Arrival at this page follows, in the great majority of cases, from a single circumstance: a link that once led to a business listing has been deliberately supplanted by the address of the present article, the listing that formerly occupied that position having been removed or suspended. What follows sets out, without euphemism, the grounds upon which such a removal is undertaken, the manner in which it is verified before being acted upon, and the courses of redress available to the proprietor of the affected listing.

The substitution of an explanatory page for the removed address, rather than the silent deletion of the record, is itself the first and smallest expression of a principle that governs all that follows. A deletion conducted in silence would merely generate a further dead link for whoever had bookmarked, cited, or crawled the original address; a live and explanatory page keeps the reference resolvable while disclosing, plainly, what has changed. A curated directory falls to be judged not by the number of rows held within its database, but by whether each continues to point toward something real.

Jasmine Directory has operated since 2009 as a human-curated catalogue rather than an automated harvest of the web; roughly nine listings in ten were entered by hand, through editorial review, and at no point has paid acquisition been used to inflate its size. Therein lies the reason for the severity with which it audits itself: a directory assembled slowly, through judgement, forfeits its value at the moment its outbound links cease to resolve, and the scholarly record on web decay is unequivocal that they will so cease unless their failure is actively forestalled. The policy described here is therefore no punitive instrument directed at site owners, but a discipline of maintenance directed at the integrity of the directory itself.

Two clarifications belong at the outset. The first is that the replacement of a listing by this page is, in the overwhelming majority of instances, reversible, the closing sections setting out how. The second is that no removal is ever unconsidered: detection is assisted by software, yet no listing is removed upon a single machine observation, every removal passing through verification and, where the failure concerns content rather than mere availability, through human editorial judgement. The reader convinced that the site remains healthy is invited to consult the workflow below before concluding that an error has occurred, and thereafter, the conviction persisting, to follow the contact route at the close; a sound listing is far rather re-examined than lost.

Jasmine Directory graphic headed ‘Human-Edited, Every Day Since 2009’
A human-curated directory, edited by hand every day since 2009.

Why links perish, and why a directory must attend to it

The phenomenon bears a settled name within information science: link rot, the progressive failure of formerly valid addresses, studied together with content drift, the silent mutation of a page away from what was originally referenced. Their conjunction is what Klein and colleagues (2014) designated reference rot; their examination of more than a million references established that one scholarly article in five already suffered from it, the proportion rising to roughly seven in ten among articles citing web resources. Spinellis (2003) derived for web references a half-life on the order of four years. Zittrain, Albert and Lessig (2014) found upward of seventy per cent of the addresses within a leading law review, and about half of those within opinions of the United States Supreme Court, no longer leading to the material once cited; SalahEldeen and Nelson (2012) measured the loss of resources shared upon social media at roughly eleven per cent within the first year.

The precise figures vary with corpus and method; their direction never does. A body of links is a perishable asset, and a directory is, in its very structure, nothing other than a large and ordered body of links. The literature documenting the decay of citations within journals and judicial opinions describes the identical hazard a directory confronts, distinguished only by greater scale and the commercial consequence attaching to it. Were nothing done, the directory would decay at a rate consonant with these studies, and the visitor following its links would meet with errors, hijacked domains, and pages no longer answering to what had been vouched for.

What a measured half-life signifies at the scale of a directory

At the scale of a directory the abstraction of a half-life resolves into something concrete. If a population of references forfeits roughly half its members across some four years, a catalogue assembled over many years and never pruned does not decay gently but compounds: the references entered longest ago have had the most time in which to fail, and since most large directories accumulated the bulk of their entries within a concentrated early period, the largest cohort is also the oldest and most decayed. The findings of Dimitrova and Bugeja (2007), and the URL corrosion reported by Prithviraj and Sampath Kumar (2014), incline the same way. A removal policy is the sole mechanism by which this compounding is interrupted; absent one, the effective size of a directory declines year upon year even as its nominal size holds level. Herein lies the reason the periodicity of the audits weighs as heavily as their existence: fresh failures arise continually, and only repeated inspection, acted upon, preserves the catalogue from sliding.

