Find sources
The desks' own sourcing, on its own - type a few words and read the real sources they would write from, without spending an article.
Find sources is in the left rail. It answers one question: what would the newsroom actually write this from?
The same sourcing, exposed as its own door
Before the newsroom writes anything it plans the search, reads the pages, and judges each one. Find sources is that step run on its own. Type a few words and read what came back - no article is spent, and nothing about it is a preview or a simulation. It is the same work, run for you to read first.
Every result says why it ranked
Each source carries its own reasoning: how much the desk trusts the outlet, and how much of your subject actually appears in the page. A page that mentions your subject once is not the same as a page about it, and the ranking says which one you are looking at.
The run also states what it understood your query to be about. That is the fastest way to find out that you and the desk meant different things by the same three words, before an article is written on the wrong one.
Zero sources is an answer
A search that finds nothing usable says so, and names which of the recorded reasons it was. It does not pad the list with something adjacent to fill the space - a thin source that made the count is worse than an empty result you can act on.
A thin answer offers to look harder rather than presenting itself as the whole answer. The deeper pass is a button you press rather than a guess made on your behalf: it costs real time, and on a subject with genuinely thin coverage it changes nothing.
What it does not do yet
Picking sources and pressing Write from these sends them into the pipeline as the article's priority sources. That path was broken until recently - the desks re-judged your own picks and discarded them - and the fix is deployed but has not yet been verified by generating an article end to end. Until it is, treat Find sources as the place you look, not as the hand-off.