General practices in XXXXXXAI-based scribe technologies have, over the past two years, been introduced into general practice at an almost explosive pace, largely driven by technologically enthusiastic clinicians and industry actors. This study explored how AI scribes influence GPs’ clinical practice, focusing on their impact on workflow, consultation routines, and professional responsibility, as well as GPs’ reflections on how technology aligns with their needs, expectations, and ethical considerations.Qualitative, semi-structured interviews were conducted with ten GPs who had used an AI scribe regularly for more than three months. Data were collected in the first quarter of 2025 and analysed using systematic text condensation.GPs reported that AI scribes saved up to two hours a day of writing notes, reduced stress, and improved the structure of consultations, enabling more attention to the patient and less to the computer. Several felt this enhanced patient engagement, and patient narratives were better represented in the note. Some missed their personal writing style and found the generated text overly standardised. The AI occasionally struggled to capture nuances in complex consultations, and transcription errors or hallucinations occurred, highlighting the need for careful review. Concerns were expressed about data privacy and regulatory uncertainty, but the GPs maintained a stance of critical trust. Most viewed AI scribes as a notable innovation, provided that human oversight of the output remained strict.According to the GPs, AI scribes improve documentation efficiency, reduce workload, and allow GPs to focus more on patients. Whether the perceived time savings are realistic when thorough review is required remains uncertain. However, the scribes also introduce new sources of error and may subtly shift clinical practice through reliance on AI-generated text. GPs did not describe any deliberate effort to check for information the scribe might miss. Responsible implementation will depend on clear guidelines, training, and further research to safeguard patient-centred care.GPs emphasised the positive effects of AI scribes and quickly integrated the tool. More research is needed on how patients are represented in AI-generated notes, what information may be omitted, and how the patients’ narratives captured by AI may affect equity in general practice.

Family Medicine relies on continuous therapeutic relationships. While digital health technologies support quality and safety, they also contribute to physician burnout by increasing documentation and alert burdens. The introduction of Artificial/Augmented Intelligence (AI) before, during, and after the consultation creates a third actor in care.  The resulting doctor–patient–AI triad requires human-centred design and governance.To synthesise Human Factors Science (HFS) evidence relevant to AI in Family Medicine, focusing on principles that protect relationship-centred care and physician wellbeing.We performed a focused narrative review using a HFS lens. Findings were organised along the consultation continuum: pre-consultation, during consultation, and post-consultation. The during-consultation phase was further divided into (i) history taking and rapport-building, (ii) data gathering for clinical reasoning and decision support, and (iii) decision making and communicating the plan to the patient.Across all phases of care, AI can support access, continuity, and clinical work.  The most promising gains in relationship-centred care appear during the consultation, where ambient AI and documentation support may reduce administrative burden and save time. If this time is protected, it may contribute to clinician wellbeing and allow greater presence, listening, and empathic engagement, thereby strengthening the doctor–patient relationship. AI-aided clinical decision support systems may also reduce cognitive load by structuring information, supporting differential diagnosis through ranked suggestions, and summarising electronic health records histories and visit outputs to improve situational awareness. However, benefits depend on human oversight at every stage. Key risks include over-reliance, privacy/consent challenges, and transcription or summarisation errors.If time saved is converted into higher throughput rather than protected relational time, potential benefits to the therapeutic relationship may be lost. Implementation, therefore, requires workflow redesign, evaluation including communication outcomes, and co-design with patients and clinicians.The HFS approach can guide Family Medicine teams so that AI can reduce burden, mitigate automation bias, and protect empathy and trust in the consultation. Keywords: Artificial Intelligence, Human Factors and Ergonomics, Primary Health Care

