Researchers Test AI for Detecting Fake Cosmetics, Finding Errors but Strong Potential
An experiment using AI to identify counterfeit cosmetics highlights both its ability to spot anomalies and the limitations that still need improvement.
The Hacker News online community shared a Grover Lab article documenting an experiment that used artificial intelligence to identify counterfeit cosmetics on the market. The author submitted photos of genuine and fake Rhode Peptide Lip Tint products to an AI model, Gemini 3.6 Flash, to test its ability to tell them apart.
The experiment found that the AI model could detect minor packaging flaws that people might easily overlook, such as an incorrect postal code or inconsistent information between the box and the product tube. However, it also produced false positives, labeling genuine items as counterfeit. This could stem from the company’s own manufacturing defects or visual interference such as lighting and shadows.
Although the current results remain inconsistent and are not 100% accurate, the author believes the technology has strong potential to become a preliminary verification tool for consumers in the future.
Across the broader industry, Visual AI and Deep Learning have been used in E-commerce for some time to detect counterfeit products by analyzing patterns and logos. At the same time, counterfeiters are also using AI to make fakes more convincing, requiring detection technologies to evolve continuously in response.
This highlights AI’s double-edged nature: both counterfeit investigators and counterfeiters are adopting the technology, meaning Thai consumers may need to be even more cautious when shopping online.