Not every citation is worth the same. This site sorts its sources into five layers, publishes the layer next to every reference, and refuses to publish a dossier whose weight rests on the bottom three.
Footnotes flatten everything they touch. A cuneiform tablet in the British Museum and a paragraph on a well-meaning website look identical in a list of references, and the reader has no way to tell which one is carrying the argument. The layer system exists to undo that flattening, and it is enforced by software rather than by good intentions.
T1 — The primary object
The ancient artefact itself: a tablet, a prism, a stele, an inscription, a papyrus, a coin, an excavated stratum. Cited by museum or excavation number and by the line or column referred to, so that a reader can go and look at the same thing.
A T1 citation on this site names the repository, the object number and the locus. “An Assyrian inscription” is not a T1 citation. “BM 91032, column III” is.
T2 — The critical edition
The scholarly publication that establishes the text of a primary object: the transliteration, the apparatus, the translation prepared by a specialist who worked from the object or from photographs of it. Almost nobody reads cuneiform off the tablet, and pretending otherwise would be a worse dishonesty than citing the edition.
T2 is where most real work happens, and it is the layer that makes a claim checkable by a reader who is not a specialist. A critical edition tells you which signs are damaged, which readings are restored and which are conjectural — information that disappears entirely by the time a claim reaches a popular book.
T3 — Peer-reviewed scholarship
Journal articles and academic monographs arguing an interpretation of the primary evidence. Indispensable for understanding what a piece of evidence means and where the disagreements are — but an argument about evidence, not the evidence.
T4 — Reference works
Dictionaries, encyclopaedias, handbooks, textbooks. Useful for orientation and for finding out what the standard position is. They summarise, and summaries lose exactly the qualifications that a dossier needs to report.
T5 — Everything else
Popular books, documentaries, websites, apologetics and counter-apologetics of every stripe. This layer appears on the site only when a page is describing a claim in circulation — “this is what is commonly said” — and never as the support for a finding.
The rule the software enforces
A dossier cannot be published unless at least two T1 or T2 sources are attached to it. A topic study needs at least one. The check runs when the page is saved and again at the moment a scheduled page is due to go live, and a page that fails it returns to draft. There is no override in the interface, deliberately: a button marked “publish anyway” is a button that gets pressed on a bad Friday.
Linked is not resolved
Attaching a T1 or T2 source to a page is not enough. Each source record carries a permanent identifier — a CDLI P-number for a tablet, a DOI, an ISBN, an Internet Archive identifier — and software resolves it: it fetches the record, compares the metadata that comes back with what the citation claims, and stores the result with a date. Until that result is resolved, the source record itself cannot be published, and no dossier resting on it can be published either.
This is the strictest rule on the site and the one most likely to be mistaken for pedantry. It is not. The single most common failure in writing of this kind is a citation copied from another text whose author copied it from somewhere else, with an error introduced two steps back and carried forward by everyone downstream. An identifier that resolves breaks that chain: it points at an object, and the object either matches the citation or it does not.
Be clear about what this proves and what it does not. It proves the identifier beside a source opens the item described, and that the metadata agrees. It does not mean anyone travelled to a museum, and this site does not claim that. The narrower claim is the one that can be tested, and a reader can repeat every test from the verification log, which lists every source with its identifier, its current state and the date it was last checked.
The failure this catches has a name here: a mismatch — an identifier that resolves to something other than the item cited. It is the signature of an invented or mis-copied citation, and it is the reason the whole apparatus exists. Every mismatch ever caught is counted publicly, month by month, and the count is never reset.
The consequence is not a slow site. It is a checkable one. There is no monthly quota, because a quota is a promise about effort and effort is not something a reader can verify. The limit is the rule above, and it has no exceptions: two T1 or T2 sources per dossier, every identifier resolving, everything published re-tested weekly, and nothing that publishes anyway. We would rather publish pages that survive being checked than pages that were never checked at all — and the difference is now something you can confirm rather than take on trust.