General practice is increasingly adopting ambient voice technology (“AI scribes”) for summarising and documenting medical consultations. Despite potential benefits in efficiency and accuracy of documentation, concerns persist about safety and data security. Unequal uptake and use of AI scribes might exacerbate existing inequalities, and therefore impact equity.The aim of this study was to examine current use and perceptions of AI scribes among GPs in XXX, and to identify GP, practice, and population factors associated with their adoption.We conducted a cross-sectional online survey via a medical research company to explore AI scribe use by GPs (n=598) in XXX and their perceived benefits and risks. We used iterative proportional fitting (raking) of age and current position to improve generalisability. Logistic regression was used to identify factors associated with use, including GP characteristics and local population data retrieved from national statistics. Thematic analysis was applied to open-ended responses.Use of AI scribes was common (current users: 40%; past users: 23%). Current users used AI scribes for an average of 60% of consultations (range 5-100%). Use of AI scribes was more likely in male GPs (OR=1.64 [95% CI: 1.17-2.31]), those regularly working in private practice (OR=2.88 [95% CI: 1.69-5.10]) and working 5-6 clinical sessions per week (OR=2.17 [95% CI: 1.39-3.40]), those who train others GPs (OR=3.10 [95% CI: 2.10-4.65]), and use XXX IT system (OR=1.65 [95% CI: 1.10-2.48]). Perceived benefits around efficiency and timeliness were endorsed across current, past and never users. Concerns about safety and medicolegal risks were common, especially among past and never users.Use of AI scribes in XXX is relatively high despite regulatory issues and recent official cease communication. Exacerbation of inequalities seems unlikely based on local population characteristics, but use varied significantly depending on GP characteristics.Therefore, potentially selective use, use of different functions, and patients’ perspectives on AI scribes need to be examined to ensure equitable use of AI scribes in future primary care.

Artificial Intelligence (AI) is reshaping healthcare, offering tools that can either enhance or threaten the core values of primary care. While AI promises efficiency, unauthorized reliance on algorithms poses risks of bias and depersonalization. To preserve the "human touch" in Family Medicine (FM), practitioners must master not only the technology but also its ethical implications. However, current FM curriculum often lacks a structured framework to teach these essential skills.This review aims to identify and categorize "candidate AI competencies" for FM curricula. It specifically addresses the research question: What knowledge, skills, and attitudes are required for General Practitioners/Family Doctors to use AI effectively while upholding the principles of person-centred comprehensive care?A systematic scoping review was conducted, covering the decade from January 1, 2015, to January 1, 2025. Five databases were searched: SCOPUS, WOS, PubMed, EBSCO, and Google Scholar. The search strategy utilized combinations of keywords including "Artificial Intelligence," "Family Practice," "Curriculum," and "Competency." Inclusion criteria focused on articles defining educational frameworks or training interventions for general practitioners/family doctors. A multi-stage screening process was employed: initially removing duplicates, followed by a title/abstract screening, and concluding with a final full-text evaluation to ensure relevance and trustworthiness.The search revealed a significant surge in literature between 2023 and 2025, driven by Generative AI. Preliminary synthesis identifies three core competency domains: Technical Literacy: Understanding how AI works and its limitations (e.g., hallucinations) to maintain professional autonomy. Ethical Guardianship: Skills to detect algorithmic bias, ensuring equity in patient care. Human-in-the-loop: Using AI as a supportive tool without replacing the empathetic doctor-patient relationship (humanism).The findings suggest that AI competency is not merely technical but deeply ethical. Training programs must shift from elective courses to longitudinal integration. Without these skills, the "fraternity" between doctor and patient risks being disrupted by a "black box" algorithm.Integrating defined AI competencies into the FM curriculum is essential to future-proof primary care. This review offers a roadmap for educators and policymakers to train physicians who are technically proficient yet defenders of humanism and equity in primary healthcare.

With increase of Large-Language-Models (LLM) the focus of its usage and potential in primary care arises. Use of Artificial Intelligence (AI) may increase efficacy and quality of primary care and change the profession and delivery of primary care.Aim of this study was to explore use of AI in daily routine among family medicine (FM) trainers, their attitudes towards AI and their opinion on its potential impact on primary care and postgraduate medical education.All registered FM trainers of the competence center for postgraduate medical education XXX (federal state) XXX 2025 were invited to participate in an online survey. The questionnaire consisted of 52 items about sociodemographic, experience with AI-usage, opinions on AI and educational experiences as trainers. It included open text questions. Data was analyzed descriptively an open-texts semi-quantitatively.A total of 137 FM trainers participated in the survey (response rate 12%), 55% identified as male and 45% as female, with a mean age of 50 years. 75% reported their practices to be fully or predominantly digitalized. AI use was common in private contexts (64%), while 34% reported professional AI use for medical purposes. In total, 43% used AI daily or several times per week. Approximately half used AI for differential diagnosis and/or therapeutic recommendations, and one quarter used it for clinical documentation. Most respondents (91%) anticipated increased healthcare system efficiency through AI. Additionally, 67% notice patients would increasingly consult AI tools before seeking medical care and 82% perceive increasing anxiety related to AI-generated information. One third of trainers address AI use with FM trainees, primarily focusing on application and critical appraisal. 75% endorsed the inclusion of AI literacy in postgraduate medical education curricula.FM trainers already apply AI in clinical tasks and expect further efficiency gains alongside increased patient concerns. They emphasize the need to prepare future family physicians (FP) for competent and critical AI use and strongly endorse integrating AI literacy into postgraduate medical training.Use of AI-models is increasing among FP and patients. FP see benefits and risks for primary care. AI literacy should be part of postgraduate medical education in FM.

