Can AI Detect Tax Fraud Better Than Humans

CA DHIRAJ OSTWAL & CO. | THE BUSINESS STRATEGIST

How India's Tax Authorities Are Using AI and Data Analytics to Detect NonCompliance

A groundlevel assessment from a Chartered Accountant's practice covering AIS reconciliation, GST mismatch notices, and what every taxpayer must do right now.

Author: CA Dhiraj Ostwal

Category: Tax Compliance | GST | IT

Word count: ~1,400 words

The Real Question Is Not 'Can AI Detect Fraud?' It Is 'What Has the System Already Flagged About You?'

Over the last four years, the Income Tax Department and GST authorities in India have moved from officerdriven scrutiny to systemdriven enforcement. The transformation is not hypothetical it is live, it is widespread, and in our practice at CA Dhiraj Ostwal & Co., we are handling its consequences every month. Clients receive systemgenerated notices with prepopulated transaction data, mismatch summaries, and tight response deadlines. The assessing officer has not even looked at the file. An algorithm has.

The question worth asking, therefore, is not whether AI in taxation India can outperform human judgment in the abstract. The operationally relevant question is: what does the system already know about your income, your GST filings, and your financial transactions and is what you have reported consistent with that data?

This article breaks down the analytics infrastructure that Indian tax authorities have deployed, the specific triggers that generate notices, and the practical steps every taxpayer should take to stay ahead of the system.

 

Annual Information Statement (AIS)

Introduced in November 2021, the Annual Information Statement is the most comprehensive taxpayer data aggregation framework the Income Tax Department has ever deployed. It draws information from 46 categories of thirdparty sources banks, mutual fund registrars, subregistrars, companies disclosing dividend payments, employers issuing Form 16, and reporting entities under the Statement of Financial Transactions (SFT). Every significant financial event in your life a fixed deposit renewal, a property sale, a dividend received, a foreign remittance is captured here, often within weeks of the transaction.

What most clients do not realise is that AIS is not just a disclosure statement for the taxpayer's reference. It is the department's discrepancy detection matrix. The moment you file your ITR, the backend system automatically crossreferences your declared income with the AIS data. Any unexplained variance interest income visible in AIS but not declared, capital gains recorded by the registrar but missing from Schedule CG, cash deposits exceeding thresholds that don't correlate with declared turnover generates a risk flag and potentially a deficiency notice or scrutiny selection.

? In Practice: We handled a case where a retired professional's five fixed deposit renewals across a financial year appeared as large credit entries in AIS. The client had reported only the net interest in his ITR, not accounting for the fact that each renewal was treated as a fresh transaction by the reporting bank. A Section 139(9) deficiency notice was generated during return processing. The resolution required a correctly reconciled computation and a written explanation something he could not have navigated without professional guidance.

 

Statement of Financial Transactions (SFT)

Under Section 285BA of the Income Tax Act, specified entities are required to file SFTs with the department. These include banks, NBFCs, mutual fund houses, subregistrars, registrars of companies, and credit card issuers. The SFT threshold triggers are precise and nonnegotiable:

• Cash deposits of 10 lakh or more in savings accounts in a financial year (50 lakh for current accounts)
• Purchase of bank drafts, pay orders, or prepaid instruments exceeding 10 lakh
• Mutual fund purchases aggregating 10 lakh or more
• Property registration involving a transaction value above 30 lakh
• Credit card payments aggregating 1 lakh or more in cash, or 10 lakh or more in any mode
• Foreign currency purchases exceeding 10 lakh in a year

This data flows directly into AIS and forms the raw input layer for the department's riskscoring model. If your ITR income profile does not logically explain these transaction values, the mismatch creates a flag.

 

Taxpayer Information Summary (TIS)

The TIS is the processed output layer a computed income summary derived from AIS data. It serves as the department's benchmark for what your income should look like. If your declared income is significantly lower than the TISderived estimate, the system may trigger an automatic processing adjustment under Section 143(1) or route the case to the assessing officer with a prepopulated discrepancy note. The TIS is also being used increasingly to validate Taxpayer's Assessment Summary before issuing Section 148 notices under the faceless reassessment mechanism.

