Labeling AI-generated content: the honesty rule for business publishing

A label is not honesty: a publisher is accountable for every word it ships, whichever tool held the first pen.

Every piece a company publishes is a claim about how that company works, and readers have started checking the claim. They want to know who wrote the page in front of them: a person who stands behind it, a machine that assembled it, or a mixture nobody will admit to. The question used to be invisible and it no longer is. A whole generation of readers can feel assembled text the way they feel a borrowed photograph, and they discount it accordingly.

The market answers with two false versions. The first hides the machine and passes its output as human craft, a small deception repeated until it hardens into house style. The second does the opposite and stamps an AI-made badge on everything, as though the tag were either a confession to be ashamed of or a trophy to boast about. Both dodge the question the reader actually carries: not who typed it, but whether any of it can be trusted.

The practice that holds is accountability without costume. The publisher is accountable for everything it publishes, whichever tool drafted the first version: facts verified before they ship, claims checked against their sources, and a voice that belongs to the company rather than to a tool's default settings. A machine can assemble a draft in minutes. It cannot be wrong on your behalf. The name at the bottom of the page is where responsibility lands, and it lands there regardless of what drafted the text above it.

Disclosure has a real place, and the place is defined by the reader's judgment. Where the reader is asked to weigh the thing itself, say plainly what it is: a client report that presents analysis, a photograph presented as a photograph, an article whose conclusions a buyer will act on. There the label protects the reader's ability to judge, which is exactly what disclosure is for. Elsewhere it is routine noise, and routine noise trains people to ignore the labels that matter.

One boundary deserves its own sentence: never let the label do the work honesty should do. A fabricated number wearing a disclosure sticker is still a fabricated number, and a manufactured testimonial tagged as machine-made is still manufactured. The badge answers how a thing was made and says nothing about whether it is true, and truth is the question the reader came with. A labeled lie is still a lie.

We hold ourselves to this in the open. We write with tools and publish under our own names, and nothing leaves this company that we are not willing to defend as if we had typed every word. Our assistant runs the same principle in the other direction: when someone asks whether she is AI, she says yes, plainly, because a company that publishes under its own name should not have its front door pretend otherwise. The same honesty runs through everything we ship.