UNLABELLED: Therapeutic successes in mouse models of Alzheimer’s disease (AD) largely fail to translate into clinical trials, with experimental drugs rarely validated in human models before being administered to humans. To address this, we developed an accessible method to culture commercially available primary human neurons and astrocytes, along with an amyloid-beta 1–42 (Aβ)-based in vitro AD model. This system enables to reliably culture primary neurons to mature stages of development, essential to model the adult brain and neurodegenerative diseases such as AD. The absence of a blood–brain barrier (BBB) in this model permits evaluation of drug mechanisms of action independently of BBB permeability, thereby informing the feasibility of developing fully BBB-penetrant therapeutics for CNS interventions based on validated senolytic pathways. Using this platform, we evaluated two senolytic regimens previously shown to be effective in AD mouse models: Navitoclax (NAV), which targets the Bcl-2 family of anti-apoptotic proteins, and the dasatinib–quercetin (DQ) cocktail, which inhibits tyrosine kinases and AKT signaling, among other pathways. We also assess the natural killer cell line NK92 to model emerging immune-mediated senescent cell ablation therapies. In synaptically mature cultures, we show that NK92 cells preferentially—but not exclusively—targeted Aβ-treated neurons and astrocytes with senescent-like phenotypes. DQ demonstrated a safe profile for human neurons, but Navitoclax exhibited non-selective neurotoxicity. These findings highlight potential risks associated with developing BBB-permeable therapies based on the mechanisms of NAV and NK cell-mediated cytotoxicity. Our work underscores the critical need for human-relevant models in the AD drug-development pipeline to improve safety and clinical translatability. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s13195-025-01822-7.