The channel and the algorithm

On the ninth of October 2025, in the Jhunjhunu district of Rajasthan, officers of the Delhi Police arrested a twenty-year-old man named Rahul.

He had been running a social media account styled AGHORI_JI_RAJASTHAN. He was buying paid advertising to place the reels in front of strangers. He was using artificial-intelligence tools to generate images and video of ghostly figures and ritual scenes to furnish the account. And the case names more than fifty alleged victims across the country. The complaint that opened it came from a woman in Delhi who is reported to have been defrauded of ₹1.14 lakh; the arresting unit was the Cyber Police Station, New Delhi; the arrest was made a thousand kilometres from the police station that made it.

Note the geography. A Delhi cyber unit, a complainant in Delhi, an accused in a village district of Rajasthan, and victims in states neither of them lives in. The jurisdiction of the offence and the location of the offender have come apart, which is itself a fact about how this trade now works.

That is one reported case, and it will be treated as one reported case. It is not an industry, it is not a statistic, and nothing in the reporting connects it to the Śābara corpus by name. It appears in this chapter for a single reason: it is the only piece of documentation this research found in which the machinery of the modern trade is described by an institution that had to put its name to the description.

The case is reported by ThePrint and, independently, by the Deccan Herald, both on 13 October 2025. Two outlets, reporting separately, on the same day. What does not go away is that neither source connects the case to the Śābara corpus by name. No sentence here should be read as though it did.

Everything else this research found was being sold.

What is on sale, and to whom

Put the words into a search box, in English or in Hindī, and this is the shop that opens.

Uploads of śābara mantra collections on the Internet Archive. A document on Scribd. Facebook groups and Facebook pages posting formulae to whoever follows them. An ebook storefront offering a śābara mantra title on a subscription trial, so that the purchase becomes a monthly one. Astrology-service sites, priced, and built to rank.

That is the trade in its present form, and it is worth naming what every item on that list has in common. Each of them is a seller. Not one is a court, an archive, a journalist or a scholar. Under the source tiers this treatise runs on, the entire visible field is unusable as evidence, and there was nothing standing behind it.

Now the second question, which is the harder one. Who is buying.

The trade does not say, and this research cannot produce a customer list. What it can produce is the shape of one transaction, and the shape is in the Rajasthan case. The reels were placed in front of strangers with advertising money, which means somebody chose the audience rather than waiting for one. The complainant was a woman in Delhi, and ₹1.14 lakh is a sum a person parts with under pressure. This chapter has no lecture for her.

A pamphlet had to be walked past. This has to be switched off.

So the obvious next move was to find out how much of this had been written down by somebody who was not selling it.

The search that found nothing

The plan for this chapter was ordinary. Chapter twenty assembles a harm ledger from the witch-hunting literature, which is documented at length and in detail, and the assumption going in was that a parallel ledger existed for the corpus itself. Somewhere there would be a fraud case naming Śābara mantras. A consumer court. An arrest. A journalist’s investigation. A report by an organisation that works on this.

So a sweep was run, on one stated day, with the strings kept. A search result whose strings, index and date go unrecorded is not a result. It is an impression with a footnote. This one was written down as it ran, and it is printed here in full so that anybody can repeat it.

Method. Six strings, three in English and three in Hindī, put to a general web search index on 18 August 2026. Verbatim, in the order run:

1. "Shabar mantra" fraud arrest FIR police case
2. "शाबर मंत्र" ठगी गिरफ्तार मुकदमा
3. "Shabar mantra" court judgment consumer complaint cheating
4. "शाबर मंत्र" पीड़ित पुलिस शिकायत
5. "Shabar mantra" NCRB OR "National Crime Records Bureau" OR NGO report victim harm
6. "शाबर मंत्र" के नाम पर ठगी बाबा गिरफ्तार समाचार

Result. Across all six, nothing of the kind this project counts as evidence. No reported judgment. No piece of journalism naming the corpus. No NGO documentation. Nothing in the NCRB series.

What came back instead was the shop this chapter opened on, and it came back in both languages: the Internet Archive uploads, four of them on the first page of one Hindī string alone, the Scribd document, the Facebook groups and devotional pages posting formulae, the ebook storefront on its subscription trial, the priced astrology services built to rank. Under the source tiers this treatise runs on, every one of those is unusable, and there was nothing else.

