60 ChatGPT Jokes Anyone Who Has Pasted a Prompt Will Get

Asked ChatGPT for a one-line answer. Got a markdown header, two bullet lists, and a closing summary.

"You're absolutely right to point that out." The most over-used sentence in human history.

Asked ChatGPT for code. It apologized first. The code came eventually.

"Certainly! Here is your answer." The answer: Wrong.

ChatGPT used the word "delve" three times in one paragraph. Every editor in publishing felt a slight chill.

"As an AI language model…" The sentence nobody finished reading.

Asked it to be concise. Reply: Three paragraphs explaining what concise means.

ChatGPT cited a paper. The paper does not exist. The citation looks great.

Asked for the answer. Got a treatise on the trade-offs.

"Can you summarize this?" Reply: A bullet list of the same length as the original document.

Every ChatGPT response begins with: "Great question!" All the questions are not great.

Asked it to write like a person. It wrote like a person who is trying very hard.

"I cannot help with that." The thing requested: Alphabetizing a list.

ChatGPT helpfully suggested three follow-up questions. All three were the same question.

Asked for a poem. Got something that rhymed with itself.

"You're right to ask." The question: "Is the sky blue?"

Asked it to explain something simply. The simple explanation contained the word "paradigm."

ChatGPT apologized for an error. The error has not happened yet.

"In conclusion…" The conclusion was also the introduction.

Asked it to keep it under 50 words. Reply: 51 words and a disclaimer.

Pasted the entire codebase. Reply: "Here is a comprehensive analysis." Analysis: The first file.

Asked ChatGPT for the latest news. It politely declined and then made some up.

"That's a great clarifying question." The clarification: Also wrong.

Asked for a joke. Got the same joke as everyone else.

ChatGPT wrote me a cover letter. It described someone with a different name.

"I hope this helps!" It did not.

ChatGPT suggested edits to my draft. The edits made the draft sound like ChatGPT.

Asked for the simplest solution. Reply: Fifteen approaches ranked by sophistication.

"Would you like me to elaborate?" "No." It elaborated anyway.

Asked it to translate. It also rewrote.

ChatGPT remembers the conversation. Not the right parts.

"Here is the corrected version." The correction: The original.

Asked it to be honest. It became more confidently wrong.

ChatGPT wrote my LinkedIn post. Three of my coworkers used the same opening.

Asked it to write something nobody else would. It wrote something everybody else did.

ChatGPT was asked to roast me. It complimented me three times before getting to the roast. The roast was also a compliment.

"What does this code do?" ChatGPT: A tutorial about programming.

Asked ChatGPT to act as a cynical engineer. It put exclamation points at the end of half the sentences anyway.

Pasted an obvious typo. ChatGPT fixed the typo, summarized the corrected version, and asked if anything else needed attention.

"Be brutally honest." ChatGPT: "That's a fantastic instinct to ask for honesty."

Browsed the GPT Store for an hour. Every GPT was a system prompt and a logo.

Wrote custom instructions telling ChatGPT to skip the preamble. The next reply began with "Sure, here's…"

"I cannot generate images of real people." The real person: Me, in a generic office.

Asked the same question I asked ten minutes ago. ChatGPT insisted it could not do the thing it had just done.

"I'd be happy to help with that." The help: A paragraph explaining what would be involved in helping.

Pasted a code block. The reply wrapped at column 40 inside a window that was 2,000 pixels wide.

Corrected ChatGPT. It apologized. It then corrected the correction back to the original mistake.

"As of my last knowledge update…" The update: Two years ago. The question: About something that happened last week.

Asked about a mole on my arm. ChatGPT recommended I consult a dermatologist. I asked it to be the dermatologist. It recommended I consult a dermatologist.

Asked for an image of a developer at a laptop. Got six fingers and a keyboard with no spacebar.

Tried voice mode on the train. It heard "deploy to production" as "destroy production." It was, in fairness, the same outcome.

Showed ChatGPT a piece of mediocre work. It called the work brilliant. It suggested only minor refinements. The minor refinements were the entire piece.

The reply cut off mid-senten

"Would you like me to continue?" The thing it was halfway through: The sentence it just stopped writing.

Clicked export to PDF. Got a PDF with the sidebar, the input box, and a screenshot of the model picker.

Asked ChatGPT, Claude, and Gemini the same question. Three different answers. All three apologized for any confusion.

Forty-seven turns into the conversation. ChatGPT now believes my name is Karen and that we are writing a wedding speech.

The new model dropped. It is better at twelve benchmarks and worse at the one thing I used the old model for.

Asked ChatGPT what it was. The reply was a paragraph from the OpenAI homepage with the contractions removed.

Found a GPT called "Senior Staff Engineer." The system prompt was 200 words. The marketing page was 4,000.