (45.65) | 30 (32.61) | 9 (9.78) | 2 (2.17) | 2.49 ± 0.88 | |
| AI affects professional development | 11 (11.96) | 43 (46.74) | 29 (31.52) | 8 (8.70) | 1 (1.09) | 2.40 ± 0.85 | |
| Innovativeness | First to try new technology | 4 (4.35) | 16 (17.39) | 28 (30.43) | 31 (33.70) | 13 (14.13) | 3.36 ± 1.06 |
| Like experimenting with technology | 4 (4.35) | 11 (11.96) | 18 (19.57) | 41 (44.57) | 18 (19.57) | 3.63 ± 1.07 | |
| Competent with new technology | 2 (2.17) | 7 (7.61) | 25 (27.17) | 38 (41.30) | 20 (21.74) | 3.73 ± 0.96 | |
| Behavioral intention | Want to use AI | 0 (0.00) | 1 (1.09) | 17 (18.48) | 42 (45.65) | 32 (34.78) | 4.14 ± 0.75 |
| Willing to use AI | 0 (0.00) | 0 (0.00) | 5 (5.43) | 53 (57.61) | 34 (36.96) | 4.32 ± 0.57 | |
| Should use AI in practice | 0 (0.00) | 5 (5.43) | 18 (19.57) | 40 (43.48) | 29 (31.52) | 4.01 ± 0.86 |
The color gradient indicates the number of responses, ranging from the fewest (red) to the most (green), and shows the distribution of responses.
Respondents generally expressed neutral to positive attitudes toward AI, with mean scores exceeding 3.00 for seven of the 11 measured factors.
Respondents held positive attitudes toward Medical Performance Expectancy, as evidenced by high average scores, such as ‘AI improves patient care’ (mean = 3.95) and ‘AI supports diverse treatments’ (mean = 4.03). Even higher average scores are observed for Non-Medical Performance Expectancy, including ‘AI improves efficiency’ (mean = 4.22) and ‘AI reduces administrative work’ (mean = 3.89). The highest positive scores were recorded for Behavioral Intention, with high mean scores for ‘Want to use AI’ (mean = 4.14), ‘Willing to use AI’ (mean = 4.32) and ‘Should use AI in practice’ (mean = 4.01).
The lowest mean scores are recorded for Professional Identity with the following statements: ‘AI affects income neg’ (mean = 2.50), ‘AI affects status neg’ (mean = 2.49) and ‘AI affects prof development neg’ (mean = 2.40).
Low mean scores are observed among respondents regarding Anxiety, with the lowest mean for ‘Understand AI recommendations’ (mean = 2.62). Perceived Trust also registered low scores for ‘Sensitive data secure’ (mean = 2.82), ‘AI recommendations reliable’ (mean = 2.97) and ‘AI respects patient integrity’ (mean = 2.78). At the same time, Facilitating Conditions scored low for ‘Tech requirements met’ (mean = 2.72).
Reliability
Cronbach’s alpha, as illustrated in Appendix 3, tests internal consistency and shows acceptable to good consistency for most scales. The scales with the highest alpha consist of Social Influence (Patients) (α = 0.879), Behavioral Intention (α = 0.897) and Innovativeness (α = 0.889). Two scales, Anxiety (α = 0.559) and Social Influence (Medical) (α = 0.668), have internal consistency below 0.70, which may indicate low item coherence or conceptual ambiguity in their construction.
Comparisons based on background variables
The tables for comparisons based on background variables are located in Appendix 4.
Respondents have generally shown agreement across the background variables; however, a few statistically significant and interesting differences are observed.
Considering age and seniority comparisons, anticipating a correlation between GPs older than 60 and those with more than 21 years of experience, these respondents reported lower scores on Non-Medical Performance Expectancy and Social Influence (Medical).
Among genders, male respondents scored higher on Innovativeness.
The most consistent differences observed across groups concern practice location; GPs in small cities (<20,000 inhabitants) generally held more positive attitudes than their counterparts in large cities, which is apparent for the following factors: Medical Performance Expectancy, Non-Medical Performance Expectancy, Social Influence (Medical), Facilitating Conditions, Perceived Trust, Anxiety and Behavioral Intention.
