What does AI mean for the future of screenwriting in Hollywood?

Since its launch in November 2022, hundreds of millions of people have used ChatGPT to write wedding toasts, college essays, apology texts, bad jokes and even worse poetry. Billy Ray — Oscar-nominated screenwriter and unapologetic human being — is not one of them.

Ray, whose writing credits include “Shattered Glass,” “Captain Phillips” and “The Hunger Games” (that now-iconic Nicole Kidman AMC ad with “Somehow heartbreak feels good in a place like this” is also his), has never even opened the ChatGPT site. Not to fix a clunky line. Not to win a bar trivia argument. Not to figure out what to do with the leftovers in his fridge.

To Ray, generative AI — already creeping into every corner of Hollywood, from script development and previsualization to casting and marketing — isn’t just another tool for creatives, like Final Draft or a Steadicam. It’s an existential threat, “a cancer masquerading as a profit center,” he says, eroding not just storytelling but the storyteller.

“My level of impostor syndrome, neuroticism and guilt is high enough while I’m working my ass off,” Ray says by phone, his voice equal parts weariness and outrage. “There’s no way I’d make myself feel worse by letting a machine do my writing for me. Zero interest.”

When AI hype and fear first swept through the entertainment industry, screenwriters quickly found themselves on the front lines — and the picket lines. During the 2023 strike, the Writers Guild won precedent-setting contract language: Studios can’t require writers to use AI, and anything generated by it can’t be considered “literary” or “source” material. Writers are free to use AI if they choose — but only with the studio’s approval, and under rules that protect credit, authorship and intellectual property.

The agreement was hailed as a landmark: the first real attempt to set limits on a fast-moving, poorly understood technology. But for Ray, those protections don’t go far enough. The tools are getting exponentially more powerful, he says, and adoption is already happening quietly, behind closed doors. “What I’m hearing anecdotally is that studios and streamers are putting more and more time and energy into exploring what AI can do for them,” he says. “The result will inevitably be chaos, bad movies, bad TV shows and a lot of people out of work.”

A longtime WGA member and former co-chair of the guild’s negotiating committee, Ray says his level of alarm is greater now than it was during the strike. That alarm is shared by many in a business where thousands of writers already hustle from project to project and where the prospect of studios using AI to shrink writers’ rooms, eliminate junior positions or even generate first drafts has added new urgency to the debate. The anxiety is not theoretical: According to the Writers Guild’s 2024 financial report, the number of members reporting earnings fell by nearly 10% from the prior year — and by more than 24% compared with 2022.

With AI technology leaping ahead at algorithmic speed, Ray is urging the union to move faster. “We need to put firewalls in place before the next round of negotiations,” he says. “That’s going to be necessary.” Though he still refuses to touch AI himself, he isn’t trying to police his peers. “I’m not telling writers they can’t use it,” he says. “But the public has a right to know when they’re watching something written by a human being. And I think they want to know.”

As he speaks about the potentially apocalyptic implications — for Hollywood and for humanity at large — Ray sounds both incredulous and downright scared. “We as a species have a limited window to get control of AI and put guardrails around it, but we as writers have an even more limited window,” he says, his frustration rising. “It makes no sense. If all Hollywood has to offer is a bunch of warmed-over AI bulls—, why would someone turn away from TikTok or YouTube?”

Across the film industry, AI has begun to permeate nearly every stage of the production pipeline: helping directors visualize scenes before they’re shot, cloning actors’ voices for foreign dubs and assisting editors in assembling early rough cuts. But of all the creative roles AI is taking on, writing may be the most controversial — and at risk. Actors can fight to protect their likeness. Directors still need a crew to execute their vision. Writers often work in solitude, in front of a blinking cursor, the very place AI is now starting to intrude.

Unlike a human writer, a “large language model” — the technical term for AI systems like ChatGPT that are trained on massive amounts of text — doesn’t grasp plot, motivation or theme in any true sense. It can stitch together scenes that feel plausible on the surface — a couple arguing in the rain, a soldier saying goodbye before heading off to war — and can sometimes even surprise you with a turn of phrase or an unexpected twist. What it can’t do is understand what those moments mean or shape them to make an audience feel something lasting.

