Delivery Apps, The War Against Fake Reviews

Baemin Blocks 25,000 Cases in Last 3 Months... Enhances AI-Powered Review Function

#Kim Kyung-gi (alias), who runs a delivery restaurant in Seoul, recently received a call from a marketing company. The caller offered to write positive reviews for his food on the delivery platform in exchange for a fee per order. Although tempted at first, Kim recorded the call and reported it to the delivery platform operator, thinking that an increase in fake reviews would ultimately harm restaurants.

#Park Jung-soo (alias), a housewife, noticed something strange while trying to order chicken through a delivery app. A store that had recently opened was receiving dozens of reviews daily. At first, she thought it was a popular "hot place" and was about to join the order queue. However, when she clicked on the nicknames of the reviewers who left only positive comments, she found that they were posting multiple reviews several times a day across various regions. She had almost been deceived by fake reviews.

The delivery app industry, which has seen a significant increase in orders due to the spread of the novel coronavirus disease (COVID-19), is waging a war against fake reviews. Since multiple companies compete in the delivery app market, reviews greatly influence actual consumer choices. Taking advantage of this, companies manipulating reviews to gain unfair profits are rampant.

According to Woowa Brothers on the 27th, Baemin blocked 25,000 fake reviews over three months starting in July. Earlier, Woowa Brothers detected about 70,000 suspected fake reviews in the first half of this year alone and took follow-up measures such as blocking review exposure and review-writing privileges. Although the number of fake reviews has somewhat decreased due to focused monitoring and improved review management measures, the industry agrees that continuous efforts are still necessary.

Photo by Getty Images Bank

In response, Baemin plans to enhance its review screening function using artificial intelligence (AI) technology. Last year, Baemin established a dedicated review screening team and is currently filtering out reviews containing profanity, falsehoods, exaggerations, or unrelated content based on AI technology. If reviews are found to include defamation or baseless slander, warnings or requests to block the posts are issued depending on the case. Repeated postings may even result in service usage bans.

In particular, Baemin designated a two-month period starting in August this year for intensive review monitoring to crack down on fake reviews. Store owners using fake review agencies face sanctions such as advertising suspension or contract termination upon detection. Legal action is underway against fake review companies. Additionally, Baemin restricts customers from rewriting reviews for an order once the initial review has been deleted. A Baemin official explained, "This measure is to prevent cases where a single order’s review is repeatedly rewritten to keep a specific review prominently displayed."

Delivery Hero Korea, which operates the delivery apps 'Yogiyo' and 'Baedaltong,' has also developed a system to filter fake reviews using AI. They focused on the fact that many delivery food reviews include photos. They introduced an AI process that automatically classifies photo reviews. This technology was completed through an in-house research project and applies deep learning technology that learns from vast amounts of data to make judgments. This system is a highly advanced model with a data recognition level that goes beyond simple object recognition, achieving a 96% accuracy rate in detecting fake photo reviews. Delivery Hero Korea explained that this system allows consumers to access only reliable information by excluding fake photo reviews when ordering.

An industry insider said, "As both the number of stores on delivery apps and orders are rapidly increasing, the reliability of reviews is directly linked to consumer satisfaction. Therefore, each company is actively strengthening monitoring and detection functions to prevent fake reviews from taking root."

This content was produced with the assistance of AI translation services.