The global digital transformation of healthcare, driven by rapid adoption of AI, telemedicine and health data platforms, now severely outpaces regulatory systems’ ability to maintain harmonized terminology, threatening the foundations of evidence-based practice and cross-national comparability. Despite rare successes such as EHR standards (HL7, DICOM), exponential growth in proprietary and undefined terms is producing “terminology complexity” undermining research synthesis, guideline development and patient safety.This study’s primary objective is to formally and quantitatively predict the “point of no return” at which regulatory capacity to govern digital health terminology collapses, irreversibly fracturing the evidence base and raising the risk of clinical semantic confusion. The project tests the hypothesis that terminology complexity is accelerating exponentially while regulatory processes remain static or slow, producing an inevitable gap.A mixed-methods framework was used. Quantitative cycle and scenario analyses parameterized regulatory cycle durations, terminology growth, and critical complexity thresholds, using 30+ sources, five analytic phases, and cross-jurisdictional (n=17 cycles, five countries) validation. Qualitative data included 24 expert interviews triangulated via a modified Delphi process. Intervention scenarios were systematically modeled for sensitivity and robustness.Modeling predicts that by Q3 2027 (median probability 76%; range 68–99%), regulatory systems will be outpaced by new, unmappable terms (threshold >384), with technology adoption cycles shrinking to under 15 months, less than half the median 2.5–3 year regulatory development window. Scenario analysis shows even “paused innovation” is insufficient for recovery; only immediate, globally-coordinated registry and harmonization reform can cut risk below 20%. Real-world signals already show a doubling of failed meta-analyses and exclusion of 60–80% of AI studies from guidelines due to undefined terms.Findings indicate a closing global window of 18 to 24 months for action. Policy interventions must prioritize open-access registries, mandatory semantic mapping, and shorter harmonization cycles. Failure to act ensures a phase shift to fragmented, locally divergent digital medicine.Quantitative analysis demonstrates that digital health’s current trajectory will irreversibly degrade evidence-based practice unless urgent, harmonized regulatory innovation is undertaken. Key recommendations include global real-time terminology registries, pre-market mapping, and funding tied to compliance. This roadmap is essential to safeguard patient safety and restore comparability in digital medicine.

Generative artificial intelligence (GenAI) is rapidly transforming knowledge work in healthcare, yet its penetration into frontline primary care remains poorly understood. Understanding current awareness, readiness, and concerns is essential to guide safe adoption. The Catalan Society of Family and Community Medicine (CAMFiC) conducted a survey to assess how primary care professionals in Catalonia engage with GenAI.To describe the level of awareness, use, attitudes, and perceived training and organisational needs regarding GenAI among primary care professionals in Catalonia.An anonymous online cross-sectional survey was distributed to CAMFiC members between 3 and 17 November 2025. The questionnaire covered demographic characteristics, digital access, frequency and context of GenAI use, perceived benefits and risks, organisational support, and preferred training formats. Descriptive statistics summarised responses and explored variation across age, profession, experience, and teaching or leadership roles.A total of 373 professionals participated (69.2% women), mostly family physicians (89.5%) in urban or semi-urban settings (82.8%). Daily GenAI use was reported by 27.9%, while 52.8% had never used it professionally. Common uses included literature search or synthesis (67%), preparing clinical sessions (37.3%), and drafting or editing text (30%). Among users, ChatGPT was most frequent (80.2%), followed by Copilot (37.8%) and Gemini (24.9%); 17.7% had a paid subscription. Self-rated knowledge was low or very low in 82.6%, and only 20.9% had received formal training. Main barriers were lack of training (80.2%), legal/privacy concerns (55.2%), and poor integration with electronic health records (42.4%).The modest uptake contrasts with the fast expansion of GenAI in other sectors, suggesting primary care may be slowed by structural and regulatory constraints. Interest in documentation and evidence-related uses indicates potential for workload relief, yet widespread knowledge gaps and governance concerns underscore the need for practical training and clear institutional guidance. Improving EHR integration and privacy safeguards may be essential for responsible adoption.Primary care professionals in Catalonia show limited and uneven GenAI use, significant knowledge gaps, and governance uncertainty, highlighting the need for targeted training and institutional frameworks before large-scale implementation.