Psychotic phenomena are among the most severe and disruptive symptoms of dementias and appear in 30% to 50% of patients. They are associated with a worse evolution and great suffering to patients and caregivers. Their current treatments obtain limited results and are not free of adverse effects, which are sometimes serious. It is therefore crucial to develop new treatments that can improve this situation. We review available data that could enlighten the future design of clinical trials with psychosis in dementia as main target. Along with an explanation of its prevalence in the common diseases that cause dementia, we present proposals aimed at improving the definition of symptoms and what should be included and excluded in clinical trials. A review of the available information regarding the neurobiological basis of symptoms, in terms of pathology, neuroimaging, and genomics, is provided as a guide towards new therapeutic targets. The correct evaluation of symptoms is transcendental in any therapeutic trial and these aspects are extensively addressed. Finally, a critical overview of existing pharmacological and non-pharmacological treatments is made, revealing the unmet needs, in terms of efficacy and safety. Our work emphasizes the need for better definition and measurement of psychotic symptoms in dementias in order to highlight their differences with symptoms that appear in non-dementing diseases such as schizophrenia. Advances in neurobiology should illuminate the development of new, more effective and safer molecules for which this review can serve as a roadmap in the design of future clinical trials.