 

The Goods and Services Tax Network has built a sophisticated analytics infrastructure that runs continuously in the background, reconciling data across GSTR1, GSTR3B, GSTR2B, and the Annual Return (GSTR9). The system's primary function is crossverification and it is designed to surface mismatches without human intervention.

GSTR1 vs. GSTR3B: The Most Common Mismatch Trigger

GSTR1 captures outward supply data declared by the supplier. GSTR2B is the autopopulated credit statement generated for the recipient based on the supplier's GSTR1. When a taxpayer's GSTR3B declares outward tax liability that is lower than what is reflected in the consolidated GSTR1 data, or when ITC claimed in GSTR3B exceeds what is available in GSTR2B, the system generates a DRC01A intimation or a formal DRC01 showcause notice.

GST notices due to mismatch of this nature are among the most frequent notices we handle in our practice. In the majority of cases, the underlying cause is a timing difference a supplier filed GSTR1 late, or a credit note was not correctly reflected in the amendment period. However, the system does not distinguish between intentional suppression and an inadvertent timing gap. That distinction must be established in the reply, backed by supporting documentation.

EInvoicing Data Matching Under IRP

From April 2022 onward, einvoicing under the Invoice Registration Portal (IRP) has been mandatory for entities above specified aggregate turnover thresholds. Every einvoice carries a unique Invoice Reference Number (IRN) generated in real time by the GSTN. The system uses this data to crossverify declared outward supplies. Businesses operating above the einvoicing threshold that report invoices without corresponding IRNs create an immediate anomaly in the GSTN's analytics layer.

? In Practice: We encountered a situation where a midsize trading firm was einvoicing for B2B transactions with larger customers but continued issuing manual invoices for certain cashpaying retail clients. The GSTN's analytics system flagged a gap between IRNgenerated turnover and total GSTR1 turnover. This discrepancy prompted a scrutiny notice asking the firm to explain the difference. The resolution required a detailed turnover reconciliation, a revised compliance procedure, and confirmation that the manual invoices were for B2C transactions below the B2B threshold.

 

ITC Irregularities and RiskBased Scrutiny Selection

The GSTN assigns risk scores to taxpayers based on a combination of parameters: ITC claim behaviour relative to sectoral averages, the compliance rating of their supplier base, filing regularity, and turnover growth patterns. Entities with highrisk scores are routed into the GST analytics system's priority scrutiny queue.

Specific ITCrelated red flags include: claiming ITC from suppliers with a history of nil GSTR1 filings, claiming ITC that does not appear in GSTR2B, sudden spikes in ITC in the March filing (indicating yearend credit adjustment without proper documentation), and ITC ratios that are statistically inconsistent with the entity's declared sector.

 

Based on our practice experience across income tax and GST scrutiny cases, the following transaction patterns most consistently result in systemgenerated notices or departmental queries:

INCOME TAX RED FLAGS   ?  Cash deposits of 10 lakh or more in a savings account during a financial year ?  Mismatch between turnover declared in ITR3/ITR4 and GSTINlinked turnover (the ITD crossreferences GST data) ?  Property purchase value reported by the subregistrar exceeding declared income or net worth ?  Dividends, capital gains, or FD interest visible in AIS but not declared in the return ?  Credit card spends above 10 lakh with no corresponding income explanation ?  Foreign remittances or LRS transactions not reflected in Schedule FSI or Schedule FA

 

GST RED FLAGS   ?  ITC claimed in GSTR3B exceeds autopopulated GSTR2B credit ?  Turnover in GSTR3B is lower than aggregate GSTR1 declared outward supplies ?  Nil GSTR3B filing despite active outward supplies visible through GSTR2B of recipients ?  Refund claims statistically inconsistent with the taxpayer's historical export or sector profile ?  Frequent amendments in GSTR1 after the due date a pattern that attracts scrutiny ?  GSTR9 (annual return) significantly diverging from the aggregate of monthly GSTR3B filings

 

It is important for taxpayers to internalise one critical shift: the majority of notices being issued by both the Income Tax Department and GST authorities today are not officerinitiated assessments. They are systemgenerated algorithmic outputs. An assessing officer has not independently reviewed the file and decided to issue a notice. A programmed parameter has been breached, and the system has automatically generated a response.

