This guide outlines a five-step process for conducting literature reviews, in particular rapid reviews adapted to healthcare policy dialogues. It first recommends clearly formulating the research question, using frameworks such as PESTLE, PICO, SPIDER or SPICE, depending on the object and type of research. The second step involves searching for publications in several databases and grey sources, followed by a reproducible sort on titles, abstracts and then full content, with inclusion and exclusion criteria specified. The third step consists in assessing the quality of the publications by evaluating the validity, relevance, objectivity, methodological rigor and timeliness of the information. The fourth step describes the extraction and organization of data, using a standard table including title, authors, year, location, objectives, type of estimate and key information, and recommends the use of bibliographic management software. The fifth step focuses on contextualizing and synthesizing the results to make them understandable and usable by decision-makers, including adapting recommendations to local realities. The document proposes operational best practices: clarifying the objective, involving knowledge users, searching in at least two databases, defining a review approach adapted to the resources, and checking a sample when extracting data. A checklist summarizes the critical tasks to be documented. The guide points out that the quality of a piece of evidence depends on its validity, not on the reputation of the journal or the country of production. It also advocates transparency and traceability.