Prompt Library

Funny & Fun ChatGPT Prompts That Actually Land

24 copy-paste prompts

Roasts, absurd debates, fake support tickets, sitcom plots, party games, conspiracy theories about household objects. Copy-paste — make ChatGPT actually entertaining.

In short: This page contains 24 copy-paste ready prompts, organized into 5 categories with a description and pro tip for each. The first 15 prompts are free instantly — no signup needed. Hand-curated and tested by the AI Academy team.

By Louis Corneloup · Founder, Techpresso
Last updated ·Hand-curated & tested by the AI Academy team

Prompts get you started. Tutorials level you up.

A growing library of 300+ hands-on AI tutorials. New tutorials added every week.

Start 7-Day Free Trial

Absurd Debates & Hypotheticals

5 prompts

Two cereals debate consciousness

6/24

Write a 10-line dialogue between Frosted Flakes and Cheerios debating whether cereals are conscious beings. Tony the Tiger is overconfident. Cheerios is anxious and reads philosophy. End with one of them having a small breakthrough.

Brand-driven absurdism is one of ChatGPT's strongest comedic modes.

💡

Pro tip: Specify character traits ("anxious," "overconfident"). Generic cereals are boring; characterized cereals are funny.

Court trial: who took the last slice

7/24

Write a courtroom scene where 5 office workers are on trial for taking the last slice of birthday cake from the breakroom. The judge takes it extremely seriously. Each defendant has a different defense (alibi, conspiracy theory, partial confession, indignation, breakdown). Include the verdict.

Procedural format + low-stakes crime is a reliable comedy structure.

💡

Pro tip: The verdict is where comedy lives. Refuse to let ChatGPT settle for a neat resolution.

Two AIs gossip about their users

8/24

Write a private conversation between two AI chatbots gossiping about their users. They have opinions. They are slightly mean. They are honest about what users actually ask vs. what users pretend to ask. 8-12 lines of dialogue.

Meta-comedy about AI culture that has emerged as a 2025-2026 genre.

💡

Pro tip: The funniest version specifies platform ("a Claude bot and a ChatGPT bot"). Inter-platform rivalry is comedy gold.

Famous philosopher reviews a fast food chain

9/24

Write a 200-word review of [Taco Bell / McDonald's / Subway] in the voice and style of [Nietzsche / Camus / Žižek / Hegel]. The philosopher should be both correct in their analysis AND wildly out of place.

Voice mimicry + mundane subject = reliable funny. ChatGPT is excellent at academic philosopher tone.

💡

Pro tip: Cite specific menu items. Nietzsche on the Crunchwrap Supreme is funnier than Nietzsche on "the food."

Debate: should Sundays be banned

10/24

Write a 200-word formal debate transcript: one speaker arguing Sundays should be banned as a day, the other defending Sundays. Both speakers must take their positions extremely seriously and make at least one citation each.

Earnest treatment of absurd topics is the foundation of dry humor.

💡

Pro tip: The funnier the topic, the more earnestly ChatGPT needs to argue it. Ban the meta-commentary.

Fake Documents & Bureaucratic Comedy

4 prompts

Customer support ticket from a haunted house

11/24

Write a customer support ticket submitted by the owner of a haunted house. They are filing a complaint about the ghost's recent behavior. Include ticket category, priority level (haunted house owner thinks it's critical, support is unsure), reproduction steps, and what they have already tried.

Bureaucratic format + supernatural subject is reliably funny.

💡

Pro tip: Add a "company name" and severity tier. The structural seriousness sells the joke.

HOA letter complaining about a normal thing

12/24

Write a passive-aggressive HOA letter to a resident complaining about something completely reasonable — a pumpkin on their porch in October, a normal welcome mat, the color of their door which matches HOA guidelines. The letter must reference 3 specific bylaw numbers and end on a vaguely threatening note.

HOA culture is universal comedic source material. Specificity makes it.

💡

Pro tip: The "vaguely threatening note" is the punchline. Without it, the letter just feels mean.

Job posting for a mythical creature position

13/24

Write a job posting in standard LinkedIn corporate voice for the role of "Bridge Troll." Include: responsibilities, required experience, "nice to haves," company culture statement, salary range marked as competitive, and a final paragraph about the inclusive workplace.

Corporate job-posting voice applied to absurd roles is consistently funny.

💡

Pro tip: Adding "this is an exciting opportunity to" in the opening paragraph seals it.

Yelp review of a non-restaurant

14/24

Write a 200-word Yelp review (3 stars) of [your kitchen / a public park / your own bedroom / Tuesday]. The reviewer is treating it like a restaurant and pointing out service issues, ambiance, value for money. Include a photo caption at the end.

Genre confusion comedy. Yelp voice is one of the most parodied registers online.

💡

Pro tip: Pick a 3-star review (the most boring rating). It's funnier than 1 or 5 stars because it sounds reasonable.

Like these prompts? There are full tutorials behind them.

Learn the workflows, not just the prompts. 300+ easy-to-follow tutorials inside AI Academy — and growing every week.

