Research methodology

How jurisdictions are chosen

A jurisdiction is published only when there is enough verifiable primary-source material to make a genuinely useful page. The test is not whether a regulator exists; it is whether we can document, from official sources, the things a reader actually needs:

• the regulator, its current name, and its own website;
• a public register a consumer can reach, and what it can be searched by;
• what the register returns, and whether it shows licensed domains;
• the complaint route, and which body can actually decide a dispute;
• the self-exclusion arrangements, or a clear statement that none is documented;
• at least one legal or technical standards document.

Where that material does not exist or cannot be verified, we do not publish the page. A thin jurisdiction page padded out to look complete is worse than an absent one, because it implies a verification route that is not there. The page count is an output of this test, not a target set in advance.

Source hierarchy

Sources are ranked, and a lower tier never overrides a higher one.

Search engines are used to locate official destinations, never to answer questions. A search summary is not a source. During this research pass that distinction caught real errors: on minimum gambling ages in Great Britain, search summaries returned a figure that had been superseded, and the correct product-specific position only emerged from the regulator’s own pages.

The research process

1. Research before drafting, and write it to disk

   Each jurisdiction and topic is researched first and recorded in a structured dossier — claim, jurisdiction, authority, source title, source URL, date checked, verification status, limitations — before any page is written. Pages are then drafted from the dossier, not from memory. This is a deliberate guard against the failure mode where a plausible-sounding sentence survives into publication because nobody could remember whether it had a source.

2. Fetch the official page, do not infer it

   Every URL published on this site was retrieved during research. URLs are never constructed from a pattern, guessed from a site’s structure, or recalled. A URL that looks obviously right is still fetched, because a plausible path that returns a 404 is exactly the kind of error that survives review.

3. Record the authority’s own vocabulary

   Where a register uses particular words for licence statuses, field labels or scheme names, those words are recorded as the register uses them. Paraphrasing a status value into something more readable makes it impossible for a reader to match what they see on the regulator’s screen.

4. Re-verify independently

   Each dossier is then checked by a second pass that re-fetches every claimed official URL and re-tests the consequential facts, working on the assumption that the first pass was careless. Its job is to find dead links, outdated organisation names, rules stated more precisely than the source supports, jurisdiction-specific rules written as universal, and deep-link patterns that do not actually work. Corrections are recorded against the dossier.

5. Publish only what survives

   Claims that fail verification are removed or qualified. Gaps are carried forward into the page as stated limitations rather than quietly dropped, because a reader needs to know which questions this site cannot answer.

How a register is documented

Describing a register accurately is harder than it looks, and getting it wrong wastes the reader’s time at the moment they are trying to protect themselves. For each register we record:

What it accepts

The exact label on the search field, so a reader can find it on the page. Whether it accepts a domain matters most, because a domain is the one thing a reader definitely has.

What it returns

The fields shown in results, and specifically whether licensed domains and a licence status are among them. A register that lists companies but not domains cannot answer "is this website covered?".

Its status vocabulary

The literal status values, recorded as the register writes them. Where the authority publishes no definition of those values, we say the vocabulary is confirmed and the definitions are not, rather than inventing plausible meanings.

What has to match

The fields a reader must compare — usually the legal entity rather than the brand, and the exact domain. This is the step people skip, so it is stated explicitly on every route.

What it cannot tell you

Coverage limits, absent fields, stale "last updated" dates, and the general point that authorisation is not a prediction of conduct.

Deep-link policy

Some registers accept a search term in the URL, which lets a tool open the regulator’s own results page directly. Most do not. The difference is not guessable, so it is tested.

A deep link is published only when the pattern was executed against the live register with a real term and observed to return matching results. Where a pattern was tried and the register ignored it — returning the full unfiltered list, or a byte-identical page — that is recorded as a negative result and the tool falls back to copy-and-open.

In this pass, 2 of 9 jurisdictions had a query pattern that actually worked. The remaining 7 are form-driven: the honest instruction is “type it into the field”, and offering a link that silently returns the whole register instead would be worse than offering none.

We do not reverse-engineer undocumented endpoints, and the checker never queries a regulator’s systems programmatically — see the checker limitations below.

When an official source is blocked

Regulators are entitled to protect their systems, and several do: bot mitigation, rate limits, challenge pages, region restrictions, and interfaces that exist only inside a JavaScript application. Automated research runs into all of these.

When that happens, the rule is absolute: a blocked source is never filled in from inference. We do not deduce what a register must contain, we do not reconstruct a legal requirement from memory, and we never substitute an affiliate page to obtain a source that loads. Specifically, we never manufacture a licence status, a licence number, an operator name, a licence category, a register result, an enforcement outcome or a regulator requirement.

Instead the fallback runs in order:

1. look for another official page from the same authority that documents how the register works;
2. look for official government or legislative documentation of the same rule;
3. use search only to locate an official destination, then fetch that destination;
4. if the position still cannot be established, state the limitation on the page and give the reader the manual procedure.

Review dates

Each regulatory page carries a Last reviewed date. It means one thing: on that date, the sources behind that page were read again. It is not updated when a page is touched for formatting, and it is never bumped to make the site appear freshly maintained.

Where an official page displays its own “last updated” date, we record it, because the age of the source matters as much as the age of our reading of it. A regulator page carrying a date from several years ago is a caveat we pass on rather than hide.

What the licence checker does and does not do

The checker is a routing and verification-assistance interface. Its limitations are structural and deliberate.

• It makes no network requests. No fetch, no background request, no hidden frame, no proxy. Regulator systems block cross-origin requests, and a tool that pretended otherwise would fail silently in exactly the situation where a reader was relying on it.
• It holds no operator records. It knows where registers are and how to use them. It does not know who is licensed.
• It issues no verdicts. There is no safe, trusted, verified or recommended state, and no score. The statuses it displays describe what a register offers or what a reader observed — never a judgement about an operator.
• Your input stays in your browser. The routing table is embedded in the page, so no request is needed to use the tool, and nothing you type is transmitted to us. See the privacy notice, which is careful about the difference between that and a claim that nothing about your visit is recorded anywhere.

Licensing information is not endorsement

This is the load-bearing statement of the whole project, so it is stated without qualification:

A licence means an operator is authorised by that authority and subject to its rules, including whatever consumer protections that regime imposes. It does not mean the operator is solvent, that it will pay a disputed withdrawal, that its terms are fair, or that gambling with it is safe for you. Licensed operators are sanctioned by their regulators regularly, which is itself evidence that authorisation and good conduct are separate things.

We publish no rankings, no reviews, no scores and no recommendations, and nothing on this site should be read as encouragement to gamble with anyone.

Known limitations of this methodology

Being useful requires naming the weaknesses.

• Point-in-time. Research captures a register as it behaved on the date checked. Regulators restructure sites and change search interfaces without notice.
• Language. Several regulators publish authoritative material only in their national language, with English pages that are summaries or are updated more slowly. Where a non-English source was relied on, the dossier records it.
• JavaScript-only interfaces. Some registers and portals cannot be read without running their application. Those are documented as far as their official descriptions allow and flagged as requiring a browser.
• Documents behind files. Where a register is published as a downloadable PDF or spreadsheet, filenames often encode a date and change at every update, so we link the register’s landing page rather than a file URL that will break.
• Absence is not proof. Where we could not find evidence of a scheme or a rule, pages say “not documented in the sources reached” rather than “does not exist”. Those are different claims and only one of them is supported.
• No independent audit. Nobody outside this project has reviewed these pages. The corrections process is the mechanism for that, and it depends on readers using it.