Abstract Background Delaying the transition from minimal cognitive impairment to Alzheimer’s dementia is a major concern in Alzheimer’s disease (AD) therapeutics. Pathological signs of AD occur years before the onset of clinical dementia. Thus, long-term therapeutic approaches, with safe, minimally invasive, and yet effective substances are recommended. There is a need to develop new drugs to delay Alzheimer’s dementia. We have taken a nutritional supplement approach with genistein, a chemically defined polyphenol that acts by multimodal specific mechanisms. Our group previously showed that genistein supplementation is effective to treat the double transgenic (APP/PS1) AD animal model. Methods In this double-blind, placebo-controlled, bicentric clinical trial, we evaluated the effect of daily oral supplementation with 120 mg of genistein for 12 months on 24 prodromal Alzheimer’s disease patients. The amyloid-beta deposition was analyzed using 18F-flutemetamol uptake. We used a battery of validated neurocognitive tests: Mini-Mental State Exam (MMSE), Memory Alteration Test (M@T), Clock Drawing Test, Complutense Verbal Learning Test (TAVEC), Barcelona Test-Revised (TBR), and Rey Complex Figure Test. Results We report that genistein treatment results in a significant improvement in two of the tests used (dichotomized direct TAVEC, p = 0.031; dichotomized delayed Centil REY copy p = 0.002 and a tendency to improve in all the rest of them. The amyloid-beta deposition analysis showed that genistein-treated patients did not increase their uptake in the anterior cingulate gyrus after treatment (p = 0.878), while placebo-treated did increase it (p = 0.036). We did not observe significant changes in other brain areas studied. Conclusions This study shows that genistein may have a role in therapeutics to delay the onset of Alzheimer’s dementia in patients with prodromal Alzheimer’s disease. These encouraging results indicate that this should be followed up by a new study with more patients to further validate the conclusion that arises from this study. Trial registration NCT01982578, registered on November 13, 2013.