Editorial standards

Where a claim can be checked, it is checked before it is published — against the game itself, against official communication from the developer or publisher, or against data we have collected.

Where a claim cannot be verified, the page says so. “This appears to work like X, but we have not been able to confirm it” is a legitimate thing to publish. Implying certainty we do not have is not.

Unreleased, datamined or leaked content is labelled as such. What exists in a build is not the same as what will ship, and presenting the two identically misleads readers who are making decisions now.

Readers should be able to see where something came from and go and look for themselves. Where a claim rests on a specific source — patch notes, a developer statement, community testing, a dataset — that source is identified.

Not all sources carry the same weight, and we do not present them as if they do:

• Official communication is authoritative but routinely incomplete. Undocumented changes are common.
• Values read from the game are strong evidence, but can be misread or context-dependent.
• Aggregated play data is powerful at volume and misleading at small sample sizes.
• Community testing is often excellent and occasionally confidently wrong.
• Datamined content describes a build, not a release.

We do not republish another outlet’s original reporting as our own. Where we build on someone else’s work, we credit and link to it.

On live-service games, being wrong occasionally is a certainty rather than a sign of a broken process. What matters is what happens next.

Every network property provides a route to report an error. When a factual error is confirmed, the page is corrected, and where the error was material the correction is visible on the page rather than swapped in quietly. A reader who acted on the old information deserves to know it changed.

Where the same wrong value appeared in several places — an article, a database entry, a tool — the fix is applied to all of them, not only to the page somebody complained about.

To report an error on a specific site, use that site’s own corrections process; it reaches the people who can check and fix it fastest. For anything about these standards themselves, use our contact form.

Where a piece involves testing or hands-on assessment, it says what was actually done: which platform, which patch or version, how long, and under what conditions.

If a game, product or service was provided to us at no cost, or access was arranged by a publisher, that is disclosed in the piece. Receiving a review copy does not come with any agreement about what we will say, and we do not accept arrangements that ask for one.

Where an assessment is a judgement rather than a measurement, it is framed as a judgement. “This feels bad to use” and “this deals less damage” are different kinds of statement.

We use software tools, including AI tools, in research, drafting, data processing and engineering. That is a description of how the work gets done, not a transfer of responsibility.

A person is accountable for everything that publishes. Facts are verified by a person regardless of what produced the first draft, and “the model said so” is not a source, not a defence, and not something a reader should ever have to hear.

We do not publish machine-generated text that nobody has read. We do not create synthetic quotes, synthetic testing results, or synthetic sources under any circumstances.

Commercial relationships are disclosed where a reader meets them, not buried in a policy page. Sponsored articles are labelled as sponsored. Affiliate links are labelled as affiliate links. Paid placements are visually distinct from editorial recommendations.

Where we earn a commission from a link, that fact does not change what we recommend, and we do not add affiliate links to a recommendation we would not otherwise have made.

Advertisers and commercial partners do not get approval over coverage — of their own titles or of anyone else’s. They do not see editorial work before publication, and they cannot have a piece removed because they dislike it.

If a commissioned piece would require us to state something we do not believe is true, we do not run it.

Every Overleveling property is an independent resource. None is affiliated with, endorsed by, or sponsored by the games or publishers it covers. Game names and trademarks are used for factual identification and commentary and belong to their respective owners.

Data derived from a live game is only meaningful alongside what it describes. Where we publish figures, we aim to state the scope — mode, region, rank band, time window — the game version they reflect, and when they were last computed.

Small samples are labelled as small samples. A ranking built on very little data is presented with that caveat attached rather than styled identically to one built on a lot.

Measured data and editorial judgement are distinguished from each other. Both are legitimate; presenting an opinion with the visual authority of a measurement is not.

We cover games, systems and communities — not individual players’ private lives. We do not publish personal information about private individuals, and we do not build content around harassment campaigns or amplify them by covering them uncritically.

Where coverage involves a real person, we apply ordinary care: give them the opportunity to respond where fairness requires it, distinguish allegation from established fact, and avoid speculation about people who have not chosen public roles.

Community contributions, where a property accepts them, are moderated for accuracy and for conduct.