To be fair, that might also describe more than a few human-written screenplays. And Hollywood has long flirted with the idea of turning writing into a system. In the 1970s and ’80s, a cottage industry blossomed around screenwriting formulas — from Syd Field’s three-act paradigm to Robert McKee’s guru lectures and the ever-resilient “Save the Cat” beat sheet. Storytelling became something you could learn, teach and sell, often quite successfully. (See: “Adaptation,” which turned the whole idea into a punch line.)

The difference now is that the machine isn’t just applying the formula — it’s trying to do the writing itself.

The late critic Roger Ebert famously called cinema a “machine that generates empathy.” But as generative AI takes on more of the creative process, a deeper question emerges: What does it mean when stories are shaped by a system that, for now at least, can’t feel — and whose users may not need it to?

“I’m scared of it,” says writer-director Todd Haynes, whose films, including “Safe,” “Far From Heaven” and “May December” (scripted by Samy Burch and Oscar-nominated for its screenplay), explore all-too-human themes of identity, sexuality and social constraint. “Creativity is born out of mistakes, obfuscation, fumblings, desire — things that computer technology can never replace.”

Who’s holding the pen now?

In the fall of 2023, two weeks after the writers’ strike ended — and with the actors still on the picket lines — Hollywood’s first “AI on the Lot” conference opened its doors downtown, a bold show of tech optimism in the midst of labor upheaval.

“We thought we were going to have picketers out front,” says organizer Todd Terrazas, who founded the nonprofit AI LA and co-founded FBRC.ai, an AI-driven venture studio launched in 2023 to bridge creativity and technology. “But sure enough, there was none of that. Six hundred people showed up and really leaned into how this technology could expand the industry and support everyone.”

Two years later, by this spring’s edition, the event had expanded and moved to the Culver Theater, drawing nearly twice the crowd with 1,100 attendees, a mix of indie filmmakers, startup founders and tech execs from Google, Amazon, Nvidia and OpenAI. The vibe was more techno-optimism than hand-wringing. But the stage was still missing something: writers.

“AI is a very touchy subject, especially for writers, because it’s so personal,” says Terrazas, who has emerged as a key connector between Hollywood creatives and the fast-expanding AI tech world. “Even though they’re experimenting with large language models to help organize thoughts or explore new characters or ideas, at the end of the day they want to be known as the one who actually wrote everything, like, ‘This was 95% me.’”

In a legal gray zone where authorship is murky and copyright law hasn’t caught up with technology, what’s at stake isn’t just credit or ego but ownership. “Writers are walking a tight line,” Terrazas adds. “They want to be very careful that they’re showing their work, documenting their process, so they can obtain copyright and stay in bounds with the studios and the guilds.”

While many writers worry about AI encroaching on authorship, a wave of startups sees opportunity — not to replace writers, they say, but to streamline the clutter around them. Amit Gupta, who co-founded the AI writing tool Sudowrite in 2020, began development by interviewing screenwriters about what they actually needed. The complaints he heard were often surprisingly mundane. “They’d say they dreaded writing the logline, the one-page treatment, the three-page treatment, once the screenplay was done,” he says. These were exactly the kind of mundane tasks his AI platform could automate.

Some studio executives may already be imagining a future with fewer writers, a field that’s historically one of the most developmentally expensive and unpredictable parts of making a movie. Since the spec script boom of the 1990s, when writers like Shane Black (“Lethal Weapon”) commanded multimillion-dollar paydays, screenwriting has carried a uniquely speculative price structure for work that’s often unproven. Robert Altman’s 1992 film “The Player” famously centered on a murder of a screenwriter, satirizing the industry’s long love-hate relationship with the written word.

But Gupta pushes back on that vision. He says AI is far from being able to write a good movie on its own — at least not yet. “You could watch it,” he says. “But you’re not going to like watching it.” What excites him more is the potential for co-creation, humans still driving the process with machines supporting rather than replacing them.

