Day 17: Use AI for Creative Work

The Concept

There is a persistent anxiety that AI and creativity are in opposition — that using AI on creative work means the work is no longer really yours. This concern is understandable, but it misidentifies what the actual tension is. The problem is not AI and creativity. The problem is using AI as a substitute for creative judgment rather than as support for it.

When AI writes something and you accept it wholesale, the work is not yours in any meaningful sense. Your perspective, your taste, your understanding of what matters in this particular piece — none of those are present. But that is a misuse of the tool, not an inherent property of it. Used differently, AI can keep creative work moving when it stalls, generate options you can react to and shape, give feedback that would otherwise require another person, and do the first-draft labour so you can focus on the harder work of editing and judgment. The best creative use of AI is not authorship. It is collaboration.

What AI cannot do is have taste. It does not know what matters to you about this specific piece, what you are trying to say that no one else would say it quite this way, or why a particular phrase is exactly wrong even if it is technically correct. Those are things only you can bring. The productive frame is this: AI keeps the work moving; you make the decisions about what is worth keeping.

What creative block actually is

Creative block is rarely a shortage of ideas. It is more often a specific stuck point — a transition that is not working, an opening that feels wrong, a structure that is not holding together, a piece of feedback that has lodged in your head and made the whole thing feel impossible. The problem is usually more specific than it appears, and naming it specifically is most of the work.

This is why today's prompt asks you to describe not just the project but exactly what you are stuck on or unsatisfied with. "I do not know where to go from here" is not a useful brief for a creative collaborator. "The second section works but it disconnects from the opening and I do not know how to bridge them" is. The more specifically you can describe the problem, the more useful the response will be — and often, the act of articulating the specific problem clearly enough to describe it starts to suggest the solution before AI has responded at all.

The four modes of AI as creative partner

There are four distinct ways AI is useful in creative work, and they are different enough that it helps to know which one you need. Ideation means generating raw options to react to — not to use directly, but to break out of the loop of your own fixed thinking. Drafting means producing a working version so you can edit rather than originate from nothing, which is often significantly easier. Critique means identifying what is not working and why — honest feedback that most human collaborators are reluctant to give. Expansion means taking a fragment, an image, a sentence, or a rough idea and developing it further, so you can see where it wants to go before you commit to it.

Most creative sessions benefit from more than one of these modes in sequence. You might start with ideation to find a direction, ask for a draft in that direction, then ask for critique of the draft, then expand the strongest element. The session today moves through a version of this arc — assessment, directions, then execution of your chosen direction. This is deliberately structured as a collaboration where you make the choices rather than just accepting whatever AI produces first.

What AI cannot do for your creative work

Knowing the limits is as important as knowing the capabilities. AI produces work that is statistically competent — it draws on patterns across an enormous body of text and tends toward the centre of what has worked before. This makes it reliable and makes the output feel polished. It also makes it tendency-prone: it defaults to the expected structure, the familiar metaphor, the safe resolution.

The most distinctive creative work lives at the edges of those patterns, not the centre. Your voice — if it is specific enough to be genuinely yours — will often be the thing AI smooths away in pursuit of fluency. This is why the power tip asks you to compare versions in your voice against its opposite: the contrast reveals what your voice actually is, which is harder to identify when you are only looking at one version. ChatGPT, Claude, and Gemini all have meaningfully different tendencies in tone and style, and it is worth trying the same creative prompt across more than one to see which output is easiest to shape into something that sounds like you.