CBDT uses AI to track taxpayer behaviour, flags mismatches and nudges non-filers: Report

In FY24, while nine crore tax returns were filed, around 7.42 crore taxpayers viewed their AIS, often checking it 3.5 times on average.

The Income Tax Department is increasingly relying on artificial intelligence to monitor taxpayer behaviour—ranging from tracking online activity on its portal to identifying repeated deduction claims and high-value transactions, according to Central Board of Direct Taxes (CBDT) Chairman Ravi Agrawal.

In an interview with The Indian Express, Agrawal explained that repeated visits to view the Annual Information Statement (AIS)—a consolidated summary of a taxpayer’s financial transactions—could indicate that an individual needs to file a return but hasn’t done so yet, or is struggling to file it correctly despite multiple attempts. This data allows the department to intervene and offer assistance.

Agrawal said the broader aim is to build a “360-degree profile” of each taxpayer using AI to detect patterns and filing anomalies. “The department is sitting on a wealth of information,” he said, noting that nearly 40 crore AIS were generated from approximately 650 crore financial transactions in the last fiscal year alone. Based on AI-led assessments, the department believes there's still considerable room for increasing return filings.

In FY24, while nine crore tax returns were filed, around 7.42 crore taxpayers viewed their AIS, often checking it 3.5 times on average. Agrawal said such data is crucial to the department’s “nudge” strategy—gently encouraging compliance without being intrusive.

One such campaign, rolled out earlier this year, resulted in the withdrawal of illegitimate deduction and exemption claims worth Rs 963 crore. It also led to the collection of an additional Rs 409.5 crore in taxes between April 1, 2023, and June 18, 2025.

In one example cited by Agrawal, a taxpayer had over Rs 1 crore in interest income from fixed deposits reflected in their AIS, but had not filed a return. “When you haven’t even accessed your AIS, we start questioning whether you are aware of your tax responsibilities,” he said, adding that AI models help the department zero in on such profiles.

The data pool is substantial—24 crore AIS statements were generated for 8.5 crore unique PAN holders, many of whom had multiple financial activities such as receiving interest, dividends, salaries, or buying/selling property.

AI is also assisting in validating deduction and exemption claims by cross-referencing them with Form 16, Form 26AS, and the AIS. For instance, the department flagged a case where a taxpayer claimed Rs 18 lakh in HRA exemption, while AIS data showed only Rs 5,000 in monthly rent paid.

Similar inconsistencies were observed in claims for deductions related to treatment of dependents with disabilities. “We’ve found PANs with multiple such exemptions, which we’re now investigating,” Agrawal said.

AI-generated red flags are reviewed by tax officers before any follow-up action is taken. “The goal isn’t to snoop or intimidate, but to leverage data intelligently,” he emphasized.

He also pointed out that improper claims under Section 80G (donations) and for tuition fees are being detected and filtered out through the use of smart analytics.

To make taxpayer communication more efficient, the department is developing a tiered e-assistance model. This would ensure that notices and correspondence are short, highly focused, and entirely digital. “The aim is to make the process simpler and more aligned with taxpayer expectations,” Agrawal told The Indian Express.