“I think that’s where the skill of the writer really comes in,” Gupta says. “If I go to ChatGPT and say, ‘Write me a short story about someone in L.A., reading an article on the film industry and hanging out with their dog,’ it will give me something generic, because that’s what the model is. But the prompt actually matters a lot. The people who are really good with this stuff are kind of mind-blowing.”

What exactly qualifies as a mind-blowing prompt is not entirely clear, but Gupta believes developing that kind of conjuring ability will become as essential as programming or writing itself. “Once you get adept at handling it with precision, it feels like a tool — not something doing the work for you. That’s going to be a very real skill set in the future.”

A fault line in the craft

If Gupta sees AI as a tireless, ego-free assistant for the grunt work of writing, others have leaned in further, treating it more like a virtual writers’ room — riffing on scenes, dialogue and structure — or even an uncredited auteur behind the curtain.

In January, Paul Schrader, the writer of “Taxi Driver” and his Oscar-nominated “First Reformed,” known for his psychologically intense, deeply human portraits of guilt and faith, caused a stir by praising ChatGPT online as a kind of creative oracle. After asking the AI chatbot to generate movie ideas in the style of various auteurs — Paul Thomas Anderson, Ingmar Bergman, Martin Scorsese, Quentin Tarantino, David Lynch — he was floored.

“I’M STUNNED,” Schrader wrote on Facebook. “Every idea ChatGPT came up with (in a few seconds) was good. And original. And fleshed out. Why should writers sit around for months searching for a good idea when AI can provide one in seconds?”

In another post, Schrader said the model instantly gave feedback on an old script he had written years earlier “as good or better than I’ve ever received from a film executive.” The experience, he said, left him certain that AI was the superior writer: “This is an existential moment, akin to what Kasparov felt in 1997 when he realized Deep Blue was going to beat him at chess.”

The backlash came fast. “Paul, is everything OK?” one commenter wrote. Ever the provocateur, Schrader showed no sign of backing down, gleefully sharing AI-generated images, including one in which he is seen conjuring characters with a magic pen.

Asked about Schrader’s AI enthusiasm, Billy Ray offered a pointed retort: “I have enormous regard for his career and for the work he’s done — he wrote ‘Taxi Driver,’ for God’s sake. But I don’t see how that’s helpful.”

Filmmaker Bong Joon Ho takes a more humanist tack. The writer and director of genre-scrambling films like “Snowpiercer,” “Parasite” and “Mickey 17” — a mix of original stories and literary adaptations — acknowledges AI’s value as a subject for sci-fi but doubts its capacity to tell stories with real depth or irony.

“We’ve seen from films like ‘The Terminator’ that AI can be a great source of drama, and we can create a lot of stories around it,” he told The Times earlier this year. “But I honestly don’t think AI programs will write a fun story about themselves. I feel like I am a better writer for those stories.”

Others worry that as AI becomes embedded in Hollywood, even human-written work will start to sound like the data it was trained on: smoother, safer, harder to tell apart. Roma Murphy, a young writer and story artist who serves as co-chair of the Animation Guild’s AI Committee — one of several new working groups formed in the wake of the 2023 strikes — describes herself as “a bit of a purist.” Like many, she is concerned about the exploitation of unlicensed material — the countless film and TV scripts that may have been scraped to train AI now being pitched back to the industry.

“I’m certainly not going to type my own ideas into the platform and just give them to it to train with,” Murphy says.

“Look, it’s much better than it was in 2022 — it can at least generate a document,” she says. “But I have yet to meet someone who was still thinking about their AI screenplay more than 12 hours later. People engage with art because they want to see some truth about humanity reflected back to them, and AI is never going to reflect a new truth. Nothing I’ve seen generated feels like anything more than a cheap party trick.”

At film schools, where many future screenwriters get their start, the question of when and how to introduce AI has become its own point of debate. USC’s School of Cinematic Arts, one of the nation’s most influential film programs and alma mater to filmmakers like George Lucas and Ryan Coogler, now offers courses like “Art in Post-Reality: Critical and Creative Approaches to AI” and “AI Magic: Revolutionizing Media and Workplace Creativity.”

