AI recommendations are no longer a small back-end feature in online shopping. They affect what consumers notice first, how much they search, which products they compare, and how confident they feel before buying. This research treats AI-based product recommendations as marketing tools and as forces that shape behaviour in digital markets. It looks at perceived relevance, transparency, personalization depth, and intrusiveness, then connects these design features with algorithmic trust, perceived autonomy, and privacy concern. Buying behaviour is not treated as one final purchase click. The outcomes include click intention, purchase intention, repeat buying, basket value, product variety, and perceived decision quality. The proposed design uses a structured consumer survey and a scenario-based experiment with different levels of personalization and explanation. That design helps test whether consumers accept AI suggestions because they find them useful or reject them because the system feels invasive or controlling. The main argument is that recommendation systems can reduce search effort and support better decisions, but only when consumers still feel that the final choice is theirs. For marketers, recommendation performance should not stop at sales uplift or click-through rate. Trust, privacy comfort, and freedom of choice affect whether consumers keep using the platform. The paper offers a research model for studying AI recommendations in marketing and behavioural economics, especially in online shopping settings where personalization has become a normal part of buying.