Artificial Intelligence (AI) is redefining marketing through data-driven personalization, automated media buying, and real-time campaign optimization. Yet, the same algorithms that enhance efficiency may unintentionally replicate historical inequities embedded in the data on which they are trained. This research explores whether AI marketing systems reinforce or mitigate consumer stereotypes. Using insights from recent empirical studies, simulated content-generation experiments, and a survey of consumer responses reported in secondary data, the paper identifies how algorithmic bias emerges and how fairness-aware design can reduce it. The findings show that when demographic data are unbalanced, AI tools often reproduce gender- and income-based stereotypes in messaging; however, the introduction of fairness constraints and human-in-the-loop monitoring substantially decreases bias indicators without materially reducing personalization accuracy. The paper concludes that ethical, transparent AI governance is not only a social imperative but a strategic advantage for brands seeking long-term consumer trust.