■ SAFE RISK ■ Creative Arts
Screenwriting survives, but the ladder into it is being sawn off. Generative tools are competent at coverage, outlines and formulaic pages, which is exactly the work that used to pay junior writers while they learned.
“AI writes scripts. Hollywood ignores both equally.”
Our AI replacement risk score — how we score jobs
The job is far less about typing than outsiders assume. A working screenwriter pitches, breaks story on a whiteboard, writes an outline that survives studio notes, drafts, then rewrites against contradictory feedback from producers, network executives, a director and eventually an actor who wants their character to be more likeable. In television it means a writers' room: pitching beats, taking notes, punching up other people's scenes, and increasingly being on set for production rewrites. Credit, residuals and guild rules structure the whole economy.
Language models already handle the parts of that process that are pattern-shaped. Loglines, one-pagers, alternative scene versions, dialogue passes in a given register, format cleanup, and studio coverage of submitted scripts are all things AI does adequately and instantly. Development executives use it to triage submissions. Producers use it to generate comparable-title analyses and pitch decks. For low-budget genre content — the direct-to-streaming thriller, the procedural episode built on a rigid template — the marginal quality gap is small enough that budgets are being tested against it.
What holds is structural as much as artistic. The 2023 labour agreements put explicit guardrails around AI-generated material and credit, and studios face copyright exposure on wholly machine-written work. Artistically, models are weak precisely where scripts live or die: sustaining a dramatic argument across two hours, making a character's contradiction feel inevitable, knowing which scene to cut. And the writer's real product in a room is judgement under notes, plus the ability to be blamed. Our risk score of 22 reflects a profession that persists at the top while its entry rungs and mid-tier volume work erode fast.
Automatability: our editorial assessment of current and near-term AI capability
The squeeze started with the 2023 strikes and is ongoing. Through the late 2020s expect smaller writers' rooms, shorter contracts, and template-driven content increasingly drafted by machine with a human polish credit. By 2030 the mid-tier working writer is the endangered category. Top-tier writers with distinctive voices and showrunner authority remain in demand past 2040 — assuming anyone is still training successors.
It can write a competent one for a rigid genre template, and it writes coverage, loglines and alternative scene versions very well. What it does poorly is sustain a dramatic argument across two hours and make a character's contradictions land. Ask it for act three and you generally get plausible beats that nobody would pay to watch.
Worth pursuing, harder to enter. The jobs that used to pay while you learned — staff writer slots, coverage work, low-budget assignments — are the most exposed. If you go in, aim to write things you own and to develop toward showrunning or directing, because leverage now sits with people who control projects, not pages.
The 2023 agreements set limits on studios requiring writers to use AI material and on machine-generated text receiving credit or reducing writer compensation. Those protections matter enormously in practice, since they keep human authorship attached to credit and residuals. They are contractual rather than permanent, so the terms of each renegotiation are worth watching closely.
Distinctiveness and authority. A recognisable voice, deep specificity about a world you actually know, and the ability to run a room or a set. Generic competence is the commodity being automated. The writers with security are the ones a producer hires because only they can write that particular thing, not because pages are needed by Friday.