BACKGROUND Online access to medical records by patients can potentially enhance provision of patient-centred care and improve satisfaction. However, online access and services may also prove to be an additional burden for the healthcare provider. AIM To assess the impact of providing patients with access to their general practice electronic health records (EHR) and other EHR-linked online services on the provision, quality, and safety of health care. DESIGN AND SETTING A systematic review was conducted that focused on all studies about online record access and transactional services in primary care. METHOD Data sources included MEDLINE, Embase, CINAHL, Cochrane Library, EPOC, DARE, King's Fund, Nuffield Health, PsycINFO, OpenGrey (1999-2012). The literature was independently screened against detailed inclusion and exclusion criteria; independent dual data extraction was conducted, the risk of bias (RoB) assessed, and a narrative synthesis of the evidence conducted. RESULTS A total of 176 studies were identified, 17 of which were randomised controlled trials, cohort, or cluster studies. Patients reported improved satisfaction with online access and services compared with standard provision, improved self-care, and better communication and engagement with clinicians. Safety improvements were patient-led through identifying medication errors and facilitating more use of preventive services. Provision of online record access and services resulted in a moderate increase of e-mail, no change on telephone contact, but there were variable effects on face-to-face contact. However, other tasks were necessary to sustain these services, which impacted on clinician time. There were no reports of harm or breaches in privacy. CONCLUSION While the RoB scores suggest many of the studies were of low quality, patients using online services reported increased convenience and satisfaction. These services positively impacted on patient safety, although there were variations of record access and use by specific ethnic and socioeconomic groups. Professional concerns about privacy were unrealised and those about workload were only partly so.