BackgroundIn the UK, over 6500 people die by suicide each year. In England alone, this is one person every 2 h. Professionals assess risk of suicide in face-to-face contacts with people potentially at risk. The National Confidential Inquiry into Suicide found that most people who took their life were classified as ‘low risk’ in their final contact with mental health services. Training for front-line staff in reducing suicide is a NHS priority. While there is considerable evidence on what to assess when exploring suicidal ideation, there is little evidence on how to ask sensitive questions to effectively identify suicide risk and how to respond in the treatment encounter to reduce patient distress and suicidal ideation. This is critical for identifying risk and putting appropriate care in place.MethodsAn electronic search will be conducted using MEDLINE, CINAHL, Cochrane Library, EMBASE and PsycINFO databases. Controlled studies of effectiveness will be identified using a predefined search strategy. The focus will be on suicidal thoughts/feelings rather than self-harm without intent to die. Two authors will independently screen articles using predefined inclusion and exclusion criteria and relevant data will be extracted using the Cochrane Collaboration data extraction form for randomised controlled trials (RCTs). Discrepancies between the two authors will be resolved by consensus or by consulting a third author at all levels of screening. We will assess the quality of evidence as well as risk of bias. A meta-analysis will be conducted if participants, interventions and comparisons are sufficiently similar, and we will perform the meta-analysis using Stata data analysis and statistical software.DiscussionThe results of this systematic review will be used to guide training and practice for health care professionals.Systematic review registrationPROSPERO CRD42015025867Electronic supplementary materialThe online version of this article (doi:10.1186/s13643-016-0211-y) contains supplementary material, which is available to authorized users.