Beyond AI Detectors: Why I Built an AI Disclosure Label

Over the past two weeks I have unintentionally found myself on an intellectual journey.

It began with embarrassment.

When I first started publishing work that had been assisted by AI, I was excited by what the technology could do. It helped organize ideas, improve structure, and dramatically reduce the time it took me to write. What I did not realize was that many experienced AI users could immediately recognize its fingerprints.

That realization was uncomfortable.

How could I honestly call myself a writer if readers believed ChatGPT had done the writing?

Instead of becoming defensive, I decided to examine the question more closely.

The result became a series of essays asking increasingly difficult questions.

Who is the author when AI helps write a piece?

What exactly are readers purchasing when they give someone their attention?

Does disclosure matter if the final ideas remain human?

Where is the line between using AI as a calculator and using it as a ghostwriter?

The more I explored these questions, the more I realized we were asking the wrong questions.

Public discussion has become obsessed with AI detectors.

Schools use them.

Publishers consult them.

Employers increasingly rely on them.

Yet these tools are often wrong. They routinely produce false positives, identifying genuinely human writing as AI-generated. Writers have reported changing their natural voice simply to avoid being flagged by automated detection systems. In one highly publicized case, an author claimed concerns surrounding AI accusations contributed to the loss of a reported $2 million book deal. Whether that particular case proves to be the rule or the exception, it illustrates something important: a false accusation can carry very real consequences.

A detector that incorrectly labels authentic human work does more than produce an inaccurate score. It can damage reputations, alter careers, and pressure writers to abandon the very style that makes their work uniquely their own.

That led me to an uncomfortable conclusion.

Perhaps we are trying to solve an ethical problem with a technical solution.

Instead of building ever more sophisticated machines to guess how a piece of writing was produced, perhaps we should encourage authors to simply tell us.

Not because they are forced to.

Because honesty creates trust.

Food has nutrition labels.

Creative Commons created standardized licenses.

Academic journals require conflict-of-interest disclosures.

Financial reporting depends upon disclosure requirements.

Yet AI-assisted creative work has no common language.

Every creator invents their own explanation—or offers none at all.

Readers are left guessing.

Creators are left defending themselves against algorithms.

Neither seems like a sustainable future.

I have come to believe that the best path forward is not a race between AI generation and AI detection.

It is an honor system built on voluntary self-disclosure.

Honor systems are not perfect. Some people will lie. Some people will exaggerate. Some people will refuse to participate altogether.

But authorship has always been, at its core, an ethical relationship between a creator and an audience.

If someone falsely claims sole authorship of AI-generated work, that is an ethical failure by the author—not a technological failure by the reader.

As AI becomes integrated into nearly every profession, I believe our goal should not be to build better machines to catch each other. Our goal should be to create cultural expectations that reward honesty and transparency.

With that philosophy in mind, I decided to stop merely writing about disclosure and instead attempt to build one.

Over the past several days, working alongside AI, I developed the first version of an AI Disclosure Label.

Rather than asking impossible questions like, "What percentage of this article was written by AI?" the application asks questions that I believe are far more meaningful.

Who generated the original ideas?

How much conceptual thinking came from AI?

How much mechanical execution—drafting, editing, formatting, coding, image generation, or layout—was assisted by AI?

Who exercised final editorial judgment?

The application then generates a standardized disclosure describing the collaborative process.

My hope is that this shifts the conversation away from simplistic labels like "AI-written" or "human-written." Those categories are becoming less useful by the day. Human creativity increasingly exists on a spectrum of collaboration with intelligent tools.

The objective is not to shame people for using AI.

Nor is it to celebrate using more AI.

The objective is transparency.

Readers deserve to understand how a work came into existence.

Creators deserve a vocabulary that is more nuanced than a simple yes-or-no declaration.

The irony is not lost on me that AI also helped build the disclosure tool itself.

That feels appropriate.

The application is itself an example of the kind of collaboration it attempts to describe.

This first version is imperfect.

Like any first draft, it already reveals problems I intend to solve. Even the label generated for this article contains classifications I would refine in future versions. That is exactly why I wanted to publish it publicly instead of waiting until everything felt finished.

Standards do not emerge fully formed.

They evolve through experimentation, criticism, and public participation.

Whether this particular application succeeds is almost secondary.

What matters is the larger idea.

If AI becomes an ordinary part of intellectual work—and I believe it will—then society will eventually need new norms around disclosure, attribution, and accountability.

Perhaps this is one possible beginning.

Perhaps, years from now, AI disclosures will become as ordinary as photo credits, nutrition labels, or Creative Commons licenses.

If that happens, I hope it is not because software became better at catching people.

I hope it is because creators decided that readers deserved to know the truth.

AI Disclosure Label (Version 1)

Classification: Human Directed / AI Drafted

Human Originality Score: 60/100

Meaningful human originality.
The process combines substantial human originality with AI assistance.

AI Thought: 30/100
Conceptual contribution: Brainstorming, research, organization, and analysis.

AI Mechanical: 63/100
Production & execution: Drafting, editing, coding, layout, media, and other execution tasks.

Generated Disclosure

AI participated in the creation of this essay through drafting, editing, formatting, and layout. Responsibility for the ideas was entirely human, while responsibility for the final form was shared. The questionnaire calculated an AI Thought score of 30/100, an AI Mechanical score of 63/100, and a Human Originality Score of 60/100.

This disclosure was generated by the AI Disclosure Label Version 1 application. As with the application itself, it should be understood as an evolving attempt to describe human–AI collaboration rather than a definitive measurement of authorship. The work was reviewed and approved by the human author, who accepts responsibility for the published content.