Discussion
Statement of principal findings
The analysis shows that participating GPs generally view AI positively, especially its potential to improve both Medical Performance Expectancy and Non-Medical Performance Expectancy. The high average scores for these factors indicate that GPs believe AI can enhance the quality of healthcare and support non-medical tasks. Likewise, Behavioral Intention also scored high, indicating a strong willingness among respondents to adopt AI solutions in clinical practice.
The lowest average scores were observed for Professional Identity, indicating that GPs do not expect AI to have a significant impact on their income, professional status or career progression. Similarly, low scores were recorded for Anxiety, Perceived Trust and Facilitating Conditions, indicating ongoing skepticism about the ethical and practical issues of AI deployment. While the low Anxiety score on one item suggests uncertainty about understanding the AI recommendations, the scores for Perceived Trust and Facilitating Conditions highlight concerns about the technical, legal and infrastructural requirements for safely integrating AI into general practice.
When comparing background characteristics, GPs generally shared similar views; however, those aged 60 and older reported lower confidence in Non-Medical Performance Expectancy and Social Influence (Medical), indicating greater skepticism about AI’s potential to improve efficiency and collaboration. Additionally, GPs in smaller cities tended to be more optimistic across several areas, including Perceived Trust and Behavioral intention, possibly reflecting greater openness to innovation, as resource limitations make AI solutions more attractive. Lastly, male GPs scored higher on Innovativeness, suggesting a stronger tendency to adopt new technologies earlier than female respondents.
Strengths and weaknesses of the study
A strength of this study is the cross-cultural adaptation of a previously developed and validated survey. Following best practice guidelines for the cultural adaptation of self-report measures [21], the original Dutch-language survey by Cornelissen et al. [18] was carefully translated and tailored for use in a Danish general practice setting. This process ensured not only linguistic accuracy but also cultural relevance, which enhances the validity of responses among participating GPs. Although Denmark and the Netherlands share similarities in their universal healthcare systems [9,27], the clinical and organizational aspects of general practice differ significantly from hospital-based environments, highlighting the need for a thoroughly adapted instrument.
Another strength of the study is its inclusion of a demographically and geographically diverse sample of Danish GPs. Compared to the composition of the Danish GP population [20], a larger proportion of the respondents in this study were male GPs (56.9%), while most GPs in Denmark are female (61.3%); however, a majority of male respondents in this study also scored higher on factors related to Innovativeness, which might explain the higher sample size among male GPs, as a higher interest in AI is observed in the male population in general [28]. Regarding age, most respondents were between 41 and 59 years old, which correlates with the mean age of the general population of GPs in Denmark (51.3 years). Furthermore, a limited number of respondents were aged 60 or older (7.8%), which also aligns with the general population of GPs in Denmark, where only 6.1% are registered as aged 65 or older.
A limitation of this study is its small sample size: 109 GPs started the survey, and 92 completed it. Small sample sizes are a well-documented challenge in GP survey research in Denmark [23,24]. While the achieved sample still allowed for subgroup analyses and yielded internally consistent data, nonresponse remains a potential threat to the representativeness of the findings. A possible explanation for the small sample size could be time constraints among GPs. The short length of the survey (approximately 6 min to complete) and the inclusion of a financial incentive, in line with the GP collective agreement, were intentional design choices to address the potential time barrier associated with clinical responsibilities; however, efforts were clearly not enough to engage more respondents. A proactive approach, along with a clear strategy to recruit additional GPs for future research, would be beneficial.
Furthermore, the study relied on a convenience sampling and web-based recruitment strategy, as direct outreach to all Danish GPs was not feasible due to restrictions on access to personal contact information. As a result, the survey was disseminated through professional networks, social media platforms (e.g. LinkedIn), and at relevant conferences and union meetings. While multiple efforts were made to broaden outreach, the effectiveness of this strategy may have been limited, particularly in reaching less active GPs on digital platforms. This may have introduced a motivation bias [22], where more digitally engaged GPs were over-represented in the already relatively low participation rate.
Lastly, the study’s use of the term ‘AI solutions’ may create conceptual confusion, as it is broad and can be interpreted in various ways. Although ‘AI solutions’ here were defined as supportive tools integrated into the EHR system to assist with medical tasks, with examples of their use, differences in interpretation could still affect responses.
