Over the last few years, I have seen a quiet but significant shift in income-tax assessments. Additions are increasingly being triggered not by field enquiries or seized material, but by algorithms—AIS flags, risk parameters, data analytics, and system-generated alerts.
Technology has undoubtedly improved detection. But at the appellate stage, a fundamental question keeps arising:
Can algorithm substitute evidence? The consistent answer from the ITAT has been: No.
In this article, let us review and explore the various nuances of this issue
1. The New Faceless Assessment Reality
Many assessment orders today are built on:
- AIS/TIS mismatches
- high-risk transaction flags
- abnormal patterns identified by systems
- third-party data auto-populated in the portal
Often, the reasoning stops there.
The addition follows with phrases like:
- “system-generated information indicates…”
- “as per data analytics…”
- “algorithm-based risk assessment shows…”
What is missing is what matters most in law—proof of income.
2. ITAT’s Core Position: Data Is Not Evidence
The Income Tax Appellate Tribunal has been remarkably consistent on one principle:
Technology may trigger an enquiry, but it cannot conclude an assessment.
Algorithmic outputs are, at best:
- indicators,
- risk markers, or
- starting points for verification.
They do not, by themselves, establish:
- nature of receipt,
- taxability, or
- concealment of income.
Unless the AO converts data into legally admissible evidence through enquiry and findings, the addition becomes vulnerable.
3. Why Algorithm-Driven Additions Fail at ITAT
a) Absence of Independent Verification
Tribunal benches repeatedly note that:
- AIS data is not sacrosanct,
- system information can be incomplete or duplicated,
- third-party reporting errors are common.
When additions are made without independent verification, ITAT treats them as unsustainable.
Data must be tested against facts—not accepted as such on faith.
b) No Finding on “Income Character”
Algorithms detect transactions, not income or their character.
A credit flagged by the system may be:
- capital receipt,
- loan,
- inter-account transfer,
- already taxed business turnover, or
- mere pass-through entry.
One can see each of these have different connotations for determining income character.
ITAT consistently holds that unless the AO demonstrates why a receipt has income character, taxation cannot follow.
c) Natural Justice Cannot Be Automated Away
A recurring problem in faceless assessments is:
- limited opportunity to explain data anomalies,
- generic responses by the system,
- rejection of explanations without reasoning.
The Tribunal has made it clear that: efficiency cannot override fairness.
An assessee must know:
- what exact data is relied upon,
- how it is adverse, and
- why the explanation is unacceptable.
Algorithmic opacity does not meet this standard.
4. Burden of Proof Still Lies Where the Law Places It
A dangerous assumption has crept in: “The system has flagged it, so the assessee must disprove it.”
ITAT has firmly rejected this approach.
The law has not changed:
- initial onus may be on the assessee to explain,
- but the ultimate burden to prove income remains on the Revenue.
Algorithms do not shift statutory burden from the Revenue to the Assessee.
5. CIT(A) Silence, ITAT Correction
In many cases, CIT(A) orders merely reproduce:
- assessment reasoning, or
- system-generated findings,
without examining whether:
- evidence exists,
- enquiry was conducted, or
- legal requirements were met.
ITAT has not hesitated to step in where technology is mistaken for testimony.
6. Practical Tips for Appeals
In ITAT matters involving algorithm-driven additions, focus on:
- Identifying AO’s reliance solely on AIS / system data
- Demonstrating absence of enquiry or corroboration
- Highlighting lack of findings on income nature
- Pointing out denial of effective opportunity
- Emphasizing that data ≠ evidence/income
In many cases, this alone dismantles the addition.
Closing Remarks:
Data and Analytics are powerful and Algorithms are useful.
But taxation is still a legal exercise, not a computational one.
The Tribunal’s message is clear:
Technology may guide the tax administration—but evidence, reasoning, and fairness are touch stone and critical for deciding the case.
For taxpayers facing algorithm-driven demands, ITAT continues to be the forum where law prevails over logic trees and risk scores.
*****
In case you have any concern and queries or need any support regarding taxation compliance and litigation, you may like to contact us.
Abhinarayan Mishra, FCA, FCS; Managing Partner, SAM Law Associates LLP; KPAM & Associates, Chartered Accountants, Dwarka, New Delhi; +9910744992, ca.abhimishra@gmail.com
Author Bio