This changes the compliance calculus in two important ways.

First, the threshold for scrutiny has dropped significantly. A manual system could realistically examine only a fraction of returns filed. The analytics engine processes every return against its data parameters. Discrepancies that would have gone unnoticed a decade ago a small AIS mismatch, a minor ITC timing difference, an interest income not declared are now flagged automatically at scale.

Second, the response window is compressed and the cost of nonresponse is severe. Systemgenerated notices typically carry 15 to 30day response deadlines. In our practice, we have seen cases where clients dismissed the initial DRC01A as a routine communication and did not engage a professional promptly. By the time they came to us, the response window had closed and the department had issued a demand order. Reopening a demand order is significantly more resourceintensive than responding to the initial notice correctly.

A note from our practice: We have also encountered situations where clients responded to systemgenerated notices without professional guidance, providing explanations that while factually correct used language or admitted facts that created additional compliance exposure. Tax notices require structured, legally calibrated responses, not informal clarifications.

 

1. Reconcile Your AIS Before Filing Your ITR Without Exception

Download your Annual Information Statement from the income tax compliance portal before preparing your return. Verify each entry. If any transaction is incorrectly reported by a bank, registrar, or employer submit feedback within the AIS portal to mark it as inaccurate or duplicated. Unaddressed discrepancies in AIS do not disappear; they become the basis of notices postfiling. AIS reconciliation is no longer optional it is a prerequisite step in responsible return preparation.

2. Conduct Monthly GSTR1 vs. GSTR2B Reconciliation

Do not leave GST data reconciliation for the annual return. Monthly reconciliation between your outward supply records and your GSTR2B ensures that credit note adjustments, supplier late filings, and amendment corrections are captured in the correct period. Reconciliation gaps discovered late during GSTR9 preparation often cannot be reversed cleanly within the return filing framework and may require written explanations to the department.

3. Treat EInvoicing Compliance as ZeroException

If your aggregate turnover crosses the applicable einvoicing threshold, every B2B supply must carry an IRN. There are no informal exemptions and no tolerance in the GSTN's reconciliation logic. Ensure your accounting software is integrated with the IRP, and that invoicelevel IRN data is archived for audit purposes.

4. Engage a Professional Review Before Every Statutory Filing

For businesses with highvalue asset purchases, significant ITC claims, multiple income streams, capital gains, or foreign assets, a professional review before filing is not a discretionary service it is a risk management imperative. A CA reviewing your return before submission is in a materially better position than a CA brought in to respond to a notice after the fact.

 

The evolution of AI in taxation India from the AIS and SFT infrastructure of the Income Tax Department to the GSTN's riskbased scrutiny engine represents a structural change in how compliance is enforced. The system is not reactive. It does not wait for an officer to notice something irregular. It processes every return, crossreferences it against multiple thirdparty data sources, and generates flags on mismatches automatically.

Taxpayers who approach their returns as annual paperwork exercises without reconciling AIS, without reviewing GSTR2B, without verifying einvoice data integrity are operating with significant exposure. The system will find the discrepancy; the question is whether you address it proactively or respond to it under notice.

At CA Dhiraj Ostwal & Co., we work with businesses and individuals across income tax and GST compliance from prefiling reviews and AIS reconciliation to drafting replies to notices, appeals before CIT(A), and GST adjudication proceedings. If you have received a notice, an AIS mismatch intimation, or a DRC01 from the GST department, do not delay your response.

Contact CA Dhiraj Ostwal & Co. The Business Strategist Shivajinagar, Pune | cadhirajostwal.com | @cadhirajostwal   For professional assistance with tax notices, AIS reconciliation, GST scrutiny, ITR review, and appellate submissions reach out for a consultation before deadlines close.

 

© CA Dhiraj Ostwal & Co. All rights reserved. This article is for informational purposes only and does not constitute legal or professional advice.