Is a Prompt a Piece of Writing?

The medium is the prompt.

Marshall McLuhan observed that the medium through which a message travels fundamentally shapes the message itself, regardless of the content. Television didn't change culture because of what was broadcast; it changed culture because television exists as a medium—because it requires passive consumption in front of a glowing screen, because it compresses complex ideas into digestible sequences, because it privileges the visual and the immediate. The same news story told on television has a different effect than the same story read in a newspaper, not because the words differ but because the medium has already made its argument about how truth should be consumed.

The arrival of AI writing tools suggests we're watching a similar transformation occur in how writing itself works—except this time, we're not yet accustomed to calling it a transformation at all. We still use the word "writing" as though nothing has changed. A user composes a prompt and receives text. The activity looks like it could be writing. But the medium has already shifted, and with it, something fundamental about the creative act.

What Used to Be Writing

For centuries, writing meant the composition of sentences. The writer faced a blank page—first with pen, then with typewriter, then with word processor—and generated language. The physical act of writing was also the act of thinking. The two were inseparable. You couldn't write a sentence until you'd thought it. You couldn't revise it until you'd already composed it. The constraint was built into the medium: one sentence at a time, in linear order, with the writer as the producer of every word.

Even as tools changed—pens became faster, typewriters more reliable, computers more flexible—the essential act remained: a human mind producing human language, one unit at a time. Editing existed, certainly. Writers revised. But revision still meant rewriting: crossing out the old sentence and generating a new one. The writer remained the compositor, the person responsible for every linguistic choice.

What's changing isn't the writer's access to tools, or even the speed of writing. It's the location of the generative act. Writing is no longer the primary creative work. Description is.

The Shift in the Creative Act

When you write a prompt, you're not composing sentences. You're composing instructions for someone (or something) else to generate sentences. The creative work has shifted upstream, toward specification and direction. Instead of "I will write a paragraph that explores the tension between convenience and privacy," the work becomes "describe the tension between convenience and privacy in a conversational tone, as if explaining it to someone skeptical."

This is a real difference, and it matters. It's the difference between doing a thing and describing how a thing should be done. Most people have never felt that these are the same activity. A restaurant owner doesn't cook every meal. A film director doesn't perform every scene. A conductor doesn't play every instrument. Yet we recognize direction, curation, and specification as creative acts in these fields. The question isn't whether directing requires skill—it obviously does—but why writing specifically has been conflated with the execution rather than the direction.

Prompt engineering, for all its awkward terminology, involves genuine creative work: knowing what good language looks like, understanding how to describe what you want, recognizing when the output is close but not quite right, diagnosing what went wrong, and iterating toward something better. This requires taste. It requires judgment. It requires literacy in a deep sense—not just the ability to read, but the ability to recognize quality and articulate why something works or doesn't.

A person who prompts well has to be able to read well. They have to understand the structures of good writing, the rhythms of language, the difference between a cliché and a fresh formulation. Otherwise, how would they know if the output is any good? The skill hasn't disappeared. It's been reoriented.

The Spectrum of Creative Direction

The film director's contribution is invisible to the audience. What they see is the film—the lighting, the performance, the shot composition—not the director's instructions or decisions. A great director is someone who knows what they want, understands the possibilities of the medium, communicates clearly to collaborators, and recognizes quality work when they see it. None of these things require the director to operate the camera, memorize lines, or physically construct the set.

Music producers work in a similar mode. They shape the sound, make decisions about arrangement and tone, guide the musicians, and sometimes request that something be redone twenty times until it's right. They don't play every instrument on the album. Often they play none. Yet the album is clearly their creative work. You can recognize a producer's sensibility across different artists.

Architecture offers another parallel. An architect designs the building—the spatial relationships, the proportions, the materials, the light—without laying a single brick or hammering a single nail. The skill is in the vision and the specification, not in the manual execution.

Where does prompting fall on this spectrum? It resembles all three in some ways. It shares the invisibility of directing (the output looks like writing, not like instructions). It shares the iterative refinement of producing (try it, listen, adjust, try again). It shares the specification work of architecture (describing constraints, requirements, desired outcomes). Perhaps the closest parallel is editing—the work of recognizing what's good, spotting what's wrong, and knowing how to fix it without necessarily writing the original text yourself.

The discomfort with calling prompting "creative work" might stem from how thoroughly it masks its own process. A film is clearly directed. A building clearly expresses an architect's vision. But a piece of text generated by AI and refined through prompting just looks like a piece of text. The medium has made itself invisible.

The Question of Literacy

Here's an interesting constraint that rarely gets mentioned: you can't prompt well without being a good reader. You need to understand what language does. You need to recognize when something is vague versus precise, when it's genuine versus formulaic, when it's earned versus cheap. These are the things you learn from reading widely and attentively.