Jasmine Directory graphic headed ‘Dead Links Removed Every Single Week’
Dead links are reviewed and removed on a weekly basis.

The grounds upon which a listing is removed

Set out below is the working catalogue of conditions under which the address of a listing is suspended and replaced by the present page. The first six are the grounds most frequently invoked; those that follow are additional, and no less real, conditions which the audits surface. Not one constitutes a judgement upon the merit of the business standing behind the listing: they are statements concerning the link and its destination, not the company.

1. A dropped or expired domain

Where a domain is suffered to lapse, is advertised for sale, or rests upon a holding page, the listing no longer leads to the business that was reviewed. Of all the failures the dropped domain is the most consequential, for expired domains are routinely re-registered by third parties and repurposed — for spam, for affiliate redirection, for link schemes — and to persist in vouching for such an address would be to lend the editorial credit of the directory to whoever next assumes control of the name. Listings upon dropped domains are accordingly removed without awaiting the abuse experience teaches is to come.

2. A persistent 404 (Not Found)

The enduring 404 is the dead link in its textbook form. It degrades the experience of browsing, frustrates the visitor who reposed trust in the catalogue, and undermines the referential value of the directory entire. Search engines themselves counsel the resolution of such errors, and a curated directory cannot hold itself to a standard lower than the platforms alongside which it hopes to be trusted. A listing resolving to a 404 across repeated examination is removed.

3. A 301 redirect toward a different root domain

A permanent (301) redirect dispatching the visitor toward a different root domain than the one reviewed sunders the identity between listing and destination. The review fee, the editorial assessment, and the categorisation attach, each, to one specified domain; a redirect substitutes another in silence, and this is the classical vector by which directory equity is resold. Addresses are removed, on the like reasoning, where a nominally temporary (302) redirect has been left standing in excess of a year, a redirect that never reverts being permanent in all but its status code.

4. Content altered so as to violate the terms and conditions

This is content drift bearing a dimension of compliance. A site reviewed as, for instance, a legitimate dental practice may afterward give way to gambling, to adult material, to counterfeit goods, or to thinly veiled link-selling. The address yet resolves, and may even return a healthy 200, so that an automated checker alone would not flag it; editorial review alone apprehends it. Where the content of a destination has mutated into what contravenes the published terms, the listing is removed irrespective of its technical health.

5. A failed recurring payment upon a standard (annual) listing

Standard listings are recurring, billed annually. Where the renewal fails — through a lapsed card, a cancelled subscription, an expired instrument of billing — and goes unremedied, the listing is removed automatically at the term of its paid period. The ground is administrative rather than editorial, and the most readily reversed of all, as the route of resubmission below makes clear.

6. A 403 (Forbidden) confirmed from every continent

A 403 informs the crawler that retrieval is forbidden. Since a solitary 403 may signify a transient block, a rate limit, or a geographic restriction rather than a genuine outage, no action proceeds from one observation: the response is confirmed from multiple vantage points across continents before it is treated as real, and only a 403 reproducing globally — the resource being inaccessible to the ordinary visitor, and not merely to one range of addresses — results in removal. This verification from many vantages is what separates a defensible decision from a false positive.

Further grounds surfaced by the audits

Beyond the six principal grounds, further conditions occasion removal, each a documented mode of failure and each verified before action: the expiry or misconfiguration of an SSL/TLS certificate; the flagging of a destination by reputable safe-browsing services, or its service of malware or phishing, in which case suspension is immediate; cloaked or malicious redirection, whereby crawler and human visitor are shown divergent destinations; the soft 404, alive in its status code yet dead in substance; the persistence of server errors across an extended interval; the drift of content into thin, scraped, or machine-generated filler; the manifest change of ownership by which a domain comes to serve a foreign purpose; and geographic restriction so comprehensive as to exclude the ordinary visitor altogether.

Across the whole of these, the unifying test is simple: does this address still resolve, securely and globally, to the legitimate business that was reviewed? Where the honest answer is no, the listing is removed, and a removal page set in its place, that the reference itself may not become one further dead link upon the web's accumulating heap of them.