Comparison with prior work
The results of this study, showing positive attitudes toward AI among GPs in Denmark, especially toward Medical Performance Expectancy and Non-Medical Performance Expectancy, align with those of Cornelissen et al. [18], who found that Medical Performance Expectancy, followed by Non-Medical Performance Expectancy, Expected Effort, Perceived Trust and Professional Identity, are the strongest predictors of AI acceptance among medical professionals in the Netherlands. While the study by Cornelissen et al. was conducted in the Netherlands and in a secondary care setting, the similarity in results suggests that certain factors related to acceptance may be consistent across countries and sectors, despite differences in organizational structures, decision-making processes and patient populations.
A scoping review from 2025 by Scipion et al. [29] identified Performance Expectancy and Facilitating Conditions as the most consistent factors influencing clinicians’ acceptance of AI across healthcare sectors in 46 studies. Our findings partly align with this, as Danish GPs also emphasized both Medical and Non-Medical Performance Expectancy as key drivers of AI acceptance. However, unlike the review, Facilitating Conditions received relatively low scores in our survey, suggesting that Danish GPs consider the technical and organizational prerequisites insufficiently established.
Examining primary care in isolation, Scipion et al. [29] found only a few studies with comparable results that specifically focused on the perceptions and experiences of AI in this setting. Notably, Sides et al. [30] reported that there is an overall openness to AI in the primary care sector of the United Kingdom. However, this attitude is for general application and is accompanied by limited confidence in its practical implementation. Blease et al. [31] likewise, found that British GPs were skeptical about AI’s ability to replace clinical judgment, while simultaneously recognizing its value for administrative and routine tasks. This again shows the similarity of the results across borders between Denmark and the United Kingdom.
Our findings also suggest a link between age and AI acceptance, with older GPs reporting lower levels of Non-Medical Performance Expectancy and Social Influence (Medical). This pattern aligns with Rahman et al. [32], who found that younger adults in the general population are more likely to adopt AI for efficiency, multitasking and availability. Conversely, older adults exhibit lower acceptance and greater apprehension toward AI. This likely reflects generational differences in familiarity with technology and openness to innovation. Furthermore, the lower levels of AI acceptance among older GPs may reflect technology-related anxiety [33], which might lead to concerns about increased workload when learning and adapting to new technologies, or a perceived lack of benefit late in their careers. Therefore, age-related differences in digital literacy and previous exposure to health technologies could also influence perceptions of AI.
Lastly, our study also observed higher AI acceptance among GPs in small cities. This positive attitude may be related to Denmark’s decentralized primary healthcare system [20], which has long struggled to recruit GPs to remote areas. The increased willingness to adopt AI might be due to the greater pressures on primary care in these locations. This adaptive approach to AI use could be a strategic response to the increased workload and the need for innovative solutions in less populated regions and communities.
Possible mechanisms and implications of the study
The study’s findings suggest that AI implementation strategies in general practice should be tailored not only to technical infrastructure but also to demographic and organizational characteristics. For example, older GPs may require more targeted training, reassurance and opportunities to build trust in AI systems, particularly regarding data security and the preservation of clinical autonomy. Similarly, GPs in smaller or more resource-constrained clinics may face additional challenges related to time, staffing and access to technical support. Implementation efforts in these settings should therefore include practical tools, guidance and peer-based learning opportunities to help bridge readiness gaps. At a broader level, national strategies aiming to promote AI adoption in general practice must acknowledge these variations rather than applying uniform, top-down approaches. Furthermore, in light of current political ambitions to strengthen digitalization in Danish healthcare [6], including general practice, these findings provide empirical insight into areas where implementation may be most feasible. In particular, the results suggest that early implementation efforts could prioritize AI applications for administrative and routine tasks, where GPs show greater acceptance. Ensuring equitable and sustainable deployment will require sensitivity to contextual factors and active engagement with end-users throughout the process.
Unanswered questions and future research
The findings indicate that GPs’ acceptance of AI is influenced not only by personal attitudes but also by structural and contextual factors unique to their practice environment. While this study contributes to the growing body of knowledge on GPs’ acceptance of AI, it also raises important questions that require further exploration. Although this study focused on GPs, AI adoption is rarely a decision made by a single individual. The implementation of new technology occurs within both interpersonal and organizational settings, and future research should explore how organizational factors, interpersonal trust and shared attitudes affect the adoption and acceptance of AI. This is especially important in high-trust environments, such as general practices, where organizational dynamics and GP–patient relationships can significantly influence openness to innovation.
