Hidden ads in AI: when does persuasion become manipulation?

LLMs enable advertisers to shape outputs at deeper levels—not just mentioning products, but subtly reframing information, redirecting behavior, or reshaping long-term preferences. Most deployed systems focus only on surface-level product placement, leaving more consequential forms of commercial influence unmeasured and undisclosed. The paper proposes a taxonomy of influence tiers and argues generative AI advertising requires new frameworks for detection, measurement, and trustworthiness.