The dead link and the drifted link are distinct problems

It serves clarity to part two modes of failure. The first is link rot in the narrow sense: the address returns the resource no longer — a 404, a dropped domain, a forbidden response reproducing everywhere — the visible failure, which an automated checker apprehends reliably. The second is content drift: the address yet resolves, returns a healthy 200, and shows something materially other than what was reviewed. Klein and colleagues (2014) gathered the two beneath the single heading of reference rot because the second is the more insidious, passing every automated test of health while it dismantles the meaning of the reference. Here lies the structural reason the process cannot be wholly automated: a machine confirms that a page loads; editorial judgement alone confirms that the page which loads remains the business vouched for.

The listing-removal workflow at Jasmine Directory A seven-stage vertical process: continuous monitoring, automated detection, multi-continent verification, secondary validation, editorial review, decision and URL replacement, and resolution path. How a listing is identified, verified, and removed 1 Continuous monitoring The catalogue is audited several times in the year. Outbound links are not presumed stable but treated as a perishable asset, consonant with the literature of decay. 2 Automated detection Crawl-based audit tools flag candidate failures: 404s, redirects, 403s, server errors, certificate faults, and suspicious patterns of status are surfaced at scale. 3 Verification from many vantages Each candidate is re-examined from several points across continents. A block is acted upon only where it reproduces globally, never from a single range of addresses. 4 Secondary validation A second, independent pass with differing tooling reduces false positives. Transient outages are allowed the interval in which to recover before any listing is touched. 5 Editorial review A reviewer apprehends what no automated tool can: content drift, violations of terms, parked pages, and soft 404s that return, deceptively, a healthy status. 6 Decision and replacement Confirmed failures are removed. In place of deletion, the address is replaced by this explanatory page, that the reference may resolve rather than rot. 7 The route of redress The proprietor may resubmit (standard plan) or make contact with the same domain repaired. One review fee covers one domain, so a corrected site re-enters without a second fee. Verification precedes removal at every stage; nothing is removed upon a single observation.
The workflow of removal. Detection is automated; verification is conducted from many vantages, and the decision itself is editorial, so that a transient block or a one-region outage is never mistaken for a dead destination.
Jasmine Directory graphic headed ‘Verifiable DUNS, TIN and EU VAT’
Verifiable corporate identifiers: D-U-N-S, TIN, and EU VAT.

How a decision to remove is reached

Detection and decision are therefore parted by design. The automated audit tools excel at the flagging of candidates at scale, and that is the whole of what is asked of them. The candidate list passes thereafter through verification from multiple geographic vantages, a secondary pass with independent tooling, and, where the failure concerns content rather than status, human editorial review; only a failure surviving the whole of that is treated as real. The care of the method mirrors the decay studies themselves, whose authors distinguished the transient from the persistent because the credibility of their findings depended upon it.

Two species of error, each worth averting

A process of removal may fail in two contrary directions, and one well designed weighs both. A false positive removes a listing that was in truth healthy — a site returning a transient error within the window of audit, or blocking one crawler while serving all else — and the cost falls upon the proprietor and upon the directory. A false negative retains a listing that has in truth failed, and the cost falls upon the visitor, dispatched toward an error or a deception, and upon the directory's credibility. The aggressive cull minimises the one while it maximises the other. The multi-stage design described here — automated breadth for detection, redundancy of geography and tooling for verification, human judgement for content — endeavours to drive both rates downward at once: the intervals of grace suppress the false positive, the editorial review of drifted content the false negative. The expense is the price of declining to choose between the two mistakes.

Jasmine Directory graphic headed ‘Eight Awards for Editorial Discretion’
Editorial curation recognised by eight industry awards.

The principle beneath the mechanism: the preservation of editorial discretion

All that precedes is mechanism. The reason beneath it is a single editorial principle, and it is owed a plain statement: Jasmine Directory retains a high discretion over that which it lists, and removal is the exercise of that discretion, not a departure from it.

Editorial discretion is the faculty that distinguishes a curated directory from an automated index. The index aspires to completeness, promising little of any address. The curated directory strikes the contrary bargain: it undertakes that inclusion shall mean something — that a human assessed each entry and adjudged it worth a reader's time — and that this judgement is maintained rather than rendered once and abandoned. A listing admitted in good faith may, across the years, become a dropped domain, a hijacked redirect, or a destination drifted into what would never have been admitted; and were the discretion to remove it wanting, the original act of curation would by slow degrees be made a falsehood.

