Wikipedia:Applying reliability guidelines

Guidelines on reliable sourcing, from WP:RS and also among various other pages, can be challenging to prioritize. Editors at times place undue emphasis on one policy while neglecting others that are more pertinent. This guide is intended to help identify which policies and guidelines are the most relevant to evaluating a particular source.

Verifiability
The first matter to consider is the Verifiability policy, which requires that readers be able to independently verify any information in Wikipedia articles by consulting a reliable source. Note that the policy demands that information be verifiable, not verified. It is not necessary to identify any particular reliable source for a claim as long as there is a consensus that one is certain to exist and could in theory be consulted. This is why no citation is required for obvious statements like, "fire is hot." There are four exceptions – material in articles that must be actually verified by citation to a specific source or sources: The ability to actually consult the source can still be mostly hypothetical, such as when the source is behind a paywall, in another language, or in a remote location. The allowable difficulties are not unlimited, though. A source must be "published", which means that it must be committed to some physical medium and legally available in some way to the public at large. This excludes such things as unrecorded conversations, personal experience, private mail, and works whose copies have all been destroyed.
 * direct and closely-paraphrased quotations
 * material that has actually been challenged (e.g., by being removed, questioned on the talk page, or tagged with citation needed, or any similar tag)
 * a claim that is "likely to be challenged"
 * contentious material, whether negative, positive, or neutral, about living persons

Types of Claims
The quality required of a source depends on the type of claim it is meant to support. Sources that are considered flawed may still be used for innocuous facts that are not subject to serious dispute. Contentious claims, or topics that are particularly technical, need sources in which editors have confidence. Exceptional claims require exceptional sources. If the claim is about a living person, stringency is significantly higher in proportion to the contentiousness of the claim. Only the best sources can be used for contentious claims about living people.

A special class of claims are those that assert a person holds an opinion. In such cases the source is judged on its trustworthiness that the person does indeed hold the opinion, not that the opinion is true. An author may be assumed to be reliable about their own opinions, but for opinions of other people, all the policies about living people apply. The question of whether the person holds the opinion may be much less (or much more) contentious than the opinion itself. That doesn't mean there is free rein to include all kinds of outrageous opinions in the encyclopedia - consensus must be reached that it is worth including.

There may be special considerations for claims in certain domains, particularly medical claims.

Reliability
Whether a source is reliable depends on both the source and the claim, with the ultimate criterion being the likelihood that the claim is true. That doesn't mean that an editor's opinion of the truth carries any weight. Instead, we look to several properties of the source as proxy indicators of its trustworthiness. The first is that there is editorial oversight, in that there are individuals besides the author who have authority or control over the material that is published. Examples include employers, editors, fact-checkers, peer reviewers, and legal advisers. The aggregate influence of these kinds of gatekeepers, encompassing the number of individuals involved, their expertise in the subject matter, and their degree of control over the publication, constitutes the principal driver for a source's reliability.

News and academic journals are the most common types of sources, but sources can also be government organizations, think tanks, museums, and more. What matters is the fact-checking process. Publishers vary in how much they reveal of their inner workings. Peer-reviewed journals are usually the highest quality sources, and will typically publish information about their review process. News organizations will usually identify their principal editors, and may also promulgate a code of ethics that those editors are expected to apply to content in the publication. Even if a source reveals nothing about its processes, a track record of issuing corrections is evidence that a fact-checking process exists and functions.

A full consideration of a source doesn't look just at the source, but also how others perceive it. Sources may not actually follow the procedures that they claim to use. Other sources might directly comment on it, for better or worse. If another source uses the first source uncritically, it can be taken as an endorsement. On the other hand, if a source is significantly out of step with others when it comes to statements of fact, it should be regarded with suspicion. This is particularly true if a source consistently fails to update or correct breaking news, while others do. However, deviation in how sources interpret facts is bias, and should be considered separately from reliability.

Decisions about reliable sourcing do not themselves need to be based only on reliable sources. Social media sentiment is worth paying attention to, especially statements from current or former employees of the source. There is no limit to what attributes could be used as proxy indicators for trustworthiness. Whether or not a given attribute matters should be determined by consensus. The overriding concern however is whether the source engages in effective fact checking. Editors should avoid undue emphasis on one specific indicator when there is no actual doubt that a claim is verifiable.

Self-publishing
Works are self-published when there is absolutely no gate-keeping between the author and publication. This includes books from vanity presses and a large proportion of web content, such as personal blogs, social media, Facebook, LinkedIn and YouTube. The determining factor in online content is not the kind of site it is on, but whether it was subject to editorial control. Self-publishing denotes a complete lack of editorial gate-keeping, not only a lack of gate-keeping by a web hosting platform. Reliable content could be uploaded anywhere.

Most reliable sourcing guidelines are subject to consensus, but one firm rule is that self-published sources can never be used for claims about living people except about the author themselves. Generally, self-published sources can only be used for unexceptional claims about their authors. The exception is for claims by experts within their respective fields of expertise. To be recognized as an expert, an author must have other publications in the topic area that have been recognized as reliable.

Conflict Between Sources
All sources have some kind of bias. Bias does not preclude using a source, though strong biases make it advisable to present contested claims as the opinion of a particular source. Sources that are themselves involved with or particularly close to their subjects are primary sources. Primary sources can be used to describe their subjects, but should not be used to judge, interpret, or contextualize. If the connection to the topic is strong enough to constitute a conflict of interest, then the source should be treated as if self-published.

When sources disagree, it does not necessarily indicate that one of them is unreliable. The first thing to consider is whether both statements are correct in light of some unstated assumptions or context. Circumstances or scholarly consensus may also have changed between the publications of two sources, in which case it might be necessary to deprecate the older source while still considering it reliable in principle. When one source comments explicitly on inaccuracies of another, it may be that the source in question is unreliable for particular claims.

Editorial Discretion
Verifiability is necessary, but not sufficient grounds for including a claim in an article. Policy does not prescribe an answer to every circumstance, so editors must proceed by forming consensus.

There's a common but misguided fatalism among editors who feel everything in a reliable source must be regarded as true, but editors are meant to interrogate their sources. If a source is inaccurate, other secondary sources cannot be depended on to notice the inaccuracy. To the contrary, even personal experiences of editors can be valid grounds to doubt a source. It is critical to distinguish between claims, which must satisfy policies like WP:V and WP:OR, versus editorial decisions about how to present these claims, which editors may support with their own arguments. The best way to handle a dubious source is often to highlight contrasting information. Including verifiable material is never original research, even if the effect is to confirm or detract from another claim. The article only cannot imply without support that one claim affects another. Policy is not meant to guard against the reader drawing inferences.

Even if accuracy is not in question, it is not necessary to cede editorial control of the encyclopedia over to the sources we cite. "Wikipedians are not mere copyists, bound to repeat simple statements absent context or without thought." Wikipedia should reflect the range of views in its sources, but is not bound to treat its subjects in the same way as a newspaper, for example, because Wikipedia is not a newspaper. Each source has its own purpose and audience that differ from those of an encyclopedia. If secondary sources do or do not comment on some aspect of a primary source, editors should take that into account while forming their own conclusions as to what is of encyclopedic interest.