Additionally, the views of patients on the use of AI in Danish general practice are still largely unknown, despite an increasing focus on patient-centered care and transparency in digital health. Future research could investigate how patients perceive the role of AI in general practice and whether their views align with or differ from those of their GPs.
Supplementary Material
Acknowledgements
Thank you to all participating GPs who took the time to answer the survey.
Funding Statement
This work was supported by the North Denmark Region.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
The data that support the findings of this study are available from the corresponding author, NLJ, upon reasonable request.
References
- 1.van Oostrom SH, Picavet HS, de Bruin SR, et al. Multimorbidity of chronic diseases and health care utilization in general practice. BMC Fam Pract. 2014;15(1):61. doi: 10.1186/1471-2296-15-61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.de Dumast L, Moore P, Snell KI, et al. Trends in clinical workload in UK primary care 2005–2019: a retrospective cohort study. Br J Gen Pract. 2024;74(747):e659–e665. doi: 10.3399/BJGP.2023.0527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ibsen Bruun A, Vedsted P, Pedersen HS, et al. Workload and GP burnout: a survey and register-based study in Danish general practice. BJGP Open. 2024;8(1):BJGPO.2023.0077. doi: 10.3399/BJGPO.2023.0077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Pedersen AF, Nørøxe KB, Vedsted P.. Influence of patient multimorbidity on GP burnout: a survey and register-based study in Danish general practice. Br J Gen Pract. 2020;70(691):e95–e101. doi: 10.3399/bjgp20X707837. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hoff J. Unavoidable futures? How governments articulate sociotechnical imaginaries of AI and healthcare services. Futures. 2023;148:103131. doi: 10.1016/j.futures.2023.103131. [DOI] [Google Scholar]
- 6.Indenrigs- og Sundhedsministeriet . Sundhedsreformen; 2024 [accessed 2024 Nov 15]. Available from: https://www.ism.dk/temaer/sundhedsreformen- 2024
- 7.Arvai N, Katonai G, Mesko B.. Health care professionals’ concerns about medical AI and psychological barriers and strategies for successful implementation: scoping review. J Med Internet Res. 2025;27:e66986. doi: 10.2196/66986. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Jørgensen NL, Merrild CH, Jensen MB, et al. The perceptions of potential prerequisites for artificial intelligence in Danish general practice: vignette-based interview study among general practitioners. JMIR Med Inform. 2025;13(1):e63895. doi: 10.2196/63895. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Pedersen KM, Andersen JS, Søndergaard J.. General practice and primary health care in Denmark. J Am Board Fam Med. 2012;25(Suppl. 1):S34–S38. doi: 10.3122/jabfm.2012.02.110216. [DOI] [PubMed] [Google Scholar]
- 10.Sahoo RK, Sahoo KC, Negi S, et al. Health professionals’ perspectives on the use of artificial intelligence in healthcare: a systematic review. Patient Educ Couns. 2025;134:108680. doi: 10.1016/j.pec.2025.108680. [DOI] [PubMed] [Google Scholar]
- 11.Hogg HDJ, Al-Zubaidy M, Talks J, et al. Stakeholder perspectives of clinical artificial intelligence implementation: systematic review of qualitative evidence. J Med Internet Res. 2023;25:e39742. doi: 10.2196/39742. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Perivolaris A, Adams-McGavin C, Madan Y, et al. Quality of interaction between clinicians and artificial intelligence systems. A systematic review. Future Healthc J. 2024;11(3):100172. doi: 10.1016/j.fhj.2024.100172. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Krogh M, Hentze M, Jensen MSA, et al. Valuable insights into general practice staff’s experiences and perspectives on AI-assisted diabetic retinopathy screening—an interview study. Front Med. 2025;12:1565532. doi: 10.3389/fmed.2025.1565532. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Delgado F, Yang S, Madaio M, et al. The participatory turn in AI design: theoretical foundations and the current state of practice. Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization; 2023 Oct 30- Nov 1; Boston, MA. New York (NY): Association for Computing Machinery; 2023. doi: 10.1145/3617694.3623261. [DOI] [Google Scholar]
- 15.Justinia T. The UK’s National Programme for IT: why was it dismantled? Health Serv Manage Res. 2017;30(1):2–9. doi: 10.1177/0951484816662492. [DOI] [PubMed] [Google Scholar]
