Story Math Guy Gets a Calculator
Screen scribblers who understand how LLMs read structure will write scripts that survive AI filtration. Those who don't may never know what stopped them.
I’m a story math guy. That’s a confession, not a brag.
Thirty years ago I sat in a Robert McKee seminar and watched him map a screenplay onto a whiteboard like an equation. Something clicked that never unclicked. Structure went from vague creative instinct to near-obsession. Every script I’ve read, every scene I’ve scribbled, I’m running the numbers. Act breaks. Inciting incidents. All Is Lost moments. The full diagnostic.
The catch: knowing math doesn’t mean you can always do the math.
Understanding structure and executing are two entirely different muscles. A scribbler can hold the McKee framework in their head, the Blake Snyder beat sheet in one hand, Truby’s genre mechanics in the other, and still get lost inside a draft. Once you’re inside a scene, inside a character, inside the heat of a first draft, structural X-ray vision is the first thing to go. You know what a great script looks like from thirty thousand feet. Seeing it from where you’re standing is a different problem entirely.
THE TOOL I DIDN’T KNOW I NEEDED
Then the current generation of LLMs arrived. ChatGPT, Gemini, Claude. I started running my scripts through them.
Not for dialogue polish. Not for coverage. For structure analysis. Deep structural analysis. Asking a model to read a full screenplay through the lens of McKee, Snyder, Truby, or my own Modular Story Method and come back with a real read. What’s working. What’s broken. Where my Reel structure holds and where it caves.
The results have been genuinely awesome.
These models can hold an entire feature simultaneously and apply multiple structural frameworks in a single pass. The feedback is detailed, grounded, and calibrated to whichever theoretical framework you invoke. Ask for a McKee read and you get a McKee read. Ask for a Truby character web and you get one.
At the creative level, this freed me up to be… creative.
PERMISSION TO MAKE A MESS
Early drafts should be messy. Every experienced scribbler knows this. Follow a scene into the dark. Let a character blah-blah-blah. A subplot go sideways. That’s how you find unexpected beats that make a script feel alive instead of assembled.
I’ve always known that. Still found myself pulling up short in a first draft because structure anxiety. Second-guessing if a scene ran too long, midpoint was landing in the right neighborhood, whether I was buried in Reel Three with no clear path out.
Having a structural analyst on call to read the draft and run diagnostics changes that relationship. Go deep in a scene. Let it breathe or let it explode. Go long on purpose. The structural pass is waiting on the other side of the draft, and it’s not going to miss anything. That confidence is a creative asset.
The math gets checked later. The free ride scribbling happens now.
THE OTHER SIDE OF THE CALCULATION
Companies already use AI to screen job applicants. Resumes go in, an algorithm filters them out, and a human being never looks at half of them. The applicant never knows. The company saves time. Careers shredded sans a single set of human eyes touching the work.
Scripts are documents too.
I’m not speculating this might happen someday. It’s already happening somewhere. Studios, streamers, and production companies are flooded with submissions and hungry for any system that reduces intake. Run the script through a structural model. Doesn’t clear the McKee threshold? Pass. Midpoint isn’t landing in the right window? Pass. Doesn’t pattern-match against whatever internal criteria someone decided matters? Pass, pass, pass.
Good or bad? I genuinely don’t know. A well-calibrated model with sound criteria might surface stronger material faster than a stack of overworked readers. It might catch things a tired reader skips on a Friday afternoon.
The harder question: criteria set by whom, optimizing for what? Structural conformity is a different thing from a great script. Some of the best scripts ever scribbled would’ve failed a rigid beat-sheet audit in the first draft. A model that can’t distinguish between breaking the rules badly and breaking the rules brilliantly becomes a filter and a ceiling at the same time.
Screen scribblers should know it’s coming.
Run your current draft through an LLM with a specific framework prompt (”Analyze this script using Blake Snyder’s beat sheet. Identify where each beat lands and where it’s absent.”). Treat the output as a structural X-ray, not a verdict.
In your next first draft, schedule a structural pass for after you finish. Write a note atop page one: “Check the math later.” You have permission to follow the muse down the rabbit hole.
Research how studios and streamers at your target tier currently handle script intake volume. Ask your reps directly. The more you know about how material is being screened, the better you can position your submission.
Those same structural tools that make your drafts better are probably the same tools that may be screening your work before a human reads it. No matter what, always be scribbling!