Towards Equitable AI: Detecting Bias in Using Large Language Models for Marketing