Wikipedia:Always cite contested material in the lead

WP:V is unambiguous: ...four types of information must be accompanied by an inline citation to a reliable source that directly supports the material:

• direct quotations,
• material whose verifiability has been challenged,
• material whose verifiability is likely to be challenged, and
• contentious material about living and recently deceased persons.

The guideline in WP:CITELEAD, which allows editorial consensus to sometimes omit citations in the lead, applies only to things that do not fall into any of those four categories. WP:V, as core policy, overrides WP:CITELEAD when the two contradict; specifically this means that CITELEAD can allow us to omit citations only in cases where the text has not been challenged. As soon as someone challenges any statement in the lead, it must have a citation clearly associated with it, and that citation cannot be removed.

For WP:BLP-sensitive statements, the requirement is even more severe. Uncited BLP-sensitive statements in the lead are clear violations of WP:BLPREMOVE and should be removed on sight; any attempt to restore them without a citation should be instantly reverted (and the WP:3RR does not apply to such removals.)

Why aren't citations in the body sufficient?

As a policy matter, WP:V requires that the citation accompany the text it directly supports. Citations in the body do not do so; it is often unclear which citations in the body support which statements in the lead. Placing something somewhere on the same page is not accompanying it.

Beyond that, because citations in the body do not accompany the text they supposedly support, it is easy for the lead and the body to end up out of sync - changes, even individually minor grammatical ones, can collectively add up to the point where text no longer reflects the citations it was originally intended to. Text (and sources!) in the body can be removed without anyone noticing that there was accompanying text in the lead; text in the lead can be rewritten by well-meaning editors who believe they're not changing the underlying meaning, or even by editors who do change the underlying meaning without understanding Wikipedia's sourcing requirements.

When this happens to text with a clearly-associated supporting citation, later editors can easily double-check and verify uncertain statements, correcting these issues. When it happens in an uncited lead, however, there is often no recourse beyond reviewing the entire article.

Because of these things, citations in the body cannot satisfy WP:V for statements in the lead; they cannot be reasonably verified to support any individual statement there.

When text in the lead of an article is challenged or removed, it is never, under any circumstances, appropriate to simply say that it is cited in the body. Per WP:V, The burden to demonstrate verifiability lies with the editor who adds or restores material, and it is satisfied by providing an inline citation to a reliable source that directly supports the contribution. Vaguely waving one's hand at the entire article and reassuring editors that a citation exists somewhere in that potentially massive list of hundreds of citations does not, in any way, shape, or form, meet the requirement to provide an inline citation that directly supports the contribution. Policy requires that the specific citation be presented so it can be examined and verified; and the required method to present it is to add it, inline, to accompany the statement it supports.

But WP:CITELEAD says...

WP:V is core policy and is not subject to editorial consensus. WP:CITELEAD cannot override it, and specifically cannot allow a local consensus to override it; the requirement to supply associated citations for every statement that has been challenged is non-negotiable. At best, the option for a consensus to neglect citations can apply only to text that does not meet any four of the criteria under WP:V for which an accompanying citations are required.

That said, there is a reasonable interpretation of CITELEAD that is consistent with V; specifically, people don't have to add citations to the lead simply for the sake of adding citations. If the sole reason someone wants to add citations to the lead is because they think citations are obligatory, and nobody is challenging or has challenged the actual text (and it doesn't meet any of the other criteria listed under V that require citations), then citations can still be omitted. In short, CITELEAD does establish that people shouldn't go around adding citations to leads solely for the sake of adding citations.

This is, however, exceptionally rare; in the vast majority of cases, when someone demands citations for the lead of an article, it is because they're challenging the underlying text. In those situations, citations must be provided, and invoking CITELEAD is inappropriate.