AI in Hollywood is already common if the term includes search tools, internal process automation and writers testing prompts for ideas. That is very different from proving that studios are broadly producing finished films and television with generative systems. Matthew Belloni, the Puck founding partner and longtime Hollywood reporter, made the “everyone is doing it” case in an interview with WIRED, but it is an informed characterization, not an industrywide adoption count.
Belloni separates the business into two buckets. The first is what he calls corporate AI: software used to speed up processes or improve search. The second is generative AI, which produces or changes text, images and video. The distinction is not pedantic. It is the difference between finding an asset faster and asking a model to help make the asset.
How is AI being used in Hollywood?
At the assistive end, Belloni said screenwriters he has spoken with use tools such as ChatGPT to suggest scenarios, jokes or descriptions of a location. He did not say those suggestions make it into final scripts. A language model generates text by predicting likely next pieces of language from patterns in its training, rather than retrieving a finished answer from a filing cabinet. How LLMs work when they answer a prompt explains that mechanism.
MIT Sloan Management Review similarly described generative systems as capable of proposing storylines, character arcs and dialogue. It also noted that the technology can generate entertainment-relevant text, marketing material and images. Those capabilities describe possible uses, not a reliable measurement of how often productions deploy them.
The more consequential experiments sit further up the production chain. Belloni pointed to animation, storyboarding and visual scene work as likely early sites of disruption. He said generative tools can be used to complete visuals or create storyboards without artists, though he also said the AI-generated films, shorts and pilots he had seen were not very good. That is an assessment, not a technology benchmark, but it fits the practical attraction: changing pixels after a shoot can be cheaper than rebuilding a set and bringing people back.
Belloni also described a product tied to Ben Affleck’s former AI company as a way to manipulate footage already shot. In the interview, he said it could extend the edges of a frame, generate additional coverage or avoid certain reshoots. The description has not been independently validated here, and it should not be confused with a demonstrated end-to-end movie-making system.
Why is generative AI still contentious?
Public use carries a reputational cost. Belloni said AI retains a stigma in parts of Hollywood and recounted backlash after Amazon announced AI-related projects involving human creators; according to his account, one creator left after online criticism. He also relayed Netflix’s statement on a recent earnings call that 300 shows used generative AI, but the available material does not define “use” or provide the call record. The number should be treated as a company claim, not a census.
The dispute is larger than taste. MIT Sloan Management Review flagged questions over copyright, ownership and systems producing characters influenced by existing intellectual property. Jonathan Taplin of USC’s Annenberg Innovation Lab warned that training practices may disregard copyright and that creative workers could lose work. Those are his views, not settled findings, but they explain why a harmless-looking prompt for a joke has become part of a labor and rights fight.
The available evidence shows experimentation across the workflow, especially at the lower, assistive end. It does not establish that generative AI has taken over Hollywood’s finished entertainment. The slogan outruns the audit.
This story draws on original reporting from WIRED.
