Rigour has not become less possible. It has become less visible.

1. The presumption of fabrication

Every citation in Indian scholarship now carries a presumption against it. The reader assumes, until shown otherwise, that the authority was produced by a machine and not read by a person. The presumption did not arise from suspicion of the author. It arose from the record.

The record is short and recent. In June 2023 a federal court in New York sanctioned two lawyers for filing a brief built on judgments that did not exist, generated by a chatbot and never opened.¹ A public database of judicial decisions dealing with fabricated or misrepresented authority has since grown from that one case to 2,207 decisions across more than forty jurisdictions; in thirty-two of them the person relying on the machine was the judge.² India entered the record in December 2024, when a Bengaluru bench of the Income Tax Appellate Tribunal decided a ₹669 crore trust matter on four authorities that could not be found, and recalled its own order eight days later.³ In July 2026 the Supreme Court set aside an insolvency admission because the tribunal had rested it on six authorities, each of which was non-existent, or a wrong citation of a real judgment carrying a paragraph that was never written, or a correct citation carrying such a paragraph; the Court sorted them into exactly those three classes and declared zero tolerance for the Bar and the Bench alike.⁴ In September 2026 it set aside a customs penalty of ₹425 crore because the adjudicating officer's order relied on cases that did not exist, citations that were wrong, and real judgments cited for propositions they did not decide.⁵ In each Indian instance the fabrication sat not in a lawyer's brief but in the decision itself.

The lesson the system drew from this record was reasonable against the tool and ruinous against the author. A presumption that attaches to a machine's output has been transferred, wholesale, to everyone who cites. It falls on the scholar who opened every source as heavily as on the one who opened none. Both now carry the burden of proving a negative, and neither has been given the means to discharge it.

2. How the system failed

The system answered the record with three instruments. Each was sound in its purpose and each, in its effect, has stripped the honest author of any way to be recognised as honest.

Verification was defined downward. The first instrument was the instruction to verify. But verify what? The Supreme Court's three classes of bad authority ask a single question: does it exist? That question is the right first one and the wrong last one. The customs order set aside in September 2026 had already passed it: several of the judgments it relied on were real, and were cited for propositions they did not decide.⁵ The database is clear that this is the shape of the next failure: of its 2,207 decisions, 937 involve genuine authority misrepresented, not invented.² The leading empirical study of commercial legal research tools found that they produce false answers in roughly one query in six to one in three, and counted among those failures any false assertion that a source supports a proposition; one vendor's answer was that every linked citation is real.⁶ That is a promise that the case exists, not that it says what the author needs it to say. Some courts have written the harder rule. The United Kingdom's Upper Tribunal, since November 2025, requires a signed statement that each authority exists, can be located from the citation given, and supports the proposition for which it is cited.⁷ No Indian rule yet states that third limb. Existence has become the ceiling of verification when it should have been its floor.

Detection cannot rebut the presumption. The second instrument was software that reads prose and pronounces on its provenance. The best-known study of such detectors ran ninety-one essays written by non-native speakers of English for a language-proficiency test through seven widely used tools; on average three in five were marked as machine-written, while essays by American schoolchildren were almost all passed as human.⁸ The detectors reward unpredictable word choice, and so penalise what the authors call constrained linguistic expression. Most Indian scholarship is written in English as a second or third language. The presumption therefore falls hardest exactly where the work is hardest, and no improvement in the software cures the defect, because style was never evidence of authorship. An honest author cannot prove innocence by prose.

Imported standards make Indian citations look like errors. The third instrument was older than the other two: the foreign citation manual, adopted by Indian journals and law schools because none of our own existed. Those manuals were written for the sources of other jurisdictions, and treat an Indian judgment, statute, Gazette notification, tribunal order, Law Commission report or vernacular text as foreign material to be cited by analogy. Indian authors have therefore improvised for decades, and improvisation has a cost it never had before. To a reviewer primed for fabrication, a citation in an unfamiliar form, to a reporter the reviewer cannot open, is a citation presumed invented. An unverifiable citation and a fabricated one now look the same on the page, and are treated the same.

This is what it means to say the system has failed. Rigour has not become harder to practise. It has become impossible to show. The author who checked every proposition and the author who checked none present identical pages, and the reader, with reason, trusts neither. The claim to rigour has been robbed of its legitimacy, not because rigour declined, but because the means of recognising it were never built.

3. What a standard restores

A citation standard is not a formatting convention. It is a protocol of accountability. A citation is the author's warrant that a source exists, that it can be found where the citation says, and that it says what the author says it says. The standard fixes the form in which that warrant is given, so that any reader can test it. Remove the standard and the warrant is still given, but in a form no one can test, which is where Indian scholarship now stands.

This manual cannot stop a machine from inventing a judgment, and does not try. Its object is the presumption, not the machine. It restores the legitimacy of a claim to rigour in four ways.

It makes the Indian source locatable. Every form in the chapters that follow is built for the source as it exists in India: the reporter, the neutral citation, the paragraph, the notification, the order. A citation given under this manual can be opened by a reader in a known place in a known way. The test of rigour becomes an act of a minute rather than a suspicion that never resolves, and a citation that fails the test fails it visibly. That is the difference between a standard and a style.

It cites the proposition, not the existence. The pinpoint is not decoration. It is the author's statement of where, in a source that may run to two hundred paragraphs, the proposition relied on actually lies. An author who gives the paragraph has read the paragraph, and a reader who opens it will know within a minute whether it says what is claimed. This is the third limb of the Upper Tribunal's statement of truth,⁷ written into the citation itself rather than into a certificate filed beside it. The failure that passed every existence check in September 2026 would not have passed this one.⁵

It makes the author answerable by declaration. Chapter 13 requires the author to declare what assistance was used, for what, and when, and to accept responsibility for the accuracy, originality and consequences of everything so produced.⁹ A presumption cannot be rebutted; a declaration can be held against its signatory. The declaration therefore does what detection software cannot: it moves the question from how the prose reads to what the author has undertaken, and it gives an honest author a signed instrument with which to be believed.

It is ours. The forms in this manual were written for Indian sources by someone who has had to cite them, not adapted from a manual written for another country's courts. A claim to rigour made in a borrowed grammar is always made at a discount, because the reader must first decide how much of the strangeness is the standard and how much is the author. A claim made in a standard built for the material carries no such discount. The presumption of fabrication is, at bottom, a presumption that the author could not have done the work. A standard that lets the work be seen is the only answer to it.

That the manual's author says so is no argument. The record set out above is.

4. What this chapter asks of the author

The rules that follow are the mechanism. This chapter asks for the discipline behind them, in three undertakings.

1. Cite to the proposition. A source is cited for what it decides or says, at the paragraph or page where it decides or says it. A citation whose pinpoint the author has not read is not a citation.
2. Treat a source you cannot open as a source you cannot cite. If the reporter, the order or the text is not in front of you, it is not in your footnote, whatever a tool reports about it.
3. Declare, and own what you sign. Assistance is not misconduct; concealment is. The declaration required by Chapter 13 is the author's instrument of credibility, and it is to be made exactly and dated.⁹

Rigour has not become less possible. It has become less visible. This manual exists to make it visible again.