Abstract: Generative artificial intelligence (AI) has become a powerful and increasingly pervasive information technology for producing digital content. My dissertation investigates how, when, and why the locus of generative AI involvement in content production—whether in expression or substance—affects people's evaluations and decisions in business and professional work contexts. In the first essay, I examine how the use and disclosure of AI augmentation for the expression of online reviews influence consumers’ evaluations of reviews and purchase decisions. Building on the continuum model of impression formation and using multiple experiments, I find that AI augmentation enhances perceived writing quality of a review, but the disclosure of this use reduces perceived review authenticity and sways consumer purchase decisions. I further find that review valence moderates the relative impact of review authenticity (vs. writing quality), such that the impact is greater for positive reviews than for negative reviews. In the second essay, I examine how users evaluate and respond to generative AI systems when the same system produces different outputs in response to identical or semantically equivalent inputs—a characteristic that I refer to as generative variability. I conceptualize generative variability as multidimensional, distinguishing generative expressive variability (variation in expression across multiple outputs) from generative substantive variability (variation in substance across multiple outputs). I examine these two types of variability in grounded content generation tasks—common in individual, workplace, and business settings—in which an existing informational base constrains but does not fully determine the desired communicative output. Extending attribution theory and through multiple experiments, I find that generative expressive variability increases perceived AI creativity and trustworthiness, which in turn enhance user satisfaction with the AI system and can strengthen the intention to use it. By contrast, generative substantive variability decreases perceived AI trustworthiness, leading to lower user satisfaction and weaker intention to use. Together, the two essays advance an inference-based perspective on how people evaluate and respond to the content produced with generative AI, and offer practical insights for online content platforms, generative AI providers, organizations, consumers, and AI users.