Don’t Quit Your Day Job, AI

Have you heard the one about the AI comedian?

No? Neither have I.

One person’s laughter is another person’s shoulder shrug. We’re not all wired to laugh at the same things. My friends tell me I’m funny, and I like to think I am, but they’re my friends. It’s not like I’m doing open mic at the ComedyPlex when I post a bad pun on Facebook. I can think of very few ways that I’ve put myself out there publicly to get a laugh out of someone.

One of those ways is the famed caption contest in The New Yorker. For the uninformed and non-snooty, The New Yorker runs a weekly caption contest where they publish a cartoon and invite readers to create a funny caption for it. Readers vote on the top three finalists for a final vote. The top-rated one is the winner. Sometimes, I say begrudgingly, it is even funny.

I made it to the top three over 10 years ago, and lost, and I hadn’t played at all in the intervening years until just a few months ago.

After striking out a few times, I thought: Why am I bothering to create original content? As every content creator knows, nobody does that anymore. They enter prompts in ChatGPT or another AI platform, and their content is magically spat out, search-engine-optimized, formatted for HTML, chrome-wheel fuel-injected, and ready for posting.

So, if this is how content is generated in the age of AI, is it possible that AI could win The New Yorker caption contest?

I decided to turn this into an experiment, because if AI could win or even make it to the top three, I first had to determine whether AI is funnier than I am. (And to clarify: funnier to humans, because who knows what these platforms talk about when they get together and plan our destruction.)

Here was my plan: submit my own captions for two contests in a row, see if any make it to the top three, and then submit a third created by AI. If the good readers of The New Yorker selected me in one (or both) of the first two rounds, I’d still submit an AI version as a test, but my AI prompts would have a lot more rap-like dissing to them. (“Your move, Claude. Nobody listens to techno!”) If they selected AI but not me, I would feel bad for a short time, but eventually I would understand that it had access to the sum of all human knowledge, which is mostly hilarious. If we both failed to capture the attention of editors and readers at The New Yorker, it would be a wash.

Here are the results.

Week 1

This one was a toughie, and I have to admit, I didn’t give it my A game with this clunker:

“You asked me to go for a walk. I didn’t know it would be a waddle.”

I knew in advance that I wasn’t going to make it, and I was right. No top three for me.

Week 2

I felt a lot more confident this time around, but I thought others would surely spot the same point of entry, so I hit it with a one-two punch. Maybe it was too much?

“Sometimes I feel like there’s a big hole at the center of my life, and I’m just frittering away.”

The finalists came around, and the big hole turned out to be a self-fulfilling prophecy.

So now the question was whether AI could go where my jokes had not gone before. Was it up for the task?

Week 3

I have to admit that this one was a toughie. It’s difficult to tell what the hell is going on in this cartoon, which, admittedly, would not make it unique among New Yorker cartoons.

I threw this at Claude, the platform I’ve been preferring lately.

“They said the neighborhood had good schools. They didn’t mention the recitals.”

For me, that was a combination of a tiny smirk and a “meh.” I didn’t feel particularly threatened. I could grin and bear it.

Giving AI the benefit of the doubt, I remembered the many times I asked a question on different platforms and received wildly different answers. Could Google Gemini do any better?

“It’s a nice change of pace, but I still think the bird feeder was cheaper.”

Mon dieu! First I 😳’d, then I 🤦🏻‍♂️’d. What bird feeder? Where’s the joke? You’d think a large language model would understand a non-sequitur.

On to Microsoft CoPilot:

“He’s great, but once you hire him, he never leaves.”

Ugh. That’s about as funny as Bill Gates.

Should I try with one last round on ChatGPT, the granddaddy of AI? I’d been trying to avoid the platform, as their ties to defense spending have troubled me of late. (Actually, it’s because I’ve felt my prompts were not being taken seriously, such as “Rewrite the opening paragraph of A Tale of Two Cities as told by Snoop Dogg.”)

Copy and paste. Go for it, OpenAI.

“I guess the noise ordinance doesn’t cover woodwinds.”

Meh.

I submitted Claude’s version, confident that it was a draw. And indeed I was correct. AI didn’t bring its AI game, and it was not better at humor than I am. But it was no worse, either. I would have to be satisfied with the outcome, but at least I’m better at not threatening humanity with extinction, so I have that going for me.

My next test: can AI be more self-deprecating than I am? I’m rooting for Gemini to win by pointing to the bird feeder joke. “Oy, what was I processing?” Now that’s a good prompt.

Note: all cartoons Copyright 2026 The New Yorker.