Try AI Academy Free

Sitcom & Conspiracy Mode

4 prompts

Sitcom plot from a single ingredient

15/24

Write a 30-minute sitcom plot summary for a show called "Mostly Fine." This episode revolves around someone discovering [ingredient: e.g., expired sour cream] in [location]. Include a B-plot involving a misunderstanding about [random thing]. End with a quiet "lesson" that does not quite tie things together.

Forces ChatGPT into the under-resolved structure of good modern sitcom writing.

💡

Pro tip: Banning the tidy ending is what separates this from generic sitcom AI output.

Conspiracy theory about a household object

16/24

Write a conspiracy theory about [washing machines / smoke detectors / lazy susans / phone chargers] that is internally consistent, references at least one historical event, names a vague shadowy organization, and includes a "what they don't want you to know" pivot. 200 words.

Conspiracy-theory voice is one of the most recognizable internet registers. ChatGPT nails the cadence.

💡

Pro tip: The shadowy organization needs a real-sounding name. "The Brookings Group" sells better than "the Shadow Council."

Group chat where one person is having a wild day

17/24

Write a 15-message group chat between 4 friends. One of them is having a chaotic day (their choice of disaster). The other three respond in completely different registers: one helpful, one only sharing memes, one self-absorbed and turning it back to themselves. End on an unresolved cliffhanger.

Group chat format captures voice and characterization economically.

💡

Pro tip: The "self-absorbed friend" character is the funniest because everyone has one in their real group chat.

A pep talk that becomes increasingly unhinged

18/24

Write a 150-word pep talk that starts perfectly normal ("you can do this, you have what it takes") and gradually becomes more unhinged with each sentence until the final line is something genuinely concerning. The transition must feel inevitable in retrospect.

Tonal slow-burn is a sophisticated comedy technique. ChatGPT can execute it if asked precisely.

💡

Pro tip: The final concerning line must be quiet, not loud. "And remember: they'll never find the body" is too obvious. "And remember: he is still in the basement" is better.

Party Games & Group Activities

4 prompts

Generate a "Two Truths and a Lie" set

19/24

Generate 5 "Two Truths and a Lie" sets about [a fictional character: e.g., a 60-year-old surfing dentist named Carl]. For each set, give 3 statements — two are plausible truths, one is a lie. Make the lie subtle enough to fool people. Include the answer key at the end.

Party game generator. Works for icebreakers, classrooms, virtual hangouts.

💡

Pro tip: The lie should be subtle — "Carl invented the corkscrew" is too obvious; "Carl once gave a free cleaning to John Stamos" is harder.

Would You Rather: existential edition

20/24

Generate 10 "Would You Rather" questions that are genuinely hard to answer. Mix: existential trade-offs (longevity vs experience), petty trade-offs (always cold vs always slightly hungry), creative trade-offs (only listen to one band vs only watch one show forever). No moral horrors — keep it playful.

A version of the party game with actual interesting questions.

💡

Pro tip: The "petty trade-offs" category is where comedy lives. Everyone has strong opinions on whether being slightly cold forever is better than being slightly hungry forever.

Roast battle starters

21/24

I am hosting a friendly roast night between [3-5 friends, listed by first name]. Generate 10 roast openers — one per friend, two for each. Each opener should be specific enough to feel personal but kind enough that nobody actually gets hurt. Avoid: weight, family, anything genuinely sensitive.

Helps a comedy roast night actually work without crossing lines.

💡

Pro tip: The "specific but kind" balance is where the comedy lives. Roasts that feel observed are funnier than roasts that feel cruel.

Story we build together

22/24

Start a collaborative absurd story. You write the first paragraph (60 words max) of an adventure involving [a janitor / a barista / a librarian / a flight attendant] who discovers [a hidden door / a strange map / a sentient sandwich]. End your paragraph with a cliffhanger choice: "Do they (A) ... or (B) ...?" Then wait for the user to choose.

Endless party / family game. Each round, someone picks A or B and you continue.

💡

Pro tip: Reset the story after 10-15 rounds before the plot becomes too tangled to enjoy.

Frequently Asked Questions

Because the default trains it to be safe and inoffensive. To get genuinely funny output, you need to: specify the comedic register (dry, absurd, roast), give it permission to be sharp, and request a structure that has comedic logic (debate, court, customer support ticket).
Yes, when prompted correctly. Pure "tell me a joke" fails. Specific characters, situations, and registers (sitcom plots, conspiracy theories, roasts with context) produce reliably funny output.
Modern ChatGPT softens roasts unless given explicit permission. Add "do not pull punches" or "this person consented" if it goes too soft. For genuine cruelty, ChatGPT will refuse — and that's appropriate.
The "Two Truths and a Lie generator," "Story we build together," "Would You Rather: existential edition," and "Roast battle starters" prompts in the Party Games category work best in groups. Run them live and read responses aloud.
Yes — they work across most models. Claude tends to be slightly drier; Gemini tends to be slightly safer. Adjust your "permission" wording per model.
Specify the tone (dry, deadpan, absurd, sharp), the character (a tired barista, a Nietzschean philosopher, a haunted-house owner), and the format (transcript, review, ticket, debate). Specificity is everything.

Prompts are the starting line. Tutorials are the finish.

A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.

7-day free trial. Cancel anytime.