Two of the six strings are worth reading more closely, because of what separates them.

The English strings pulled in fraud journalism that has nothing to do with this subject at all: stock-trading arrests, digital-arrest scams, a seer picked up in Odisha. The search engine was matching on fraud and arrest and ignoring the phrase. The Hindī strings did better, and what they returned is the interesting part. String 4 brought back real crime reporting from real outlets. ETV Bharat on a man arrested in Chhattisgarh for defrauding people of ₹52 lakh on the pretext of tantra-mantra. Amar Ujala, dated 31 July 2026, on a case registered in Madhya Pradesh over an assault on a nurse committed under cover of occult ritual. ETV Bharat again on a man arrested in Himachal Pradesh for rape after a woman approached him for ritual help.

That reporting exists. It is dated, it is attributed, it names police stations and sums of money. And in none of it does the word śābara appear.

So the finding can be stated exactly, and it is reproducible from the page.

The harm reporting in this domain attaches to tantra-mantra as a general category, and never to this corpus by name. What attaches to this corpus by name is a shop front.

That is what the sweep of 18 August 2026 establishes, and its limits are ordinary and worth stating: six strings are six strings, one index is one index, and a court record that has never been digitised will not appear in any of them. The result is a statement about the searchable web on a stated date, and it should be cited as that and nothing larger. What would extend it: the same six strings in an Indian legal database, and in the vernacular press archives that do not surface in general search.

Chapter nine met the same wall from the doctrinal direction and chapter twenty meets it from the direction of harm. Neither owns it. The ledger borrows the one line it needs and leaves the anatomy alone.

A chapter that had gone looking for scandal and found some would have told you something about one fraudster. A chapter that went looking and found only shop fronts tells you what the present condition of this material actually is.

The banned library

The same pattern held everywhere in this research, and this is the point at which it stops being an anecdote and becomes an argument.

Chapter nine tested the popular claim that Śābara mantras are exempt from the kīlana doctrine, the idea that mantras are “locked” and require unlocking. The claim is everywhere in the popular literature. The search for a source ran in English, in IAST transliteration and in Devanāgarī, and every single hit was unusable: Gumroad products, Scribd uploads, Patreon posts, blogspot pages, astrology services, mantra-selling sites, and one Hindī devotional magazine from 1940. The doctrinal claim that half the internet repeats about this corpus has no source that this project is permitted to cite, because it has no source that anyone can check.

Under the source tiers this treatise runs on, the entire discoverable discussion of Śābara mantras is banned as evidence.

Read that as a description of the field rather than as a complaint about it. The material is not rare. It is superabundant. Type the phrase into any search box and you will be given more text than you can read in a week. Almost none of it carries a date, an author who can be identified, an edition it worked from, or a claim that could in principle be shown to be wrong.

So the meta-finding of this book gets said here, once, in the chapter that earned it. The trouble is not that the popular material on Śābara mantras is unreliable, which everybody already suspects. The trouble is that the reliable material does not exist, and that its absence is systematic. Two pages in a 1981 handbook. A footnote in a 1955 ethnography. A clause in a poem of about 1574. A trade catalogue nobody had cited for this purpose. Beyond that, a commercial literature that no method available to a careful reader can convert into evidence at all. A subject with millions of readers has a bibliography that fits on a single page.

And here is the generalisation, which holds well beyond this corpus and is the reason this treatise exists at all.

The questions with the most popular demand have the least accessible scholarship.

Millions of people want to know what Śābara mantras are. The demand is enormous, sustained and entirely unmet by anything a careful reader could rely on, because everybody writing about the subject is either selling the rites or debunking them, and both postures produce the same defect. The seller cannot afford a finding that would spoil the product. The debunker has no reason to establish a text, a date or a manuscript, since the conclusion does not depend on them. Between the two of them, the ordinary scholarly work has simply never been done. Goudriaan and Gupta gave the corpus two pages in 1981. That remains close to the whole of what a reader can stand on.

That is the commercial case for this book, and it is also the intellectual one, and this is the chapter where it gets said out loud.

What the pamphlet already was

Now the part where the argument has to resist an easy sentence.

The easy sentence is that the internet degraded something. That a living transmission was corrupted when it met the platform economy. It is a satisfying sentence, and this research cannot support it, because the commerce was there before the platforms and the previous chapter is full of it.

