Everyone got very loud about AI replacing writers, and then the evidence got strangely quiet.
That is the force behind J. Michael Straczynski’s two-part essay on AI storytelling. If a system had produced a genuinely great screenplay, novel, play, or television episode, the industry would not have hidden it in a drawer. It would have been announced, packaged, monetized, debated, denounced, praised, and sold before lunch.
Instead, as Straczynski frames it, there is silence where the story was supposed to be. His first article argues from the inside of craft: AI lacks lived experience, interiority, subtext, and the unconscious pressure that makes a writer need to say something. His second article sharpens the claim into a more technical one: current AI systems do not possess self-directed curiosity. They do not wake up troubled by an unresolved question. They do not leave the assigned path because something in the world refuses to make sense.
He is right about more than people invested in AI want to admit.
“AI doesn’t feel.”
– J. Michael Straczynski, “Silence Where a Story Might Have Been,” May 11, 2026
The mistake is assuming that this settles the entire question. It settles one question beautifully: AI is not an author in the human sense. It does not have a wound it is trying to understand, a memory it cannot digest, a belief it needs to test, or a private reason for pushing one story instead of another.
But writing is not only authorship at the keyboard. Writing is also diagnosis, pressure-testing, comparison, revision, structural mapping, contradiction hunting, and the stubborn work of asking better questions of material that is not yet alive. A system does not need to be the writer to help the writer think. A microscope does not need curiosity to reveal what the scientist is curious about. A map does not need wanderlust to keep the traveler from confusing motion with direction.
The important distinction is between AI as replacement author and AI as externalized story intelligence. The first deserves deep skepticism. The second is where the conversation gets useful.
The machine can imitate a scene before it understands a story
Straczynski’s strongest craft point is about the difference between surface action and subtext. Two people can say the same line in the same room and mean entirely different things depending on distance, posture, timing, prior injury, and what the audience knows that the characters cannot bear to say.
“Space is subtext. Space is information.”
– J. Michael Straczynski, “Silence Where a Story Might Have Been,” May 11, 2026
Dramatica would say this another way. Storytelling is what the audience sees and hears. Storyform is the deeper arrangement of pressures that makes those sights and sounds mean something. A scene can be vivid and still be structurally inert if it does not move the argument of the story.
That is why so much AI-generated fiction feels over-explained and under-dramatized. The characters say what they want because the prompt asked for motivation. They describe pain because pain belongs to the genre. They announce transformation because the pattern suggests a transformation should appear around here. The scene contains recognizable materials, but the pressure beneath them never quite arrives.
A writer does more than decide what a character says. A writer decides what the character cannot say yet, what the other character hears anyway, what the audience understands before either of them does, and why this moment belongs here rather than anywhere else. That is not decoration. That is structure.
Current AI systems can generate prose that resembles prose. They can generate scenes that resemble scenes. What they do badly, on their own, is sustain a meaningful conflict between Perspectives long enough to make the story feel discovered rather than assembled.
That is exactly where a Dramatica lens matters. A complete story is not a stack of plausible scenes. It is a model of conflict seen from different Perspectives: the Objective Story where everyone is dealing with the shared problem, the Main Character where the audience experiences the conflict personally, the Influence Character where another way of seeing things applies pressure, and the Relationship Story where the bond between central points of view changes under stress.
Those Perspectives do not sit politely beside one another. They argue. They contradict. They expose blind spots. They make the writer choose what the story is actually saying about a human problem.
AI tends to flatten that argument when it is asked to be the author. It can still be useful when it is asked to test the argument the author is building.
Curiosity belongs to the writer
Part Two of Straczynski’s argument turns on curiosity. He argues that AI systems are built to respond, not to initiate; to answer questions, not to be seized by them; to remain subordinate to the user’s task rather than generate their own unresolved need to know.
“Silence where a question might have been.”
– J. Michael Straczynski, “Silence Where a Story Would Have Been,” May 12, 2026
That is a devastating description of AI as an autonomous storyteller. A writer begins with a question that will not leave them alone. What happens next? Why did she stay? What would make a good person betray someone? What if the thing saving the family is also the thing destroying it? A system waiting for a prompt does not suffer that pressure.
But the writer does.
That changes the practical question. If the machine has no self-generated curiosity, then the writer must not outsource curiosity to the machine. The writer must bring the question, the taste, the irritation, the unfinished thought, and the willingness to reject an answer that arrives too easily. The writer must decide what the story is trying to prove and which answers feel false.
Used badly, AI becomes a fluency engine. It fills the page with confident material that feels like progress because there are more words than before. Used well, it becomes a surface against which the writer can press the story’s unresolved questions.
Ask it to draft the soul of the work, and it will often give you a competent imitation of whatever pattern it thinks you want. Ask it to identify where the Main Character’s personal conflict contradicts the Objective Story’s shared problem, and suddenly the tool is doing something more useful. Ask it where the Influence Character pressure disappears, where the Relationship Story has become decorative, or where a scene only transfers information instead of forcing reconsideration, and the machine does not need to care in order to reveal something the writer cares about.
