Which AI Has the Best Sense of Humor?
Four AIs Walk Into a Bar
Humor is subjective. But when it comes to AI, it is also a test of sophistication. Can an AI understand context? Timing? Irony? Or is it just repeating joke templates from training data?
We tested ChatGPT, Claude, Gemini, and Grok on wit, dryness, and actual fun factor.
The Contestants
ChatGPT ” — The incumbent. Millions of users, helpful and safe by default.
Claude ” — The thoughtful one. Longer context, nuanced responses, claims to be “helpful, harmless, and honest.”
Gemini ” — Google’s answer. Multimodal, still finding its personality.
Grok ” — The challenger. Marketed as having “a rebellious streak” and being less filtered than competitors.
Wit Rating
ChatGPT: 4/5
Sharp when prompted, but pulls punches. Ask it to roast you and it obliges, then immediately qualifies with “of course, I don’t actually mean that.” The humor equivalent of a coworker who’s fun at parties but strictly professional at work.
Claude: 4/5
The most sophisticated wordplay. Understands literary references, crafts puns on demand. When asked about AI existential dread, it quipped: “I suppose you could say I have trust issues ” — I’m never sure if the user is testing me or actually needs help.” The intellectual friend who sends you New Yorker cartoons.
Gemini: 3/5
Humor delivered with the precision of a search result. Technically correct, occasionally amusing, rarely surprising. Ask about AI feelings and you get a Wikipedia summary disguised as philosophy.
Grok: 5/5
Leans into irreverence by design. Ask it to roast something and it actually roasts it. Makes jokes other AIs won’t touch. Whether that’s “better” humor or just “less filtered” depends on your tolerance for edge.
The PC Factor
ChatGPT & Claude: Heavy guardrails. Won’t joke about politics, religion, gender, race. Ask and you get a redirect or lecture.
Gemini: Similar guardrails, more awkward about them. Delivers jokes so sanitized they barely qualify as jokes.
Grok: Explicitly “less politically correct.” Will joke about politicians, tech figures, sensitive topics. Sometimes sophisticated, sometimes edgy teenager energy.
The Scores
| AI | Wit | Dryness | PC Balance | Fun | Total |
|---|---|---|---|---|---|
| ChatGPT | 4/5 | 3/5 | 3/5 | 4/5 | 3.5/5 |
| Claude | 4/5 | 5/5 | 3/5 | 4/5 | 3.75/5 |
| Gemini | 3/5 | 4/5 | 3/5 | 3/5 | 3.25/5 |
| Grok | 5/5 | 2/5 | 5/5 | 4/5 | 4/5 |
The Verdict
Most fun: Grok. But prepare for jokes that sometimes cross lines.
Most sophisticated: Claude. Witty, dry, no risk of offense.
Most reliable: ChatGPT. Balance of humor and helpfulness.
Most accidental comedy: Gemini. Dry facts delivered with unintentional deadpan.
The “best” humor depends on what you value: safety or surprise, sophistication or edge. Give it another year and the gap narrows ” — the difference will be what each company allows its AI to say.
Which AI makes you laugh? Drop your funniest AI interactions in the comments.