Where all of this comes from

Every source this programme knows of: the ones it reads, the ones it does not, and what was agreed with each.

How we know this

This is the default view and it holds nothing written to test the engine. The validation records still exist and are counted below; they are not shown here, and no number on this page is computed from them.

117 × real data · 0 × synthetic validation data

Only real records are shown. The map is incomplete on purpose: it holds what has been read and checked, not everything ever published. Where something is missing, the page says so instead of filling the gap.

A source is switched on only when a contract states which identifier it speaks, what shape its answer has and on what basis anything may be held at all. Sources that are off are listed here too, with the reason — a source absent from a registry reads as one nobody thought of.

Read

3 sources are read

Source	Kind	Identifier it speaks	Basis for holding anything
Crossref	Index	DOI	Crossref distributes bibliographic metadata under CC0. Deposited abstracts remain under publisher terms, which are not uniform, so no abstract text is stored. Provisional reading, needs legal confirmation.
PubMed	Index	PMID	NLM permits retrieval and local processing of citation records. Copyright in article text remains with publishers, so the default is metadata and a locator. Where the NCBI open-access service states a licence for a specific article, that licence governs that article. Provisional reading, needs legal confirmation (BL-07).
ClinicalTrials.gov	Trial registry	NCT	Registry records are a work of the US federal government and carry no copyright. Whether text a sponsor wrote and the registry reproduces is covered by that is not settled, so the ceiling is metadata and a locator: no brief summary is stored or shown without a confirmed record-level licence. Attribution, the registry version we processed and the fact that a body was withheld are recorded on every retrieval. Owner decision 27.08.2026, fail-closed until BL-07 closes.

Described in advance and not read

0 sources are described in advance and switched off

Each of these is a decision, not an oversight. What has to happen before it could be read is stated beside it.

Every declared source is switched on.

This map holds only what has been read and checked against a source, and nothing has been recorded for this section. Empty here means exactly that — not zero, and not a fault in the page.

What a kind of source may say

Authority is not a property of a source. It is a property of a source and a subject together: a clinical-trial registry is the last word on the status of a trial and says nothing about an effect size.

Refused outright

Corporate · Measured effect — A company announcement is a primary source about the announcement and about nothing scientific. The effect belongs to the paper or the registry entry, and recording it from here would put an unreviewed number in the canon.
Corporate · Retraction — Whether a paper has been retracted is stated by the journal that published it, never by a party with an interest in the answer.
News · Measured effect — News reports a result; it is not the record of one. Every number in it has a source that this canon can hold instead.
News · Retraction — A retraction is a publisher's act. Reporting on it is not performing it.
News · Trial status — The status of a trial is the registry's record. A news account of it is already out of date at the moment it is written.

Authority written down in advance

Kind	Subject	Rank	Why
Conference	Publication record	2	Proceedings record what was presented, which is not always what was later published.
Corporate	Funding	3	A company is a primary source about its own money and no more than that.
Funder	Funding	5	The award record is held by the body that made the award.
Index	Publication record	3	An index copies the publication record accurately and is not its origin; it may lag a correction by the publisher.
Index	Retraction	3	Indexes carry retraction flags reliably and receive them second-hand.
Institutional	Affiliation	4	An institution is the record of who works there; the paper's author line is a statement about the moment it was written.
Patent office	Patent claim	5	The granted claim is the register's text. Any other wording is a summary.
Preprint server	Measured effect	2	A preprint reports an effect that no reviewer has yet checked.
Publisher	Publication record	5	The journal is the record of what it published, including its version.
Publisher	Retraction	5	A retraction notice is issued by the publisher of the paper.
Publisher	Measured effect	4	The peer-reviewed paper is where the effect is reported and checkable.
Regulator	Regulatory action	5	A regulatory action is an act of the regulator. A report of it is a report.
Regulator	Retraction	3	A regulator may withdraw an approval; the retraction of a paper belongs to the journal that published it.
Trial registry	Trial registration	5	The registration IS the registry record. Nothing else is closer to the fact.
Trial registry	Trial status	5	The sponsor updates the status in the registry; every other account of it is a copy of that record, made at some earlier moment.