Mental diseases, including depression, have been spreading with the growing, nearly epidemic speed, sometimes with life-threatening consequences. This creates a necessity for early detection of textual manifestations of depressive states. Social media represent a milieu with nearly-spontaneous user speech; the latter has become an invaluable source for detecting how people perceive, subjectively experience, and verbally manifest depression. Russia has for decades been among top three countries by the levels of clinical depression, while academic and industrial efforts in the field of depression detection have been relatively scarce and sparce, and virtually never reviewed. Employing both the PRISMA and the scoping review approaches, we trace the development of methods and baselines of detection depression by Russia-based scholarly groups who mostly do research in the Russian language. We identify 52 papers published since the late 2000s and review them for methodologies and baselines reached, textual and behavioral markers of depression explored, and their relation to subjective well-being markers. By that, we detect the ‘AI turn’ in depression detection that has led to critical rise of efficiency in detecting the depression markers, but the findings need to be re-assessed due to attribution fallacies. We also critically assess the papers to detect the lacunas in application of the findings; we show that, till 2025, none of the achievements of several research groups resulted into production of industrially accepted web or mobile applications for doctors or ordinary users. We conclude by insisting on closer connections between academically developed methodologies and social aid services for citizens.