AI-powered verification to strengthen Bengal’s Special Intensive Revision of electoral rolls

The Election Commission of India (ECI) is set to introduce an artificial intelligence–based verification system during the ongoing Special Intensive Revision (SIR) of electoral rolls in West Bengal, with the aim of preventing the inclusion of fake, duplicate, or deceased voters. According to senior officials, the move has been prompted by a surge in complaints regarding the misuse of voters’ photographs, particularly those of migrant workers.

It was stated by officials that AI-enabled facial matching technology would be integrated into the verification process to analyse facial similarities across voter photographs. Through this method, instances of the same individual appearing in multiple locations on the rolls are expected to be detected. The system has been designed to flag cases where an image matches more than one entry, after which field verification will be initiated.

Officials said that complaints about the photographs of migrant labourers being misused during voter registration had increased significantly, making technological intervention essential. To address this, AI tools will be deployed to scan the existing voter image repository and identify abnormal patterns such as duplicate registrations or mismatched demographic details. The technology commonly used in identity authentication systems relies on machine learning algorithms trained to detect unique facial landmarks, even when images vary in lighting, resolution, or angle.

Despite this technological enhancement, it was emphasised that booth-level officers (BLOs) would continue to remain central to the verification process. Door-to-door visits will still be required, and photographs will be taken directly by BLOs to ensure authenticity. Even when booth-level agents (BLAs) submit completed forms, BLOs will be obligated to visit households for signature verification.

It was also stated that strict accountability norms would be applied. If a fake or deceased voter is detected after the enumeration and form-filling stages are completed, responsibility will rest with the BLO of the concerned polling station. BLAs, meanwhile, will be required to obtain handwritten statements from voters confirming that the forms were filled in their presence, ensuring an additional layer of documentary verification.

Facial recognition and AI-driven matching systems have been increasingly adopted across government identification platforms worldwide. The technology works by converting facial features into mathematical representations known as facial vectors. These vectors are compared across large databases to detect duplication or fraudulent identities. In the electoral context, such systems have already been tested in limited capacities in other states for cleaning voter rolls, although full-scale deployment remains experimental. The version now being introduced in West Bengal will operate as a back-end screening tool rather than a public-facing system, ensuring that only field officers make final decisions.

By combining AI-powered detection with human verification, the Election Commission expects the SIR process to become more accurate and resilient against voter fraud, without compromising procedural safeguards or ground-level oversight.