Someone has started running my posts through ChatGPT and pasting the results back at me as comments. Same account, every time. The output is confident, fluent, and doesn’t quite mean anything, like a very polite person talking in their sleep.
It took me a while to work out why it bothered me, because it isn’t the AI. I use AI every day. It’s that nobody’s home. There is no one behind the comment. Nothing is being vouched for. It is text with no author in the sense that matters, which is not “who typed it” but “who is answerable for it.”
And that, it turns out, is the entire AI-disclosure debate, being had back to front.
We have spent a year building machinery to answer one question: was a machine involved. Watermarks, the EU’s transparency rules, the “seems like AI slop” button. All of it detects the same thing, and I have written before about why that thing barely matters. “A machine was involved” is now true of almost everything, and it tells you nothing about whether the work is any good.
So people ask the natural next question: fine, but shouldn’t creators disclose it anyway, for trust? The answer is yes, disclose more, but not that. The thing worth disclosing was never the tool. It is the accountability.
Two questions that actually build trust
When you read something and decide whether to trust it, you are not really asking “did a human type every word.” You are asking two other things, usually without noticing. Is a real, named person answerable for this? And is it what it claims to be?
Those are the disclosures that matter. Not the keystrokes. The responsibility, and the honesty of the representation.
A blog I researched with AI, wrote in my voice, checked, and put my name to is not a trust problem. I am answerable for every claim in it. The tool that helped is nobody’s business, the same way nobody asks whether I used spellcheck, or a thesaurus, or stood at a whiteboard first.
Where disclosure genuinely matters
There is a real line, and it has nothing to do with AI. It is impersonation. Content becomes a trust problem the moment it pretends to be something it is not.
A testimonial nobody actually gave. A statistic nobody actually measured. A “here is what I learned in twenty years” post from someone who learned none of it. An expert’s byline on a piece the expert never read. Those betray trust, and they betray it exactly as much whether a human or a machine manufactured them. The machine just made them cheaper to manufacture.
That is the real disclosure obligation, and it is an old one wearing a new coat: do not claim experience you did not have, results you did not get, or endorsements nobody gave. “Made with AI” is a distraction from it. You can wear that label and still be lying about everything that counts.
What this means if you make things for a living
For a brand, the trust play is not an AI-disclosure policy. It is putting a real, answerable point of view behind the work, and never faking the handful of things that have to be real. Your reviews. Your case studies. Your numbers. The lived experience you claim.
Get those right and it does not matter what drafted the intro paragraph. Get them wrong and no amount of transparency about your tooling will save you, because you were never going to be caught out on the tooling.
The label that carries trust has never been “a machine was involved.” It is “I stand behind this.” Sign your work. Mean it. That is the only disclosure that has ever mattered, and the only one that ever will.