AI in theatre: humans still ahead in a 2025 play study

An opinion column in the Egyptian Gazette published on November 7 argues that AI should assist theatre artists, not replace them. The evidence usually cited alongside it is a 2025 peer-reviewed study in which the human-generated play scored higher on all indexes of creativity. There's no deal, funding round or product launch behind this story. It's an argument, one small study and one university project.

What the column actually says

Dr Mahmoud Kamel, a professor at the Academy of Arts, wrote the piece, and it's labelled part one of two. He builds it on Shane Pike's book AI, AR, and VR in Theatre and Performance. His list of what AI can do is practical: suggest ideas, help develop scenes, prepare preliminary texts, produce images for stage design, and take on routine tasks.

His limit is just as clear. A scene produced by a computer doesn't automatically become a good dramatic scene, he writes, and its value depends on how the artist examines, selects and develops it. Fair enough. I'd call it a sensible position rather than news.

The study behind the headline

The study, published on 20 May 2025 in Humanities and Social Sciences Communications, compared two plays: Jordan Harrison's Marjorie Prime (2015) and AI: When a Robot Writes a Play (2021). The AI play came from THEaiTRobot, a theatre script generation tool, and ran to a full sixty minutes.

The authors didn't write the machine off. They say the AI play showed significant creativity potential, close to the human play in several indexes.

Now the caveats, and they're big. It's one comparison of two plays. The AI script is from 2021, so it tells you little about the models theatre-makers could use today; I'd guess a rerun with current tools would look different. I'd treat the result as a snapshot of a specific tool, not a verdict on AI.

Who pays and who gains

The only money in these sources is academic. Stanford's Michael Rau got a seed grant from Stanford HAI to put AI into live performance, according to a January 2025 HAI interview. One of his experiments feeds audience suggestions into a large language model, turns the script into synthetic voices, and sends each role to an actor's earbud. Actors can then perform unrehearsed.

Rau is candid about the payoff. His words: "Honestly, we don't know what the benefits will be yet." He also lists drawbacks: copyright concerns baked into the models, a homogenising effect on creative work, and a risk that artists suffer if too much of the making is handed to machines.

So the gains so far are research results and a few experimental shows, not revenue. Nobody in these sources reports ticket sales, cost savings or a paying customer.

The stakes for the AI market

Theatre isn't a market that moves AI company earnings. Still, the pattern is useful if you build or buy creative tools. Every source here puts the human in the selecting seat, and none argues for autonomous writing.

My read: the products that win in this niche will probably help a director compare and discard drafts, not generate more of them. Drafts are cheap. Judgment isn't. If a vendor pitches theatre companies on replacing writers, I'd ask for evidence first, because I couldn't find any in these sources.

What to watch next

The Gazette's second instalment is still to come, and I'd expect it to stay at the level of principle. More useful would be studies that test newer models across more plays, since a single 2021 script can't carry this debate for long. Until then, the honest summary is that the argument is settled by opinion, and the data is thin.