Even Store Employees Didn't Know... Kindly Answering Difficult Questions, 'AI Effect' Boosts Ad Clicks by 20% Jump

Naver, Korea's First Generative AI Advertisement
Interview with the Developer of 'Clova for Ad'

"What sneakers did actress Jung Yu-mi wear on 'Lee Seo-jin's New York New York'?" This is a question that is difficult to ask store staff even if you want to purchase a product seen on a YouTube channel. It is common to search social networking services (SNS) or blogs to find the product name and then look for it in stores. However, the story changes when generative artificial intelligence (AI) acts as a salesperson. It understands the consumer's language and adds product explanations, saying, "The sneakers Jung Yu-mi wore are Nike's 'Zoom Vomero 5'. They are lightweight and popular for their versatile design."

Photo by YouTube channel Sibooya capture

In January, Naver launched 'CLOVA for AD', an advertisement based on the generative AI 'HyperCLOVA X'. They are currently testing their first advertisement in partnership with the sports brand Nike. When consumers engage in conversations as if talking to store staff, the AI, trained on advertiser data, guides them to products. We asked Jinsuk Lim, Advertising Product Planning Leader; Isak Kim, Business Solution Planning Leader; and Jisoo Jung, Intent Aware System Leader, who lead CLOVA for AD, about the potential of AI advertising.

Photo by YouTube channel Sibooya capture

CLOVA for AD starts with resolving consumers' questions. When asked, "Which sneakers would you recommend for someone who walks a lot?" it provides recommended products along with information about materials and comfort. It suggests related questions to help clarify purchase intent and supports the process through to the final purchase.

As a new experiment, there were many challenges to overcome. First, since it was unclear what questions consumers might ask, setting the scope of possible answers was a problem. Narrowing the scope would fail to satisfy diverse needs, but answering any question would dilute the core function of advertising. Because corporate reputation was at stake, they had to thoroughly prepare for hallucination phenomena where AI might provide incorrect answers. Leader Jung explained, "We requested service reviews from employees unrelated to the project to find an appropriate range for answers," and "To solve hallucination issues, we built multiple safety nets by blocking profanity, competitor information, and more."

They also discovered new possibilities for AI in marketing. They focused on the significant information asymmetry between sellers and consumers. Consumers remember products by nicknames or images of celebrities wearing them, while sellers communicate using cryptic product names. Leader Kim emphasized, "Consumers ask about the model 'Nike Air Force 1 '07 LV8 Athletic Club Rush Orange' by its nickname 'Nike Carrot Force' instead," and "AI plays an important role in understanding consumer language and providing more accurate information."

From the left, Isak Kim, Naver Biz Solution Planning Leader, Jisoo Jung, Naver Intent Aware System Leader, Jinseok Lim, Advertising Product Planning Leader
[Photo by Naver]

There were also instances where questions were asked about brands rather than specific products. Examples include "Why does the Air Jordan 3 have an elephant pattern?" or "What socks go well with these shoes?" These are questions even store staff find difficult to answer. In other words, AI performs roles that are challenging even for store employees.

As consumers' questions were answered, responses began to appear. Among consumers who interacted with CLOVA for AD, 30% moved to the seller's site to search for products. Ads exposed through CLOVA for AD had about a 20% higher click-through rate compared to traditional banner ads.

Naver plans to increase advertisers for CLOVA for AD. Leader Lim said, "In the short term, we will continue testing with advertisers who have secured enough data to reduce hallucinations," and "We are also considering pricing models for technology and advertising." Increasing dwell time by continuing conversations with AI is another goal. To this end, providing personalized ads to consumers is also under consideration. For example, even when asking about the same running shoes, different products would be recommended based on gender or age. Leader Kim added, "Consumers receive personalized information through AI, and advertisers can develop marketing strategies that meet consumer desires," concluding, "It will be an assistant that benefits both advertisers and consumers."