How Algorithm-Driven Recommendations Are Changing Brand Protection

In this article, Crystal Lo, Head of Latin Counsel’s China Desk, examines how algorithm-driven recommendations are beginning to reshape traditional concepts of trademark confusion and brand protection.

For decades, trademark law has relied on one central test: would an "average consumer" be confused? We’ve always assumed this average consumer is a human being. But AI is now taking over many of our everyday tasks – from browsing webpages to comparing prices and making purchases.

But how does that test work when the buyer isn’t making the choice by themselves?

Today, more and more online shopping is enabled by AI recommendation tools, voice assistants, and smart search engines. When someone asks an AI assistant to order "shampoo" or "coffee beans," the algorithm picks the brand – the very moment when the traditional test for likelihood of confusion starts to break down.

Usually, the real risk in an AI-driven marketplace isn’t that a human gets confused by two similar logos, but that the algorithm silently makes the decision for them.

If a buyer asks a smart assistant for a specific brand, but the platform’s algorithm is programmed to suggest a high-margin generic alternative instead, the consumer never gets to compare the two. There isn’t a ‘likelihood of confusion’ under traditional trademark law; they’ve simply been steered away before they even had a chance to choose.

In trademark terms, this changes how we think about initial-interest confusion and point-of-sale decisions. Traditional factors like visual similarity or ‘imperfect recollection’ don’t really apply to an algorithm processing product metadata behind a black box.

As automated shopping tools become more common, brand owners and IP counsel will need to rethink how we present our cases:

Who is being evaluated? If an AI assistant routinely substitutes one brand for another, do we evaluate whether the human was misled, or whether the platform is improperly diverting sales?

Platform liability: Where is the line between fair commercial recommendation and unfair brand substitution when an algorithm prioritizes sponsored products over the brand requested by the user?

Clean brand data: Brand protection is no longer just about monitoring physical shelves or domain names. It’s also about making sure your product metadata is clear enough that recommendation engines don’t misclassify your mark.

AI-driven commerce is inevitable, and it will fundamentally change how products reach consumers.

For trademark professionals, protecting a brand will require looking beyond traditional marketplace confusion and paying close attention to how algorithms interpret, display, and substitute clients’ marks. As the line between human choice and automated recommendation continues to blur, legal frameworks will have to evolve to keep pace.