Method

Why a fixed pattern set

The obvious approach is to search certificate transparency logs for every domain containing the brand string. We started there and stopped. The public crt.sh substring search truncates: a query for one brand's exact domain returned 1,652 records, while the substring query for the same brand returned 350. A separate collection run across 36 brands came back with between zero and two domains each. Numbers built on that source would understate reality while being presented as a measurement.

So we test a fixed set of 648 patterns identically for every brand. This answers a narrower question with a reliable number: of the same obvious variants, how many are taken. It is a benchmark that can be compared across brands and repeated by anyone.

The steps

1. GenerateEvery pattern in the set is combined with the brand token and each tested TLD.
2. ResolveEach candidate is checked by DNS. No resolution means the domain is not in use and we drop it.
3. RetrieveEach resolving domain is fetched over HTTPS, then HTTP.
4. ClassifyThe response is sorted into live third-party, parked, operator-owned, unrelated or dead.

Classification rules

A domain counts as live third-party only if it returns a page, the page carries the brand as a standalone word, and the response does not resolve to the operator's own domain. Substring matching was not enough: it found one brand's name inside another independent company's name and counted a separate business as a hijack. We match on word boundaries instead.

A page is parked if it carries known registrar or sale-page markers, or if its text runs under 600 characters. Short bodies are holding pages in practice.

What this does not measure

It does not find creative variants outside the pattern set, misspellings, or internationalised domain names. It does not measure traffic, and a live page is not evidence that anyone visits it. It does not cover the search results themselves, which we run per client rather than publish.

Patterns that generate independent English words were removed from the set after testing showed they produced most of the false positives.

Sample

The sample is public gambling brands with a known primary domain. Brands we have a commercial relationship with are excluded in either direction: we do not measure our own clients for a public index, and we do not publish research on a brand while we are working against it for someone else.