The product rating system for online shopping, once a helpful way of gauging how good a product is, has become far less trustworthy to consumers, in part because of the use of bots and generative AI to try to deceive shoppers.
Innovation is needed, according to Donald Valoyi, founder and CEO of homegrown e-commerce platform Zulzi, who made the remarks in a recent interview with the TechCentral Show (TCS).
“Ratings have a problem because everything has scaled: there are a lot of people online, but there are a lot of bots, too,” he said.
Valoyi suggested that one of the ways the ratings system could be made more effective is to link shoppers who know each other and show prospective buyers ratings only from people they know personally, thereby using a community-driven approach to building trust.
According to Bob Group MD Andy Higgins, the sanctity of product review systems is critical to the success of online shopping platforms.
Product reviews significantly influence buyers’ decisions, often serving as a key differentiator in whether a customer proceeds with a purchase or abandons their cart. Authentic reviews build trust and credibility, fostering long-term loyalty. When this system is compromised, the platform’s reputation suffers, leading to a loss of consumer confidence, said Higgins.
Since better ratings lead to increased sales, third-party sellers have an intrinsic incentive to game the system. And if customers can’t trust the ratings, they are likely to use an online store less or migrate to a rival platform.
‘Zero tolerance’
“Amazon has zero tolerance for fake reviews,” said Robert Koen, MD for sub-Saharan Africa at Amazon, in response to questions from TechCentral. “We have robust and longstanding policies that prohibit review abuse, and we suspend, ban and take legal action against those who violate these policies. We consistently monitor and enforce our policies so customers can shop in our store with confidence.”
To help with that enforcement, some online retailers, including Amazon, make use of machine learning models that analyse “thousands of data points”, including sign-ins by sellers and relations to other accounts that may pose as buyers for the sake of adding fake reviews, said Koen. The algorithms also look at review history and other markers of fraud.
Read: New Checkers Sixty60 app to launch as sales soar
According to Amazon, more than 250 million suspected fake reviews were removed from its online store globally in 2023.
Machine learning tools are adept at detecting fake reviews from bots, which are a relatively easy way for sellers to create the impression of a large number of positive reviews. Review cherry-picking, where positive reviews are given more prominence than negative ones, is another way that customer perceptions can be gamed on online shopping platforms.
One way to make product ratings vulnerable to fake reviews is by allowing third-party data to contaminate the system, such as when reviews from a third-party seller’s own site are ingested by an online store. For South African online retailer Takealot Group, keeping third-party reviews away from its platform is key to keeping its reviews sanitised.
“We do not aggregate reviews or use third-party data in any way. You are only able to review a product if you have actually bought it on Takealot. Product reviews are an incredibly important tool for shoppers to make informed decisions about their purchases,” said Karla Levick, head of brand and communications at Takealot Group.
According to Bob Group’s Higgins, the quest for positive outcomes is not the only reason product reviews are gamed; some sellers attack the review system with malicious intent. For example, review farming is when a group of consumers are paid to post reviews about a product. It is harder to detect by machine learning algorithms because the links between the so called consumers of a product and its seller are either non-existent or sparse at best. Unscrupulous sellers sometimes try to tarnish the reputation of competitors by combining review farming with review bombing, where a large number of negative reviews are posted about a rival’s products.
Higgins said generative artificial intelligence tools are a double-edged sword for the integrity of product reviews. On one hand, they can be used to detect fake reviews more accurately and consistently; on the other, AI can also generate fake reviews that seem more authentic to the prospective buyer.
‘Vital role’
“The evolution of product reviews will likely continue as e-commerce platforms seek new ways to enhance trust and transparency. Encouraging verified feedback, providing more nuanced review formats and employing advanced technologies like AI to ensure authenticity will play an increasingly vital role. Ultimately, the credibility of product reviews is essential for maintaining a thriving, competitive e-commerce environment,” said Higgins. – © 2024 NewsCentral Media