Hence the discretion is described as high, and no apology is made for it. The directory undertakes to keep a listing not because it was once paid for or once legitimate, but only so long as it continues to satisfy the standard upon which it was admitted. Reference rot being no rare accident but a measurable and well-nigh inevitable process — a half-life, in the framing of Spinellis, and not a freak event — a directory that does not actively prune is negligent toward those who consult it, suffering its catalogue to fill with the very dead and drifted references researchers have spent two decades documenting.

The discipline already practised by libraries and by journals

None of this is peculiar to the web directory. A research library does not retain every volume without regard to condition; it deaccessions, withdrawing what has decayed or ceased to belong, as an act of stewardship and not of failure. A scholarly journal issues corrections and retractions where a published item proves unreliable. The legal-citation community's response to reference rot, documented by Zittrain and colleagues (2014), was to build infrastructure for preservation. In each case the institution holds discretion over its corpus and exercises it to defend that corpus's reliability; to remove a decayed listing is the directory's equivalent of deaccessioning a ruined volume. Curation without the power of withdrawal is no curation at all.

Removal is not the failure of curation but its continuation beyond the moment of admission: the judgement that once found a listing worthy returns, in time, to determine when it is so no longer; and in the exercise of that judgement, inclusion retains its meaning for all that remain.

Jasmine Directory graphic headed ‘Editorial Backlinks, Weighed Above the Rest’
Editorially placed links, weighed above automated alternatives.

That dead links do genuine harm, and to three parties at once

The temptation is to regard a dead link as a trivial nuisance soon forgotten. The research, and the directory's own experience, hold otherwise: a dead link inflicts distinct and cumulative injury upon three parties together — the directory that hosts it, the visitor who follows it, and the business that owns it. The diagram below sets the three in relation; the prose thereafter takes each in turn.

The propagation of a single dead link across three parties A single dead link at the top cascades into harms to the directory, then to the visitor, then to the proprietor, with the injury recirculating among them. The propagation of a single dead link A SINGLE DEAD LINK its consequences do not remain singular THE DIRECTORY The corpus that hosts the link Erosion of the trust that curation exists to establish Accrual of reference rot across the entire corpus Obstruction of the crawlers on which indexation depends THOSE WHO CONSULT IT The visitor who follows the link A click expended upon an error or a vanished page Redirection toward parked, forsaken, or hijacked pages Exposure to malware, phishing, or deceptive redirects THE PROPRIETOR The business that owns the listing Forfeiture of the custom a working link would have carried Reputational injury when a name resolves to a derelict page A lapsed domain reconstituted as a rival's or spammer's asset Inconsistent data that weakens its presence across the web The injury recirculates: each party's loss becomes the mechanism of the next.
A dead link is never the private failure of one record. The injury propagates outward and back again, the directory's diminished credibility falling upon the visitor's experience and the proprietor's reach alike, so that the three losses compound rather than merely add.

The injury to the directory

For the directory, dead links are corrosive in a literal sense, eating away at the very asset of which it is composed. Reference rot accumulating over time rather than striking at once, an unattended catalogue degrades by slow and invisible degrees until a large fraction of its links no longer resolve — the condition Klein and colleagues (2014) and Zittrain and colleagues (2014) documented in corpora more carefully tended than the average directory. Each dead link obstructs, besides, the crawlers by which the directory is indexed; and every tolerated failure converts, in silence, the directory's headline magnitude into a fiction, a catalogue of many thousands in which thousands no longer resolve being, in use, a far smaller catalogue than it professes.

The injury to those who consult it

For the visitor, a dead link is a direct breach of the implicit compact a curated directory extends. The visitor did not arrive by chance but followed a link the directory vouched for. When that link returns a 404, alights upon a page advertising the domain for sale, or exposes the visitor to a malicious redirect, the harm is no mere wasted click but a forfeiture of confidence that generalises, research upon user trust finding consistently that a single deceptive experience taints the perceived reliability of the platform entire. Content drift is the subtler instance — the page loading, returning its healthy 200, and showing something other than what was promised, in certain respects the graver failure, for it is not so much as flagged as one.

