Fake reviews are a real menace for online shoppers. If you have ever bought something online based on glowing reviews only to receive a disappointingly subpar product, you know what I mean. A new study published in the International Journal of Information and Communication Technology proposes an AI-powered system that can not only detect fake reviews, but also trace how they spread.
Why existing tools keep falling short
Most existing fake review detection systems focus on the text of a review. That approach worked for a while, but fake reviewers have gotten smarter. They now pair carefully written text with misleading images to make their reviews look authentic. Text-only tools struggle to catch this, and that’s a real problem for shoppers and honest sellers alike.
The researchers addressed this by building a system that looks at multiple signals at once. It analyzes the review text using two different methods, a text convolutional neural network and pre-trained language models, to capture both surface-level and deeper meaning in the words. It also factors in reviewer behavior, since fake accounts tend to have default profile pictures and system-generated usernames, unlike real users who tend to personalize their accounts.
Can AI really catch a fake image too?
The short answer is yes. Review images are analyzed separately using a residual network, a type of deep learning tool commonly used for processing visuals. Once all these signals are gathered, the system fuses them together to make a final call on whether a review is genuine.
When a review is flagged as fake, a Transformer model kicks in to map its origin and track how far it spread through the network.
Tests on a large dataset from JD.com showed that the system achieved a recognition accuracy of 94.2% and a tracing accuracy of 93.5%, outperforming all existing methods it was compared against. This kind of accuracy could eventually mean fewer misleading reviews and more trustworthy ratings to shop by.