The easiest way to add generative AI to shopping is to build a chatbot. A shopper asks a question, the bot recommends products, compares options and generates an answer. Coupang’s recently disclosed approach is less flashy, but it may fit the act of shopping better.

The key is that the customer does not have to start a conversation.

On product pages, Coupang uses AI to summarize key information, turn complicated specifications into visual explanations and pull recurring strengths and weaknesses out of reviews. In electronics and digital goods, its AI product comparison feature can use browsing history and product data to proactively surface comparable products and explain the differences.

A shopper looking at a 16-inch laptop, for example, might be shown a 17-inch version of the same model or an option with more storage, along with a concise explanation of why the difference might matter.

The significance is not the recommendation itself. It is where the AI sits.

Coupang is treating AI as a way to remove friction from an existing purchase flow rather than as a new destination the shopper has to visit.

Consumers do not want to use AI because they are buying a refrigerator. They want to avoid reading dozens of specifications and hundreds of reviews — and to reduce the chance of making a bad choice. A separate chat interface asks them to learn another interaction pattern. AI that appears only when it is useful on the product page can do its job without demanding attention for the technology itself.

There are obvious risks.

AI-generated summaries can overstate what matters, recommendation criteria can favor sellers unevenly, and aggressive personalization can narrow the shopper’s options. If the system misreads product information or review data, a smoother interface simply produces a bad decision faster.

That is why the metrics to watch are not just “AI feature users.” Conversion, returns, what customers buy after comparing products and customer complaints will say much more about whether the system is actually helping.

Coupang’s experiment offers a practical clue about where generative AI may settle in retail.

The best shopping AI may not be the one that makes customers think, “I should use AI.” It may be the one that leaves them with only the sense that choosing a product became slightly less annoying.