According to Holly Willis, chair of the Media Arts and Practice Division and co-director of USC’s new AI for Media & Storytelling initiative, student attitudes toward the technology vary widely.

Willis points to the first AI-focused class USC offered in 2023, launched during the height of the strikes. “The students came in very wary,” she recalls. “They weren’t even telling their friends they were in the class — out of fear of reprisal. Some were saying, ‘Why am I paying for this education when you could just prompt and make a film?’”

But even as the school has integrated AI across a range of filmmaking disciplines, one area remains off-limits: screenwriting. “We’ve been very intentional about protecting that early phase when students are still figuring out who they are as writers,” Willis says. “They need space to develop their own voice and stories before turning to tools like this. Understanding how the technology works is important, but so is safeguarding that vulnerable creative moment.”

Writing in the gray zone

Oscar Sharp arrived in the future a little earlier than most in Hollywood. Nearly a decade ago, the British filmmaker set out to see what would happen if a computer tried to reverse engineer a science-fiction screenplay using nothing but genre tropes. “My writer friends joked with me, quite reasonably, ‘You hate writing so much that you’d build a machine to do it for you, even if it’s really bad,’ ” Sharp says dryly. “There’s some truth to that.” But his real aim, he says, was to see what the genre’s average story looked like when processed by an early AI model.

The result was 2016’s “Sunspring,” a nine-minute short scripted by a custom-built neural net —dubbed Benjamin — trained on dozens of sci-fi films, mostly from the ‘80s and ‘90s. Created with AI researcher Ross Goodwin, the film stitched together a surreal, dystopian mashup of familiar — if often nonsensical — beats, delivered by Thomas Middleditch and the cast with deadpan sincerity. A year later, Sharp followed with “It’s No Game,” a short set during a fictional AI-inspired writers’ strike, featuring David Hasselhoff performing AI-generated dialogue distilled from his past work in shows like “Knight Rider” and “Baywatch.”

In truth, Sharp’s AI experimentation was less about replacing writing than exposing the underlying code of storytelling itself. “It’s looking for statistical patterns — like, what similar things have happened before,” Sharp, a discursive and reflective speaker, says on a video call. “But those patterns were themselves created by feedback loops. So if you train something on them, you’re just deepening those same loops.”

Today, Sharp is still experimenting with AI but only very occasionally and never to outsource the work. He’s more cautious about how publicly he engages. “I’ve kept a pretty low profile about this sort of stuff,” he says, aware of how charged the debate has become within the industry. He suspects he’s not alone. “Far more people are probably using it than are comfortable saying they are,” he says. “It’s widely available and extremely effective for those who employ it in particular ways. It would be very weird if that wasn’t happening.”

Sometimes he uses it not as a collaborator but as a kind of negative muse, a foil to push against.

“I’ve asked it to write a really bad scene — just let it go kind of mad,” he says. “Then I rewrite every damn word, often doing the opposite of what it gave me. It’s actually pretty adaptive for a writing process. Writers have always looked for ways to get the ball rolling. Whether they’re Hemingway and they get drunk to get the ball to roll is up to them. But in terms of a process, it’s not that different.”

What worries him more is what happens if, over time, that ball keeps rolling in the same direction. “Set it to make money and AI will produce feedback loops, loops that make things less good,” Sharp says. “That gets you McDonald’s. But humans still want mother’s home cooking too.”

During the 2023 strike, Sharp marched with fellow writers, many holding signs aimed squarely at the moment’s anxieties. Some read: “Alexa will not replace us.” Or “AI came up with 10 suggestions for this sign: THEY ALL SUCKED.”

One afternoon, as he marched in a circle in the summer heat, a delivery robot — one of dozens now trundling through L.A. neighborhoods — rolled past.

A dark, unmistakably human thought crossed Sharp’s mind. “There I am, walking round and round with these folks,” he says, “and I remember thinking, ‘They should send a fleet of those robots down here with AI protest signs on their backs to walk the circle for us.’ Because it’s really hot out here and nobody wants to be doing this.”

For now, Sharp plans to keep experimenting quietly, pushing back against the technology he once treated as a curiosity — and wondering how long that will still feel like a choice.