The success of all-trans retinoic acid (ATRA) therapy for acute promyelocytic leukemia (APL) provides a rationale for using retinoic acid (RA)-based therapy for other subtypes of acute myeloid leukemia (AML). Recently, several studies showed that ATRA may drive leukemic cells efficiently into differentiation and/or apoptosis in a subset of AML patients with an NPM1 mutation, a FLT3-ITD, an IDH1 mutation, and patients overexpressing EVI-1. Because not all patients within these molecular subgroups respond to ATRA and clinical trials that tested ATRA response in non-APL AML patients have had disappointing results, the identification of additional biomarkers may help to identify patients who strongly respond to ATRA-based therapy. Searching for response biomarkers might also reveal novel RA-based combination therapies with an efficient differentiation/apoptosis-inducing effect in non-APL AML patients. Preliminary studies suggest that the epigenetic or transcriptional state of leukemia cells determines their susceptibility to ATRA. We hypothesize that reprogramming by inhibitors of epigenetic-modifying enzymes or by modulation of microRNA expression might sensitize non-APL AML cells for RA-based therapy. AML relapse is caused by a subpopulation of leukemia cells, named leukemic stem cells (LSCs), which are in a different epigenetic state than the total bulk of the AML. The survival of LSCs after therapy is the main cause of the poor prognosis of AML patients, and novel differentiation therapies should drive these LSCs into maturity. In this review, we summarize the current knowledge on the epigenetic aspects of susceptibility to RA-induced differentiation in APL and non-APL AML.