Think you’re funnier than ChatGPT? Think again. A recently released study from the University of Southern California found that AI-generated jokes performed better than those written by humans.

The researchers behind the study—which was published last week in the journal PLOS ONE—recorded the reactions of participants to jokes written by ChatGPT 3.5, and others written by regular people. They found that nearly 70% of the participants rated the ChatGPT jokes as funnier than those crafted by humans, as compared to only 25% who liked the human-created jokes more.

In a second experiment, the researchers challenged ChatGPT to develop new headlines based on the style of satirical news website The Onion. 200 participants were asked to rate the humour of the original Onion headlines and the new ChatGPT-generated ones. The result? A draw. The participants found the ChatGPT headlines just as funny as the ones written by professional comedy writers.

So does this mean that comedians and comedy writers need to be worried? Perhaps, but there are some caveats. For one, ChatGPT was not creating these jokes from scratch—the prompts and set-ups did a lot of the heavy lifting. In the second experiment, ChatGPT was able to copy The Onion’s immediately recognisable, well-honed headline style, but imitation is far easier than coming up with a comedic style all on your own.

Most importantly, at least for comedians, so much of what makes a good joke is the delivery. Until AI can get on stage and charm an audience into laughing at dead baby jokes, it still has a long way to go before it can replace comedians. Still, the study hints at the potential disruption that generative AI and LLMs can cause in the comedy industry.

“The implications are more positive for people who merely want to reap the benefits of elevating their everyday communications with a dose of humour,” the study’s authors Drew Gorenz and Norbert Schwarz write. “But for professional comedy writers, our results suggest that LLMs can pose a serious employment threat.”