The injury to the proprietor

The least intuitive injury, and the one site owners most often overlook, is that which a dead link works upon the proprietor himself. A lapsed or broken listing forfeits, first, the referral and the custom a working link would have carried; yet the cost runs deeper. An abandoned domain does not remain neutral — expired domains being re-registered and repurposed, a business that suffers its domain to drop surrenders a ready-made asset, complete with whatever residual authority it bore, to a spammer or a competitor. There subsists, besides, a structural cost upon which the literature of local presence insists: vanished or inconsistent business data weakens an entity's footprint across the web, the consistency of a business's particulars across many sources being itself a signal upon which the systems of search and answer rely. The vexing part is that the business has accomplished the hard part — it exists, it is real, it belongs — and stands to forfeit that benefit to the easy part left undone.

Why the three injuries compound rather than add

The three do not, in reality, repose in separate compartments but feed one upon another. A dead link injures the visitor; the injured visitor forfeits confidence in the directory; the directory's diminished credibility lessens the value of every listing it carries, the well-maintained not excepted; and that lessened value descends again upon site owners as a class. The damage is thus not additive but compounding, the loss of one party becoming the mechanism of another's. Herein lies the deeper reason a policy of removal is an act of collective stewardship rather than a transaction with a single owner: in removing one failed listing, the directory defends the credibility from which every remaining listing draws its worth. Dead links are, within a shared resource, a species of pollution: severally minor, collectively corrosive, best addressed by the party possessed at once of the vantage and the discretion to act.

Jasmine Directory graphic headed ‘Trusted by SEO Professionals’
A directory relied upon by search-engine-optimisation practitioners.

The life-course of a dead link, and why prevention costs less than recovery

A listing seldom dies of a sudden but decays through a recognisable sequence. The first stage is, as a rule, invisible: a domain's renewal approaches while none attends to it; a migration alters a structure of addresses without redirection; an old page quietly disappears upon an update. The second stage is the failure itself — the renewal lapses, the page returns its 404 — and from this moment the listing is dead, though the proprietor, having ceased to think upon it, commonly perceives nothing. The third is detection: the audit apprehends and verifies the failure, and the listing is removed and replaced by the present page. The fourth, in the worst case, is expropriation: the dropped domain is re-registered by a third party, and whatever residual value it bore is now at labour for another.

The instructive feature is where intervention is cheap and where it is dear. At the first stage — the renewal of a domain in good time, the configuring of a redirect across a migration — it costs minutes and forestalls the whole cascade. At the fourth, the domain once lost to a third party, it may cost a great deal or prove impossible, the listing equity built over years admitting of no simple transplant to a replacement domain. The asymmetry is stark: the cheapest action available forestalls the dearest outcome possible, and between them stands nothing but attention.

The preventive measures are few and undemanding: the renewal of the domain well in advance of its expiry, ideally under automatic renewal and with the instrument of billing kept current, the lapse of a domain being the most damaging and least recoverable of failures; the preservation of addresses across a migration, or proper permanent redirection toward the new location upon the same root domain; the maintenance of a valid certificate; the keeping current of the payment instrument attaching to the standard plan; and a reading of the listing itself, once or twice in the year, as a stranger would read it. Each is an intervention of the first stage; none demands technical expertise, and together they neutralise nearly every ground of removal before ever it can arise.

The courses of redress, should a listing have been removed

Should the listing have been replaced by the present page, the situation is, almost invariably, recoverable, the route depending upon which of two cases obtains.

Where the removal followed a lapsed or failed renewal upon the standard, recurring plan, resubmission through the ordinary process suffices: the site is reassessed against the prevailing standard, precisely as a new submission would be, and — provided it continues to satisfy the criteria upon which it was first admitted — restored.

