Meesho's AI crackdowns on counterfeit listings, deactivates bad actor seller accounts

Meesho has intensified its crackdown on counterfeit listings as non-compliant listings have seen an 80% reduction since February, now representing a mere 0.1% of platform views.

Leveraging artificial intelligence and data science technologies, social ecommerce marketplace Meesho has deactivated 42 lakh counterfeit listings and 10 lakh restricted products in the past six months, it said in a statement.

Meesho's counterfeiting prevention strategy, "Project Suraksha", lies in its adoption of cutting-edge tools to enable swift detection and flagging of suspicious listings and bad actor seller accounts.

Through these measures, Meesho ensures only legitimate sellers operate on its platform. In 2023, Meesho's AI-driven systems identified and blocked over 12,000 bad actor accounts.

“Our journey towards addressing counterfeits began in 2018 when our systems were simpler, relying more on manual checks and basic level text matching. In 2019, we formed our first dedicated AI team at Meesho, marking the beginning of our AI journey," Sanjeev Barnwal, Founder and CTO, Meesho, said.

"Over the years, we have applied AI across the value chain to make ecommerce simpler and relevant to our customers and sellers,” he added.

The company's technology-driven approach relies on various tools and techniques to combat counterfeit products effectively and maintain a trustworthy marketplace.

One of these tools is Image OCR (Optical Character Recognition), which scans product images to extract text and brand logos for thorough checks. It plays a crucial role in identifying infringing and counterfeit products by investigating proof of authorisations and determining authenticity.

Another tool, Image Match, launched at the beginning of 2023, matches product images against curated brand images and historical data from previous deactivations.

Logo Detection is another tool specifically trained for leading brands present on the platform. It effectively identifies whether a brand's logo is legitimate or fraudulent, helping to ensure customers receive authentic goods.

Fuzzy RegEx (Regular Expression) Matching is employed to broaden the search by matching existing text to alternative keyword matches. This plugs loopholes and fixes gaps like special characters or deliberate misspellings of brand names, making it more challenging for counterfeit products to evade detection.

Additionally, its recently launched Nearest Neighbour Search tool operates similarly to the Image Match function. It identifies the closest match to a particular product or brand, mapping each brand to a cluster and vetting new listings against the original brand images.