The trade catalogue of the Lakṣmīveṅkaṭeśvar Press of Kalyan and Bombay, examined in chapter seventeen, is a sales document. It is organised by section, it carries prices, and its tantra-mantra titles sit together in one numbered run. The Aghorītantra is advertised as saral bhāṣā mẽ, in simple Hindī, which is a selling point aimed at a reader who cannot manage Sanskrit. The Bṛhad Indrajāl is offered with its 138 yantras: a count of plates, and this chapter records it as a count and nothing more. The Uḍḍīśatantra comes with a Hindī ṭīkā, a commentary, because a bare text would not move. And these titles are shelved in the same catalogue as stage conjuring manuals and instructions for making fireworks.

That is not a tradition later spoiled by the market. That is a market, in print, with a price list, in the era of the steam presses.

The catalogue itself carries no date. Chapter seventeen examined it and refused to assign one, because the copy consulted shows no year on the pages examined, and this chapter inherits the refusal. What the catalogue proves is the trade. What dates the press is a different document, Goudriaan’s note on the printed Bṛhat-śābaratantra editions, which records the same Lakṣmīveṅkaṭeśvar house issuing one in Śaka 1841, 1918 or 1919 of the Common Era.

Goudriaan and Gupta, cataloguing the Uḍḍīśa editions in 1981, close the note by remarking that the market was then full of low-priced printings of the same sort. Their handbook is in copyright and nothing from it is reproduced anywhere in this treatise; that sentence is a summary in this book’s own words, and a reader who wants the compilers’ own sentences will find the note listed in the sources. Projit Bihari Mukharji’s chapter on the subject is subtitled “Occult Forces in the Bazaar Economy,” and the printed grimoire in that account is a commercial genre from the start. Mukharji is cited here by argument rather than by page, because the chapter text was never retrieved, and nothing is put in his mouth in a form that could be read as a quotation. What would close it: the chapter itself, in Rethinking Markets in Modern India.

There is no pure prior state that the internet corrupted.

Anyone writing about this material owes the reader that sentence before they write anything else about the present, because almost every popular treatment of the digital trade is built on its opposite.

The catalogue and the targeting system

Something has changed, though, and it can be named precisely. It is a change in the direction of the transaction.

A catalogue is a bibliography. The Kalyan and Bombay list sits in a shop, or arrives by post, and it waits. It exercises no judgment about who reads it. A man who wants a book about yantras has to already want it, has to find the shop, has to have the price, and has to ask. Every step of that sequence begins with him.

A recommendation engine runs the other way. It does not wait to be consulted. It builds a model of who is likely to respond, and then it goes and finds them, and the signals it responds to are the signals of a life going wrong: the searches people make about a marriage, an illness, a court date, a child who will not speak to them. Paid placement puts the offer in front of that person. Synthetic imagery furnishes it with an authority nobody had to earn or sit for. Subscription tiers convert a one-time purchase into a relationship with a monthly price.

The pamphlet was a product on a shelf. The channel is a system that identifies distress and arrives with a solution.

That is a real difference and it is worth stating carefully, because it is a claim about mechanism and not a claim about authenticity. The printed pamphlet was not more genuine than the video. It was in some ways less careful: an uncritical popular compilation, printed cheap, edited for salability. The difference is in who initiates, and in what the initiating party knows about the person on the other end.

And the person on the other end gets no lecture in this chapter. Someone who pays for a rite because a marriage is failing or a child is ill is doing what people under pressure have always done, and the harm ledger in chapter twenty shows where the injury in this whole world actually falls: on the accused, and through the ojhā’s role as the man who names a witch. The purchaser is not the problem, and moralising at them would be both unkind and, on the evidence, wrong.

There is no longer an edition

Here is the loss, and it is narrower and more concrete than the one usually claimed.

Chapter seventeen was able to cite a book. A specific book, with a publisher, a press, a city, and in the case of the Kāmaratna, a named ṭīkākāra, the commentator Jvālāprasād Miśra. That is not much, and it is enormously more than nothing. It fixes a text at a moment. Two copies can be compared. A later scholar can go back to the same object and find it unchanged, and disagree, and be checked in turn.

The current transmission produces no citable object at all.