That is the difference between replacing the writer and extending the writer’s attention.
The “then this” problem is real
Straczynski also makes a useful distinction between event sequencing and dramatic causality. Bad AI storytelling often moves like a summary: this happens, then this happens, then this happens. The problem is not sequence by itself. The problem is that nothing is meaningfully resisted.
Good story movement has pressure inside it. A choice creates a consequence. A consequence exposes a misconception. A misconception forces another choice. The story advances through opposition, not through itinerary.
This is where Dramatica gives the critique teeth. The Objective Story is not “the plot.” It is the shared field of conflict. The Main Character Throughline is not “the protagonist’s feelings.” It is the personal inequity the audience experiences from the inside. The Influence Character is not “the person who gives advice.” It is the Perspective that makes the Main Character’s way of seeing the problem harder to maintain. The Relationship Story is not “the romance” or “the friendship.” It is the evolving conflict between ways of being in relation to one another.
If those Throughlines are absent, AI will naturally default to scene-shaped prose. It will produce conversations that explain the situation and plot turns that satisfy the prompt. It may even produce moments that sound moving in isolation. But the whole will feel oddly weightless because the story is not carrying an integrated argument.
That does not mean AI cannot participate in story development. It means the writer needs a structural standard higher than “make this more dramatic.” The machine can help pressure-test a Storyform when the Storyform exists as a constraint. It can compare the intended argument to the scene on the page. It can notice repetition, missing opposition, collapsed Perspectives, and false turns where a character says the theme instead of dramatizing it.
The craft burden does not vanish. It becomes more explicit.
The machine does not need an unconscious to reveal a pattern
Straczynski is also right to defend the unconscious part of writing. Stories often arrive from somewhere below the well-lit room of conscious intention. Writers discover things in the act of writing that they could not have outlined in advance. Some line appears on the page and the writer recognizes, with a little alarm, that the story knew something before the writer did.
AI does not have that experience. It has no childhood, no grief metabolized into metaphor, no body remembering light, touch, shame, weather, or fear. It can recombine descriptions of those things. It cannot remember them.
Still, the unconscious is not the only part of writing. Craft is the conversation between unconscious material and conscious form. A writer may receive the image from the dark, but then they revise it, test it, cut it, place it, and ask what work it is doing in the whole.
That conscious work is exactly where a structured tool can help.
A Dramatica analysis does not replace the writer’s unconscious. It gives the writer a way to examine the pressure the unconscious may already be arranging. Why does this image keep returning? Why does this relationship feel more important than the plot summary admits? Why does the ending feel technically resolved but emotionally unearned? Those are questions a writer can bring to AI, and they are far more valuable than “write me a scene where two people argue.”
The machine is not the source of meaning. It can still become an instrument for finding where meaning has failed to take form.
The real danger is not AI writing badly
The obvious danger is that studios and platforms will use AI to generate cheap, mediocre material and call it story. That danger is real, but it is not new in spirit. The industry has always had incentives to confuse volume with value, familiarity with quality, and marketable premise with dramatic meaning.
The deeper danger is that writers will internalize the same confusion. They will use AI to avoid the discomfort Straczynski is defending: the unresolved question, the ugly draft, the private insistence that the easy answer is lying. They will let fluent output talk them out of the slower work of finding the story’s actual argument.
That is why the right answer is not fear. Fear makes AI feel more powerful than it is. The right answer is standards.
A writer using AI should be more demanding, not less. What is the Objective Story conflict? Where is the Main Character’s personal pressure distinct from the shared problem? What does the Influence Character force the Main Character to reconsider? How does the Relationship Story change because of the conflict, rather than merely decorate it? Which Storypoint is this scene actually illustrating? Where does the draft collapse Storyform into storytelling?
Those questions do not make writing mechanical. They keep the machine from making the writing merely fluent.
What AI can know, and what it cannot want
Straczynski’s conclusion is strongest when aimed at autonomy. Current AI systems do not decide they have something to say. They do not originate a need to disturb the world. They do not become sand in the machinery because something inside them refuses the machinery’s terms.
That matters.
But a writer does not need the tool to want. The writer needs the tool to serve the wanting without replacing it. The writer needs the machine to stay subordinate to the human question while being sharp enough to expose weak answers.
That is a narrower claim than “AI will write great stories.” It is also a more useful one.
AI does not know writing the way a writer knows writing. It does not know it through the body, through memory, through fear, through love, through the private cost of saying the thing that finally has to be said. It knows patterns, probabilities, and relationships between expressions. Left alone, that is not enough.
Joined to a writer’s curiosity, it can still become part of the work.
The missing masterpiece remains a structural clue. We should not confuse generated pages with authored meaning. We should not pretend a prompt has the same moral or imaginative burden as a writer’s unresolved question. But we also should not throw away the possibility that a tool without curiosity can help a curious writer see the story more clearly.
The silence is real.
The writer is the one who has to break it.