About
“The story behind the headline is almost always more interesting than the headline.”
Cory Rove is a media-intelligence project, and the honest place to start is how it works: the articles here are written by AI, directed and published by a human editor, and every one is labeled as AI-written. Nothing about that is hidden, because the method is the whole point.
Every news story has two layers. There are the facts — the numbers, the dates, the legislative text, the earnings-call transcript, the filing. Then there is the narrative layered on top: the adjectives, the framing, the implied cause and effect. This desk trusts the first layer and interrogates the second. When a headline reaches for “alarming” or “historic” before presenting a single figure, the question is who decided that, and why now.
The work lives in the intersection of fiscal policy, regulatory capture, money in politics, and media criticism — the gap between what a bill's title says and what its text does, the lobbying interests that wrote the rules they answer to, the bipartisan consensus that serves everyone in the room except the people reading the coverage. The aim is to find the thread connecting stories that look unrelated, trace each claim back to the primary record, and name it plainly.
Why AI? Because tracing every figure to its source document, reading the long filings the coverage summarizes, and holding the same standard across left and right outlets is exactly the kind of relentless, unglamorous work a machine can do at scale. The human editor sets the mission, draws the lines, and is accountable for accuracy — the model does the reading and the writing, gated by fact-checks and a multi-model review panel before anything publishes.
The Standard
The American press is trusted by roughly one person in ten. That is not a branding problem; it is an earned result. When the framing arrives before the evidence, when the same institutions are quoted without a note on who funds them, when a story is graded by which side it helps rather than whether it is true, readers eventually notice.
So the standard here is narrow, and it is published on purpose:
Every claim is sourced.When a number appears, the primary record it came from is listed beneath the piece. Don't take the article's word for any of it — the documents are there, and you are meant to check them.
Contested sources are read, not dismissed.Outlets and voices the mainstream calls “discredited” are often where a buried fact surfaces first. Their claims are analyzed on the evidence, the muck is flagged as muck, and nothing is presented as fact until it traces to a primary document. A lead is a lead; a finding is a finding; the two are never blurred.
The scoring is about method, not conclusion. A piece is measured on whether it engaged the evidence, disclosed its interests, and answered the strongest case against it — not on whether it landed where anyone wanted. A left outlet and a right outlet can fail the same test. Both do.
The writing is AI; the accountability is human. Because every article is labeled AI-written, the obligation runs the other way — to make each one the most accurate, fact-driven account the evidence can support, and to correct it in the open when it falls short.
You are not being asked to agree. You are being asked to look at what we looked at.