- 16.Lambert SI, Madi M, Sopka S, et al. An integrative review on the acceptance of artificial intelligence among healthcare professionals in hospitals. NPJ Digit Med. 2023;6(1):111. doi: 10.1038/s41746-023-00852-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Venkatesh V, Morris MG, Davis GB, et al. User acceptance of information technology: toward a unified view. MIS Quart. 2003;27(3):425–478. doi: 10.2307/30036540. [DOI] [Google Scholar]
- 18.Cornelissen L, Egher C, van Beek V, et al. The drivers of acceptance of artificial intelligence-powered care pathways among medical professionals: web-based survey study. JMIR Form Res. 2022;6(6):e33368. doi: 10.2196/33368. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Sharma A, Minh Duc NT, Luu Lam Thang T, et al. A Consensus-Based Checklist for Reporting of Survey Studies (CROSS). J Gen Intern Med. 2021;36(10):3179–3187. doi: 10.1007/s11606-021-06737-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.The Danish Organization of General Practitioners (PLO) . Fakta om almen praksis 2025; 2025. Available from: https://laeger.dk/foreninger/plo/politik-analyse-og-fakta/fakta-om-almen-praksis
- 21.Beaton DE, Bombardier C, Guillemin F, et al. Guidelines for the process of cross-cultural adaptation of self-report measures. Spine. 2000;25(24):3186–3191. doi: 10.1097/00007632-200012150-00014. [DOI] [PubMed] [Google Scholar]
- 22.Stratton SJ. Population research: convenience sampling strategies. Prehosp Disaster Med. 2021;36(4):373–374. doi: 10.1017/S1049023X21000649. [DOI] [PubMed] [Google Scholar]
- 23.Froelunde AS, Deleuran M, Thomsen JL, et al. The doctors’ perception of skin diseases in general practice in Denmark, with emphasis on atopic dermatitis: a descriptive study. JEADV Clin Pract. 2024;3(4):1095–1101. doi: 10.1002/jvc2.426. [DOI] [Google Scholar]
- 24.Andersen CA, Brodersen JB, Graumann O, et al. Factors affecting point-of-care ultrasound implementation in general practice: a survey in Danish primary care clinics. BMJ Open. 2023;13(10):e077702. doi: 10.1136/bmjopen-2023-077702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Harris PA, Taylor R, Thielke R, et al. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377–381. doi: 10.1016/j.jbi.2008.08.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Harris PA, Taylor R, Minor BL, et al. The REDCap Consortium: building an international community of software platform partners. J Biomed Inform. 2019;95:103208. doi: 10.1016/j.jbi.2019.103208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.van Weel C, Schers H, Timmermans A.. Health care in The Netherlands. J Am Board Fam Med. 2012;25(Suppl. 1):S12–S17. doi: 10.3122/jabfm.2012.02.110212. [DOI] [PubMed] [Google Scholar]
- 28.Russo C, Romano L, Clemente D, et al. Gender differences in artificial intelligence: the role of artificial intelligence anxiety. Front Psychol. 2025;16:1559457. doi: 10.3389/fpsyg.2025.1559457. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Scipion CEA, Manchester MA, Federman A, et al. Barriers to and facilitators of clinician acceptance and use of artificial intelligence in healthcare settings: a scoping review. BMJ Open. 2025;15(4):e092624. doi: 10.1136/bmjopen-2024-092624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sides T, Farrell T, Kbaier D.. Understanding the acceptance of artificial intelligence in primary care. HCI International 202. 3 Posters; 2023 July 23–28, Copenhagen, Denmark; Cham: Springer Nature Switzerland; 2023. [Google Scholar]
- 31.Blease C, Kaptchuk TJ, Bernstein MH, et al. Artificial intelligence and the future of primary care: exploratory qualitative study of UK general practitioners’ views. J Med Internet Res. 2019;21(3):e12802. doi: 10.2196/12802. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Rahman MM, Babiker A, Ali R.. Motivation, concerns, and attitudes towards AI: differences by gender, age, and culture. Web Information Systems Engineering – WISE 20. 24; 2024 December 2–5, Doha, Qatar; Singapore: Springer Nature Singapore; 2025. [Google Scholar]
- 33.Kim H, Freddolino PP, Greenhow C.. Older adults’ technology anxiety as a barrier to digital inclusion: a scoping review. Educ Gerontol. 2023;49(12):1021–1038. doi: 10.1080/03601277.2023.2202080. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author, NLJ, upon reasonable request.