Earlier this year, Amazon began using artificial intelligence to feature the highlights of customers’ product reviews.
However, some sellers apparently think this AI tool seems to focus too much on the lows, Bloomberg reported Tuesday (Dec. 19). Merchants said the AI provides inaccurate product descriptions or gives too much weight to bad reviews.
“If you have a handful of negative feedbacks, and the artificial intelligence model summarizes it like it’s a consistent theme, that’s not fair,” Jon Elder, who runs the Amazon seller consulting firm Black Label Advisor, told Bloomberg. “It’s really hard to capture the monetary impact of this, but sellers are not happy.”
While inaccurate product descriptions might be easy to spot, the report noted that the AI’s exaggeration of negative feedback might not be as apparent.
A board game, for example, has a 4.7-star rating, and a three-sentence AI summary of reviews noted that “some customers have mixed opinions on ease of use,” even though just 1% of reviews mentioned ease of use in a critical light, according to the report.
Bloomberg said it reviewed dozens of review summaries and found the AI is inconsistent in analyzing customer comments and generating blurbs, with only some spotlighting critical feedback.
A spokesperson for the company told Bloomberg that it is working on refining the technology based on feedback from sellers and customers.
Amazon launched its AI review summary feature in August, designed to let shoppers see a short paragraph on the product detail page that showcases features and customer sentiment frequently mentioned in reviews to let customers decide if a product is right for them.
It’s part of several AI-powered solutions introduced by the company this year, from using the technology to spot damaged products or to decide where to place inventory for faster shipping and delivery.
PYMNTS also reported in October on the company’s efforts to use AI to attract more third-party sellers via an image generation tool that lets brands create lifestyle imagery that enhances the effectiveness of their ads.
“As an example, consider an advertiser who may possess isolated product images, like a toaster, set against a plain white background,” the report said. “Yet, when this same toaster is incorporated into a lifestyle setting, such as positioned on a kitchen counter alongside a croissant, Amazon said click-through rates can surge as much as 40% when compared to ads featuring conventional product images.”
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.