Where, by contrast, the destination was genuinely unavailable, hijacked, or broken, and has since been repaired, the proprietor is invited to make contact with the corrected site, that it may be re-verified; and here a nuance of fee bears stating plainly: a review fee attaches to a single domain. The proprietor returning with the same domain, repaired, does not pay to have a new property reviewed but completes the review of the one already paid for, and no second fee arises. Should the move have been to a different domain, that constitutes a new property requiring a fresh review, for which a new submission is the proper course.

Two further matters merit a word. Where the removal is believed an error — the site having returned a transient failure within the window of examination yet being now in health — contact is invited and the verification re-run; no process is without fault, and a sound listing is far rather re-examined than left out. And where an agency of search-engine optimisation created the listing on a proprietor's behalf, and it cannot afterward be located, contact is likewise invited: the directory abides by the law upon unsolicited mail and does not approach site owners unbidden, and any dispute of billing between an agency and its client subsists between those two alone.

Whichever route obtains, the most durable measure is the one that forestalls recurrence: the keeping of the listing, and of the domain beneath it, in good maintenance. A listing examined once or twice in the year against reality, upon a domain renewed well before its lapse, simply does not fall within any of the grounds rehearsed above. The studies below advance the same case from the side of the data: the listings that perform are the maintained ones, and maintenance is the lighter labour by far.

Jasmine Directory graphic headed ‘Studied in Springer Nature, 2023’
The directory’s methods, examined in a Springer Nature study, 2023.

The studies beneath this policy

The policy here set out is not asserted from authority alone but rests upon original research into the directory's own corpus of 14,362 listings, published in full and with its methods left open to inspection. For the reader who would examine the evidence beneath the principles advanced above, the following are commended:

The thread joining that research to the present policy is plain. The studies established that the worth of a directory resides not in its headline count but in how many of its listings are complete, current, and resolvable; a removal policy is the maintaining arm of that finding, the means by which the sound portion of the catalogue is kept from decaying in silence back toward the skeleton-and-dead-link condition against which the research warns.

Further reading

The principles advanced in the present article rest upon an established body of peer-reviewed inquiry into link rot, content drift, reference rot, and digital preservation. The sources below are provided for verification, rendered in APA style with bare links to the originals.

  1. Klein, M., Van de Sompel, H., Sanderson, R., Shankar, H., Balakireva, L., Zhou, K., & Tobin, R. (2014). Scholarly context not found: One in five articles suffers from reference rot. PLOS ONE, 9(12), e115253. https://doi.org/10.1371/journal.pone.0115253
  2. Spinellis, D. (2003). The decay and failures of web references. Communications of the ACM, 46(1), 71–77. https://doi.org/10.1145/602421.602422
  3. Zittrain, J., Albert, K., & Lessig, L. (2014). Perma: Scoping and addressing the problem of link and reference rot in legal citations. Harvard Law Review Forum, 127, 176–199. https://harvardlawreview.org/forum/vol-127/perma-scoping-and-addressing-the-problem-of-link-and-reference-rot-in-legal-citations/
  4. SalahEldeen, H. M., & Nelson, M. L. (2012). Losing my revolution: How many resources shared on social media have been lost? In Theory and Practice of Digital Libraries (TPDL 2012), Lecture Notes in Computer Science, 7489, 125–137. https://doi.org/10.1007/978-3-642-33290-6_14
  5. Dimitrova, D. V., & Bugeja, M. (2007). The half-life of internet references cited in communication journals. New Media & Society, 9(5), 811–826. https://doi.org/10.1177/1461444807081226
  6. Prithviraj, K. R., & Sampath Kumar, B. T. (2014). Corrosion of URLs: Implications for electronic publishing. IFLA Journal, 40(1), 35–47. https://doi.org/10.1177/0340035214526529

That the present page should resolve at all — that the address of a removed listing should lead to an explanation rather than to the error it would otherwise have become — is the smallest and most literal enactment of the policy it describes. A directory is, in the end, a body of references held in trust; and the discipline of removal, far from contradicting the act of curation, is that act prolonged, the same judgement that once admitted a listing returning, in time, to ask whether it yet belongs. To decline the question is to suffer the slow conversion of a catalogue into a museum of dead addresses; to ask it, repeatedly and without sentiment, is what keeps the word curated answerable to the thing it names.