Psychotic symptoms, defined as the occurrence of delusions or hallucinations, are frequent in Alzheimer disease (AD with psychosis, AD+P), affecting ~ 40% to 60% of individuals with AD. AD+P identifies a subgroup of AD patients with poor outcomes. The strongest clinical predictor of AD+P is a greater degree of cognitive impairment than in AD subjects without psychosis (AD-P). Other frequently replicated correlates of AD+P include elevated depressive symptoms. Although the estimated heritability of psychosis in AD is 61%, the underlying genetic sources of this risk are not known. We report a genome-wide meta-analysis of 12,317 AD subjects, with and without psychosis. Results showed common genetic variation accounted for a significant portion of heritability. Two loci, one in the gene ENPP6 (best SNP rs9994623, O.R. (95%CI) 1.16 (1.10, 1.22), p=1.26x10-8) and one spanning the 3-prime-UTR of an alternatively spliced transcript of SUMF1 (best SNP rs201109606, O.R. 0.65 (0.56-0.76), p=3.24x10-8), had genome-wide significant associations with the risk of psychosis in AD. Psychosis risk in AD demonstrated negative genetic correlations with cognitive and educational attainment and positive genetic correlation with depressive symptoms. We had previously observed a negative genetic correlation with schizophrenia, instead we now found a stronger negative correlation with the related phenotype of bipolar disorder. Psychosis risk in AD was not genetically correlated with AD or other neurodegenerative diseases. These findings provide the first unbiased identification of the association of psychosis in AD with common genetic variation and provide insights into its genetic architecture. Study of the genetic mechanisms underlying the associations of loci in ENPP6 and SUMF1 with psychosis in AD are warranted.