Who publishes GamGem, and how it is sourced

GamGem is an independent reference site for games, published by the GamGem Research team. It is not affiliated with, endorsed by or sponsored by any game publisher. Where we write about a game we are writing about somebody else’s work, and we say so on every page.

Most of what makes this site different is not the data — it is what is attached to it. A figure here carries where it came from and when we last looked. Where sources disagree we publish the disagreement, instead of quietly picking one.

• Every fact about the world outside this site carries a source and the date we checked it.

  A build gate reads every claim table in our own documentation and fails the build if a row has no external URL, no ISO date, or cites us instead of a source outside this site. No count is printed here on purpose — a number typed into a sentence stops being true and nobody notices.

• If we could not open a page ourselves, the row says so.

  Claims carry a first-hand flag. Where a source refused an automated reader, the badge reads “Reported to us” rather than quietly matching the rows we did read.

• A number we do not have is never printed as a dash, a zero or an estimate.

  Where a figure is unknown, the field is null and the row is not rendered. “Not established” is a first-class answer here, not an empty state.

• A floor is never rewritten as a total.

  “Over 100” is not read as 100. The parser returns null for any bounded phrase, and a test fails if that guard is removed.

• Nothing on this site states a fact that came from a language model.

  Generated text explains and arranges what sources already establish, and every explanation shows the rows it drew on. It never adds a claim of its own.

• When we are wrong, the old answer stays visible beside the new one.

  Superseded grades are kept in each claim’s history with the date and the reason, so a reader can see what we changed rather than only what we now say.

• Leaked footage, leaked images, leaked builds or anything derived from them.

  It is the publisher’s property however it reached the internet. We report that a leak happened and what outlets said about it; we host none of it and no claim here rests on it.

• Screenshots, key art, renders or any artwork belonging to a game’s publisher.

  Rockstar’s policy on posting its copyrighted material covers non-commercial fan use and says it does not extend to digital publishing. An ad-supported site is outside it, so every illustration here is our own.

• Real players’ names, real team badges and real tournament identities.

  Publisher developer policies prohibit deanonymising players. Leaderboards and pro builds here run on fictional identities and say so.

• A statistic nobody published.

  Win rates, tier lists and performance figures that do not exist in an official source are derived in the open and labelled as derived — never presented as measured.

Every claim on this site carries one of six grades. They describe where a statement came from, not whether we believe it — a leak is graded a leak whether or not it later turns out to have been right, and one on this site did. When you quote a claim from here, the grade is part of the claim.

• OfficialStated by the publisher in its own material.
• Shown, not statedVisible in published material but never stated.
• Community reconstructionAssembled by an outside project from published material.
• Leak / data-minedCirculating from an unpublished source. Never evidence on its own.
• DisputedTwo readings disagree and neither has been settled.
• Still openNobody has answered it yet.

A grade can move, and the old one is kept. Where a claim has been regraded, its history shows what it was, when it changed and why — so a reader can see what we changed our minds about rather than only what we now say.

Everything published here may be quoted, indexed, summarised and used to train a model, at no cost and without asking. The one condition is attribution, and the terms below are what attribution means. They are also stated for machines, in RSL at /license.xml, so a crawler reads the same answer a reader does.