This creates an odd gate. As AI tools become more capable and more accessible, the ability to use them well may become less accessible, not more. A person who's never read good writing, who doesn't have an internal sense of what quality looks like, will generate mediocre prompts and receive mediocre text. The AI doesn't rescue them from their lack of taste; it just produces more of it, faster. The skill isn't eliminated. It's been moved to a different stage of the process, but it's still required.

This suggests something worth noticing: the democratization of access to writing tools—the ability to produce text without writing it yourself—doesn't democratize the ability to produce good writing. Good writing still requires the same attentiveness, the same judgement, the same accumulation of reading. It's just applied at a different point in the process.

Visibility and Invisibility

Editing is rarely credited as creative work, even though editors make crucial decisions about what remains, what's cut, how ideas connect, and what matters. But editing is invisible. The reader doesn't see the editor's pencil marks. They see a finished text. Editing is recognized as creative work primarily by other people who edit, and by writers who understand how much has been shaped by editorial hands.

The same invisibility may befriend or haunt prompt engineering, depending on how the field develops. If prompting becomes an explicit, credited part of the writing process—if bylines change to reflect that a piece was written in collaboration with AI and shaped through prompting—then it will gain visibility and recognition. If it remains hidden, credited wholly to the AI or not credited at all, it may follow editing into a kind of invisible craft. Valuable but unrecognized. Influential but unnamed.

It's worth noting that production work—the engineering, mixing, and post-processing that shapes a recorded song—is almost entirely invisible to the listener. Most people who listen to music have no idea how much of what they're hearing is the result of producers, engineers, and sound designers. The invisibility hasn't prevented these from being recognized as skilled, creative professions. But it has meant that the average listener significantly underestimates how much work goes into the final product.

The Vocabulary Problem

Much of the current confusion might stem from a vocabulary that hasn't caught up to the reality of what's happening. The word "writing" has always meant, primarily, the act of composing sentences. It carries the weight of intentionality, of authorship, of direct creation. To call prompting "writing" feels like it's claiming too much, because it isn't sentence composition. To deny it the word "writing" feels like it's claiming too little, because real creative work is occurring.

But the language available—directing, editing, curating, designing—all come from other fields and don't quite fit. "Prompt engineering" is an attempt to borrow from another domain (engineering, i.e., building systems), but it sounds clunky because it is clunky. We're trying to describe something new using old words.

What might be happening is that "writing" itself has been revealed as a word that covered too much ground. It meant simultaneously: the generation of language, the selection of language, the arrangement of language, the expression of thought through language. Now that it's possible to separate these activities—to direct the generation of language without generating it yourself, to select and arrange without composing—we're noticing that the word was doing too much work.

This happened before in other fields. "Composing" once meant writing music by hand, literally composing it on paper. Eventually it meant the conceptual work of creating music, separate from the manual act of notation. A composer could now work in a digital audio workstation, sculpting sound without ever writing a note on paper. The word expanded to accommodate new tools without losing meaning, because the core creative act—the conceptual work of making musical decisions—remained the same.

The question is whether "writing" will expand similarly, or whether we need new language entirely. Will we say "she wrote this essay with AI assistance" the way we say "she composed this symphony using a digital workstation"? Or will new terminology emerge—"prompted," "directed," "crafted," something else—that more precisely describes the activity?

What Remains Unchanged

One thing worth holding steady: the human work of deciding what to say remains essential. AI can generate language, but it cannot decide which ideas matter, what's worth communicating, what story needs to be told, what perspective has been overlooked. Those are human decisions. They come from judgment, from knowledge, from caring about something enough to articulate it.

A prompt that tries to outsource even this—that asks an AI to "write something interesting about climate change" without the person knowing what they actually want to communicate—will produce mediocre work. The AI will generate something that sounds like an essay, but it won't have the animating force of genuine thought. It won't have stakes.

This is why the question "is a prompt a piece of writing?" might be less interesting than the question it unintentionally raises: what is writing actually for? If writing is about expressing thought, making arguments, sharing perspective, then the medium through which that happens—whether it's pen on paper, fingers on keyboard, or instructions to an AI—matters less than whether the thought itself is genuine. If writing is about the physical act of composing sentences, then prompting is something else entirely. But if writing is fundamentally about creating meaning through language, then what you call the activity becomes almost beside the point. What matters is whether the meaning is real.

The medium will eventually settle into a new normal. We'll develop conventions for credit and attribution. We'll find vocabulary that fits. The question of whether prompting is "really" writing will resolve itself, probably not through definition but through practice—by writers, or prompt engineers, or whatever they come to be called, simply doing the work well enough and often enough that the category becomes self-evident.

Until then, there's value in sitting with the uncertainty. Not as a problem to solve, but as a lens. The confusion about prompting reveals something true about writing that was always there: it was never only about producing words. It was about shaping them, selecting them, arranging them, knowing which ones matter. The tools have changed enough to make that visible. The rest is just learning to see what was already happening.