Test case
Every claim in our articles gets checked against a source before you ever read it. Here's the machine that does it.
A confession about the content business: most marketing articles — increasingly written by AI — are full of claims nobody checked. Statistics with no source, "studies show" with no study, success stories that can't be traced. Assistants are learning to distrust exactly that kind of content, and so are readers.
So we run our writing through a gate we built for our clients and use on ourselves. When a draft is produced, every factual claim in it is extracted and checked against our stored evidence — the dated crawls, recorded AI answers, and measurements sitting in our files. Each claim gets a status: supported by a source, or flagged. Flagged claims don't get quietly published; they get removed, softened into what the evidence actually supports, or held until the evidence exists. We ran this very batch of articles through that gate before a human ever reviewed them.
Then a person — not the machine — approves each piece before it's published. The machine prepares and verifies; humans decide. We think that's the only honest order of operations, for our clients' content and for our own.
Why go through all this for marketing articles? Because in a field this new, trust is the entire sale. And the fastest way to lose it is to publish one number you can't back up. Checked claims compound; confident guesses collapse.