A video is re-uploaded and re-edited. A post is deleted when the account is suspended. A PDF circulates with no imprint, no date and no indication of what it was copied from, and the next copy has been silently altered, and there is no way to tell which came first. Whatever else is happening in this economy, it is not producing editions.

That is a bibliographical loss. It is not a spiritual one, and the distinction matters. Nothing here says that anybody’s practice has been diminished, or that a transmission has been broken, or that the material has lost some quality it once had. What is gone is the fixed object, and the fixed object is the thing scholarship needs.

What is being lost

A book on a vernacular occult corpus is expected to carry a passage on the dying tradition. There is none here, and the reason is worth stating in the open.

The claim has sentiment and no evidence.

The research found no documentation of decline. None. No survey showing practitioners diminishing, no study of a lineage failing to transmit, no ethnography of an abandoned practice. What it found was a large, visible and evidently growing commercial presence, which is the opposite of dying. Whatever is happening to this corpus, disappearance is not it, and a chapter mourning its death would have been a chapter of invented feeling.

What can be said without romance is smaller and firmer.

The corpus has never had a critical edition. Not one. There is no apparatus, no stemma, no collation of the printed versions against each other, let alone against manuscripts. The frequency work in chapters ten to twelve, which is the most concrete evidence in this treatise about how the formulae actually behave, had to be done on a single uncritical popular compilation, and every claim built on it carries that caveat permanently attached. The syntax of the formulae, the regional distribution of the corpus, its relation to oral practice, the question of whether the printed texts descend from anything older than the presses that printed them: all of these are answerable questions, and all of them are currently unanswerable, for want of a text.

What is being lost is not belief. It is the possibility of scholarship. The loss is bibliographical, and calling it anything grander would be the invented feeling again.

Every year the printed pamphlets decay a little further, and the digital versions proliferate a little faster, and the ratio between fixed objects and unfixable copies moves in one direction. The corpus becomes harder to edit, not easier. That is a slow loss, it is undramatic, and it is the only decline this research can actually document.

One limit sits on all of it. This is Hindī-belt evidence: north Indian print, north Indian search results, a Hindī compilation, and one case running between Delhi and Rajasthan. The Ādinātha charter in chapter three claims the corpus for Kerala, Karṇāṭa and Andhra as well, and nothing in this chapter tests that claim. A Tamil or Malayalam or Telugu picture of the same trade might look quite different, and this treatise does not have it. If the charter is describing a real distribution rather than an aspiration, then everything above is a regional finding wearing national clothes.

What has not been read

Two doors were found and left shut, and they are named here so that whoever opens them knows the argument they are joining.

The first is the International Journal of Hindu Studies special issue of 2022 on Digital Tantra, and within it Hugh B. Urban’s article “Dark Webs: Tantra, Black Magic, and Cyberspace.” That is a scholarly literature on exactly this chapter’s subject, and it was not opened. It may well contain the scholarship this chapter says does not exist, and anything here that it contradicts should be corrected to it.

The second is not a door left shut. It is a door this book decided not to open, and the distinction is worth drawing clearly, because a reader who reaches the end of this chapter may feel a specimen has been withheld and wonder whether the withholding is a failure of research.

It is a position. Nothing about how an amulet or any other object of this trade is made, carried, filled or used appears in this book, and nothing will. What chapter seventeen exhibits is the form of the genre: catalogue copy, entry numbers, a count of plates, the promise of saral bhāṣā. What chapters ten to twelve exhibit is its shape in the aggregate, by frequency count. What chapter two exhibits is what the lexicographers made of the word. Function is withheld throughout, and chapter twenty sets out why. A study of the trade does not have to become a supplier to the trade in order to describe it.

The cost of that decision is real and it is named here. Material culture is the class a comprehensive study of this economy would begin from, and this treatise cannot claim to have covered it. With that angle goes Christopher Pinney’s Photos of the Gods of 2004, on the industrial production of devotional images in India, which was not opened. It is the closest documented analogue to what this chapter has described: sacred imagery made at scale by commercial presses for a mass market. A scholar who takes up the object rather than the text will find the foundation there, and will have to make the same decision about what to print.

The next chapter leaves the market entirely and goes to the village, where the same power is held by two kinds of person and called by two different names: the healer and the witch.

The catalogue waited to be read.

The channel goes looking.