Retinal optical coherence tomography (OCT) and optical coherence tomography angiography (OCTA) are promising tools for the early-stage diagnosis of Alzheimer’s disease (AD). These non-invasive imaging techniques are cost-effective and more accessible than alternative neuroimaging tools. However, the current literature lacks an extensive review of AD or cognitive impairment diagnosis using OCT or OCTA. This motivated us to examine recent deep learning studies using the PRISMA approach to systematic review. We used Publish or Perish software to locate relevant research from databases such as Scopus, PubMed, and Web of Science, obtaining an initial pool of 2725 references. We then followed the PRISMA review process to identify twelve relevant studies and two patent applications for detailed analysis. Half of the papers we reviewed described longitudinal mouse studies targeting early AD detection. Whereas earlier research used patient demographics and pre-computed features as inputs to the classical machine learning classifiers, more recent studies tend to employ end-to-end deep learning models with OCT and OCTA image inputs. However, this approach presents issues such as small datasets and an absence of scan- ning standards, which the reviewed literature addresses in various ways. We discuss the lack of open OCT/OCTA datasets (about Alzheimer’s disease) as the main issue impeding progress in the field.