In income-tax assessment proceedings, assessing officers (AO) get a full nine months to issue information-seeking notices to scrutinize and verify the claims of expenses, deductions, and incomes as declared by taxpayers in their income tax returns (ITRs).
However, recently concluded assessments for the fiscal year 2022-23 suggest that a major part of this nine-month period was spent in hibernation, followed by the sudden onslaught of final show-cause notices issued close to the 31 March deadline.
Arbitrary claims
Taxpayers’ experience over the last four years, since the introduction of the faceless assessment proceedings, has, by and large, been good, with reasonable and fair assessment orders being passed. But, in 2025, the scrutiny notices and assessment orders indicate growing instances of gross time mismanagement. Also, strange and overreaching assessment queries in the tax notices indicate that assessing authorities could possibly be using generative AI/ChatGPT.
In some of the assessment notices, the taxpayers have even been asked to file details of their monthly ration expenses, including flour, rice, spices, oil, gas, electricity, clothing, shoes, polishes, haircuts, cosmetics, perfumes, student fees, car running expenses, rent, insurance, restaurant visit expenses, etc.
The taxpayers’ practical inability to furnish such extensive and intrusive personal expenditure details has been used to make arbitrary additions to the assessment orders.
In some of the cases, the AOs became valuation experts, working out such high valuations that would give even big merchant bankers a run for their money.
In one case, the AO substituted the actual buyback consideration paid by the taxpayer company with some imaginary and substantially higher figure—perhaps with the help of some AI-generated share valuation methodology that was completely inconsistent with the applicable legislative provisions—and made a huge and arbitrary addition. To test the theory. whether generative AI was used for the valuation methodology, I asked an AI platform to calculate valuation for the same case and it threw up the exact same numbers and methodology.
In another case, the AO treated the long-term capital gains on the sale of immovable property as the taxpayer’s normal business income and denied the benefit of indexation merely on the ground that the said sale was executed by an agreement to sell and not by way of registered sale deed. The taxpayer had no business history of trading in immovable properties.
Another issue plaguing faceless assessments is that many cases are transferred to jurisdictional AOs on an ad hoc basis without providing assessees any reasons.
When JAOs issue arbitrary orders after such transfers, taxpayers have no choice but to file writ petitions before high courts or regular appeals before the commissioner (appeals), which has 600,000 appeals pending.
Proper training required
Income-tax assessment proceedings are quasi-judicial proceedings and, as such, are required to be made in a fair, just, equitable, and judicious manner, with due application of mind. AOs should do a timely, preliminary objectivity and materiality analysis to enable them to ask specific and relevant material queries from taxpayers, in a timely and proactive manner. Taxpayers should be given adequate time to scan and upload their replies and supporting documentary evidence.
Where AI tools are being used by AOs, they should be used to complement their natural intelligence and shouldn’t be relied upon completely in framing assessments. Further, the widely proclaimed benefit of faceless assessments is the random allocation of cases to non-jurisdictional AOs to do away with the element of bias and to safeguard taxpayers from any vulnerability.
In order for it to really work, faceless AOs in smaller cities should be properly trained to handle complex matters involving high-value transactions efficiently.
Views are personal.
Mayank Mohanka is the founder of TaxAaram India, and a partner at S.M. Mohanka & Associates.