Fake online reviews have become a major problem for shoppers around the world.
Whether people are buying electronics, booking hotels, or ordering food, many rely heavily on customer reviews to decide what to buy.
But not all reviews are real. Some are written to unfairly promote products, while others are designed to damage competitors.
Now researchers at the University of East London have developed a new AI system that may help detect fake reviews much more effectively than older methods.
The study, published in FinTech and Sustainable Innovation, describes a new “hybrid fusion” detection model that combines language analysis with behavioral signals to identify suspicious reviews on websites such as Amazon and Yelp.
According to the researchers, fake reviews are becoming harder to spot because modern AI tools can generate convincing text that sounds natural and human.
Traditional systems often look for simple clues such as repeated words, unusual phrases, or obvious spam patterns. However, many fake reviews today are more carefully written and can easily slip through older detection systems.
The new AI model tries to understand both what the review says and how it behaves. Instead of only checking for suspicious wording, the system also looks at patterns connected to dishonest activity.
For example, it examines whether the emotional tone of a review matches the star rating. A review giving five stars but containing negative or mixed language may raise suspicion. The system also considers factors such as review length and other unusual patterns that could suggest fake behavior.
By combining language understanding with behavioral analysis, the researchers say the system gains a much clearer picture of whether a review is genuine.
During testing, the model achieved impressive results. It correctly identified fake reviews with 93% accuracy on Amazon data and 91% accuracy on Yelp data. These results were better than several traditional detection systems examined in the study.
Researcher Dr. Hisham AbouGrad said fake reviews are becoming increasingly sophisticated and more difficult to detect. He explained that combining AI language analysis with behavioral clues provides a more reliable way to identify misleading reviews and improve trust in online shopping platforms.
Co-author Fiza Riaz added that the system goes beyond simply spotting suspicious words. By understanding context and behavior together, the model can recognize more complex patterns linked to deceptive reviews while still protecting genuine customer feedback.
The researchers say future work will focus on training the system with larger and more diverse datasets and testing newer AI technologies. They also hope the tool could eventually work in real time on large online shopping platforms.
If successful, the technology could help make online reviews more trustworthy and help shoppers avoid wasting money on poor-quality or unsafe products.
Source: KSR.
