AI in Income Tax Scrutiny: Faceless Assessment Explained

Tax administration in India has moved a long way from manual file reviews, subjective officer discretion, and random audit selection. Today, tax authorities use Artificial Intelligence, Machine Learning, and Big Data Analytics to run a more transparent, data-driven scrutiny process — one that increasingly requires expert interpretation, not just compliant record-keeping.

From real-time data ingestion to automated scrutiny selection and fully faceless proceedings, AI has turned tax compliance from a retrospective, paper-heavy exercise into something closer to continuous monitoring. And since the Income-tax Act, 2025 came into force on 1 April 2026, this framework is no longer just an executive scheme — it's now written directly into statute.

How AI-Driven Scrutiny Works: The Three Pillars

1. Multi-source data ingestion (AIS/TIS, Form 26AS, GSTN, SFT, bank records, property registries) feeds into
2. an AI and analytics engine (360-degree risk profiling, pattern matching, anomaly detection), which drives
3. automated action (CASS-based scrutiny selection, automated case allocation, faceless assessment under Chapter VIII of the 2025 Act).

1. 360-Degree Taxpayer Risk Profiling (Project Insight)

At the core of this system is centralized data aggregation through Project Insight, which pulls together a taxpayer's financial footprint across multiple channels:

• Annual Information Statement (AIS) and Taxpayer Information Summary (TIS): real-time tracking of stock transactions, mutual funds, dividend payouts, and interest income.
• Cross-portal integration (GSTN and Income Tax): automated reconciliation between turnover reported in GST filings (GSTR-1, GSTR-3B) and revenue declared in ITRs.
• Statements of Financial Transactions (SFT): high-value cash deposits, credit card spending, and real estate purchases reported by banks and registries.

AI algorithms run continuous pattern-matching between declared income and actual transaction/lifestyle data. Discrepancies generate an automated risk score — reducing manual bias, but also meaning a genuine, explainable mismatch (a gift, a loan, a one-off asset sale) can trigger a flag that needs a proper response, not just an assumption that the system is right.

2. Computer-Assisted Scrutiny Selection (CASS)

Rather than selecting returns for audit arbitrarily, machine learning drives Computer-Assisted Scrutiny Selection (CASS), operating within the risk-based selection framework now codified under Section 270(10) of the Income-tax Act, 2025.

• Targeted interventions: returns are categorized into Limited Scrutiny (issue-specific) or Complete Scrutiny based on automated anomaly triggers.
• Common red flags: sudden drops in net profit ratio, high-value refund claims, disproportionate deductions, or TDS credit mismatches.

3. The Faceless Assessment Framework

This is the area where the law itself changed most significantly.

Under the Income-tax Act, 1961, faceless assessment operated as an executive scheme under Section 144B, which meant it could, in principle, be withdrawn or altered by government notification. Under the Income-tax Act, 2025, faceless assessment is now a statutory framework, embedded in Chapter VIII of the Act — with the faceless assessment mechanism itself governed by Section 273, working alongside the regular assessment (Section 270) and best judgment assessment (Section 271) provisions. This gives it permanent legal standing that can't be withdrawn without a Parliamentary amendment.

Key features of the current framework:

• National Faceless Assessment Centre (NFAC) remains the single point of contact, coordinating specialised Assessment, Verification, Technical, and Review Units — the taxpayer and the specific officer handling their case remain anonymous to each other throughout.
• Clarified division of roles: the Jurisdictional Assessing Officer (JAO) now typically handles the initial investigation, while NFAC conducts the final evaluation — a distinction the earlier scheme left ambiguous and which had generated litigation.
• Automated allocation: randomized algorithms assign cases to Assessment Units nationally.
• AI-assisted draft order review: draft assessment orders are evaluated against predefined risk parameters to determine whether they should be finalized directly, sent to a Review Unit, or flagged for further clarification.
• Statutory personal hearing right: this is a meaningful upgrade under the 2025 Act. Previously, a personal hearing (via video conferencing) was discretionary and frequently denied. It is now a codified right that must be granted on request in appropriate cases — giving taxpayers a real opportunity to respond to adverse findings before an order is finalized.

Impact on Taxpayers and Businesses

Why This Raises the Value of Professional Representation, Not Reduces It

It's tempting to read all this automation as making a tax professional less necessary — the system flags issues, so surely you just respond to what it tells you. In practice, it works the other way. An AI-generated mismatch or risk score tells you that something doesn't reconcile — it doesn't tell you why, or what the correct, defensible response is. Interpreting a genuine AIS discrepancy, framing a response to a Section 268 inquiry, or deciding whether to request a personal hearing under the faceless framework are judgment calls that carry real financial consequences if handled incorrectly. As scrutiny becomes faster and more automated, getting the response right the first time matters more, not less.

Best Practices for Taxpayers in an AI-Monitored Environment

1. Conduct pre-filing AIS/TIS reconciliation: verify every transaction listed in your AIS matches your internal books before filing.
2. Harmonize direct and indirect tax data: keep GST return revenue figures consistent with what's shown in your ITR — this is one of the most common automated cross-checks.
3. Maintain digital document trails: retain verifiable invoices, bank statements, and deduction proofs to support your position if flagged.
4. Respond promptly to portal notifications: under the faceless framework, failing to respond to a notice under Section 268 (inquiry before assessment) or Section 270(8) (scrutiny) can result in a Best Judgment Assessment under Section 271 — often at a higher estimated tax liability than a cooperative response would produce.

Final Thoughts

AI-driven scrutiny has fundamentally changed the relationship between tax authorities and taxpayers in India — replacing discretionary manual checks with objective, algorithm-driven analytics that catch mismatches faster and more consistently than before. For businesses and individuals, this makes accurate, synchronized financial reporting a prerequisite, not an afterthought.

But automation flags patterns — it doesn't resolve them. When a notice or discrepancy does surface, having someone who understands both the technology behind the flag and the law behind the response is what keeps a routine reconciliation from turning into a prolonged dispute.

If you've received an AIS mismatch notice, a scrutiny notice, or want a pre-filing reconciliation check before you submit your return, talk to our tax advisory team — we help clients respond to AI-flagged discrepancies correctly the first time.

Note on currency: Section numbers under the Income-tax Act, 2025 reflect the Act as it stands following commencement on 1 April 2026. As with any newly commenced law, CBDT clarifications, rules, and interpretations may continue to develop — please verify current section references before relying on this article for a specific filing or notice response.