Don’t Ask How Much AI Is in It. Ask How Much of You Is.

Writers are now deliberately making their writing a little worse to prove they are human.

Roughing up a clean sentence. Leaving in the em dash they would normally cut. Adding a small stumble on purpose, so nobody suspects AI touched the work.

That is when I started wondering whether we are asking the wrong question. Was it written by AI? Assisted? Edited? Would a detector catch it? All reasonable-sounding questions, and all getting us remarkably little further.

Perhaps the more useful question is not How much AI is in the work? but How much of you is in it?

Ann Handley recently approached this mess from a wonderfully pragmatic direction. Her question is not how to make our work look more human, but what we should consciously not hand over because doing it ourselves helps us think, develop taste, grow or simply gives us joy.

I like that question because it makes a distinction we easily lose in the efficiency debate: sometimes the process produces more than the output. But it also leaves me wondering about another part of the same puzzle: when we do use assistance, what makes the resulting work meaningfully ours?

“Sometimes the process produces more than the output.”

Substantial outside help in creating something is hardly new. Plenty of highly acclaimed creative work has been made with technology, collaborators, assistants, specialists and editors. We have simply become unusually anxious about one particular new participant in that process.

Take iconic rock band Queen. For years their album sleeves proudly carried some version of “No Synthesizers.” That sounds like a declaration of technological purity, except Queen were anything but purists. They layered guitars and voices, manipulated tape and pushed recording technology hard enough to create sounds people sometimes mistook for synthesizers in the first place. Technology was all over the work. Their boast was really about where the remarkable part came from.

Famous painters make the same question more interesting. Raphael and Rembrandt worked with assistants and workshops; Andy Warhol turned his Factory into something that deliberately blurred the line between artist, studio and production system. The master’s own hand could be absent from substantial parts of a work carrying the master’s name. Yet we don’t automatically conclude that it therefore isn’t a Raphael, a Rembrandt or a Warhol. We look at conception, direction, choices and judgement, and sometimes argue vigorously about where assistance became authorship.

During my years in book publishing, I saw the same ambiguity much closer up. The work of highly individual literary authors could benefit enormously from editors who challenged, cut, reordered, questioned, suggested and occasionally saved a manuscript from its author. Nobody seriously concluded that the author’s name should therefore disappear from the cover. The famously fraught collaboration between author Raymond Carver and editor Gordon Lish shows how difficult that boundary can become, even without technology anywhere in sight.

That history matters because assistance, by itself, does not make a work less yours. Interchangeability does.

“Assistance, by itself, does not make a work less yours. Interchangeability does.”

If virtually anyone could have given the instruction and received essentially the same result, there may not be very much of you in it, however confidently your name sits at the top.

Generative AI makes this old question unusually visible because it can participate almost anywhere in the process. It can check a fact or improve a sentence, but it can also challenge an idea, build an argument, suggest examples, restructure a draft or generate most of the finished prose.

“Did you use AI?” therefore tells us remarkably little about what actually happened creatively.

I’m not sure “How much did you make yourself?” gets us much further either. Even if creative contribution could somehow be expressed as a percentage, would something that is 90% directly produced by you necessarily be more yours than something where your visible contribution was 10%?

One essential observation, connection or creative decision can be the reason an entire work exists. Conversely, you can personally type every word of something while contributing remarkably little that is distinctive.

Authorship is not a percentage. It is the presence of an essential contribution.

“Authorship is not a percentage. It is the presence of an essential contribution.”

So I find a counterfactual test more useful: would this particular thing have existed without you?

Not merely because you typed the prompt that caused an output to appear, but because you brought an observation, perspective, intention, direction, choice or judgement without which what is distinctive about this work would not have emerged.

But there is a catch. If what matters is what you bring, then we also need to keep developing what we are capable of bringing.

Which also brings me back to Ann’s point. If part of the struggle is where we discover what we actually think, handing it over may remove something more valuable than the hours we save. But the reverse can be true too. Assistance can help an idea become more fully what its creator intended. An editor can sharpen it. A collaborator can challenge it. Technology can make something possible that craft alone could not. AI can expose a weakness, connect thoughts, test an argument or find a form for something that might otherwise have remained inside your head.

That leaves honesty. We should be open about assistance where that assistance materially matters, but honesty does not require a purity contest in which every contribution by another person or a machine somehow subtracts from the creator.

And then there is the test that all this creator-focused discussion can make us forget: the audience.

A lovingly handcrafted article can be pointless. Something produced with extensive human and technological assistance can make someone stop, reconsider an assumption, solve a problem or carry an idea into their own work.

The audience doesn’t consume the purity of your process. It encounters the result.

“The audience doesn’t consume the purity of your process. It encounters the result.”

So perhaps the useful test is simpler. Here is the practice I would offer, and it takes about ninety seconds. Before the outline, before the prompt, before the draft, write one sentence:

What do I bring to this that makes it mine and worth the reader’s attention?

Not the topic. Not the angle. The thing that would be missing if someone else, or something else, produced a piece on the same subject. An observation nobody has made out loud. A pattern you noticed because of where you have worked. A position you are willing to be wrong about in public.

If you cannot write that sentence, you do not have a piece yet. You have a topic. Everyone has topics.

If you can, keep it visible while you work and use anything you like. Use the editor. Use the workshop. Use the technology. Use AI wherever it genuinely helps.

Then, at the end, read the finished thing against that one sentence. Is it still true? And is the result worth someone else’s attention?

If it is, your essential and valuable contribution survived the process, however much help it took.

If it quietly stopped being true somewhere around the third draft, you have produced something efficient and interchangeable, and no disclosure statement will fix that.