Skip to main content
Transparency

How we decide
what to publish.

GambleLABS publishes a record when its evidence can be shown, and publishes nothing when it cannot. The rules that decision follows are written down here, in advance, in enough detail that they can be used against us.

Sources publishedEvidence level on every recordCorrections permanentNo paid placement

01The short version

Five questions, asked in order

Every record on this platform has to survive the same five questions before it is published. They are asked in this order because a failure at any step stops the record where it stands.

  1. Can it be sourced? Every factual claim about a named party must reach a source record — a document, a register entry, a filing, a notice. A claim that reaches no source is not published, however plausible it looks and however many people repeat it.
  2. Is it the right kind of source? A regulator’s own notice outranks a report of that notice. A corporate filing outranks a press release describing the filing. Where we are relying on the weaker option, the record says so rather than letting the reader assume the stronger one.
  3. Does the evidence identify the party? Company names repeat across jurisdictions and brands change hands quietly. Before a company is named in a record it has to be identified by something durable — a registration number, a licence number, a register entry — not by a name that matches.
  4. Has the subject been able to answer? Where a record describes what a named party did, that party is approached before publication and its response is published alongside the record, in full, whether or not it changes what we write.
  5. What happens if we are wrong? Corrections are permanent and attached to the record they amend. A retracted record stays where it is, marked as retracted, so that anyone who read it can find out that we withdrew it.

The rule behind the rules

An empty section is a true statement. A section filled with estimates, samples or placeholder rows is a false one. Where GambleLABS has no evidence, the platform shows an empty state that explains what would fill it — and links to the reason it is empty.

See the ingestion contract

Where machines stop

The Evidence Engine reads sources, extracts entities, proposes relationships and scores its own confidence. It cannot publish. Every finding it produces is held for a named human reviewer, and an unreviewed finding is not a record.

How the engine works

02The full set

Standards and policies

Eight documents. Together they cover what we accept as evidence, where our data comes from, what we get wrong, who pays us, and what we do with what you send us.

03Editorial workflow

The route a record takes

Editorial state is stored on the record itself rather than implied by its presence in the database. These are the states a record moves through, and what has to be true before it moves.

  1. 01Draft

    Intake and drafting

    The intended source-backed workflow begins with a filed source — including its publisher, retrieval date and archive position — and the draft is written against it. A record without that relationship must disclose the gap rather than imply source provenance.

  2. 02In review

    Verification and reply

    For a source-backed record, a second person checks material claims against the filed source. Where the record names a party and describes its conduct, that party is contacted and given a stated period to respond before publication.

  3. 03Approved

    Editorial sign-off

    An editor confirms the evidence level, the source list and the right-of-reply position. Approved means cleared to publish — it does not mean published, and an approved record is not readable by the public or reachable through search.

  4. 04Published

    Publication

    The record becomes public with its evidence level, source count and recorded review state visible. A zero source count or missing review date remains visible. Nothing publishes itself: publication is an action taken by a named person.

  5. 05Archived

    Review, correction, archive

    Records carry a last-reviewed date. When something is wrong we issue a correction against the record rather than editing the claim away; a withdrawn record stays in place, marked as retracted.

04Live counts

What the record set contains

Read from the database at the moment this page rendered. Nothing here is estimated, projected, rounded up or seeded with a starting offset.

Sources on file
0
Every source record held, including those attached to unpublished drafts.
Published editorial records
17
Editorial records readable by the public. Drafts and approved-but-unpublished work are not counted.
Ingested source records
2,914
Retrieved source records currently held in the ingestion provenance store. Ingested does not mean published.
Corrections issued
0
Clarifications, corrections and retractions in the public log.
Jurisdictions reviewed
0
Of 230 tracked. The remainder are listed with their regulator and marked unreviewed.

Ingestion status

6 of 35 declared upstream feeds are connected. Where a feed is not connected, the section of the platform it would fill shows an empty state — not an estimate, and not a sample.

6 connected29 outstanding
Feed-by-feed status

05Limits

What this platform will not do

Stated as flatly as we can manage, because these are the behaviours that make industry coverage worthless.

  • Publish a claim we cannot source, on the basis that it is probably true.
  • Fill an empty section with sample rows, estimates or illustrative figures.
  • Rank, score or recommend an operator — for money or otherwise.
  • Remove a published record because its subject asked, or paid, or threatened.
  • Publish a private report, or any part of it, without the reporter’s explicit consent.
  • Edit a published claim quietly instead of issuing a correction.
  • Let an automated finding reach a reader without a named human approving it.
  • Describe a jurisdiction as reviewed when no review has been completed.
Report a case Request a correction