What don't paste the AI argues, in the nohello tradition
Don't paste the AI is a single web page with one request: if someone asks you a question, answer it yourself instead of forwarding a chatbot's reply. It calls itself a spiritual cousin of nohello.net, a site that asks people to stop opening a chat with a bare "hello", and it says you can send its link to the person who pasted a chatbot's reply at you.
It opens with a scene. Someone asked you an honest question, you put it into a chatbot, copied what came out, and sent it back.
The exchange the page describes
They want your answer, not a chatbot's
The page says the paste felt fast and felt helpful, then adds "It wasn't." It ends the section with a plain request: the world has plenty of people who do not want to read or think things through, so do not join them.
The page does not say AI output is wrong or can be spotted. It says AI is a great tool for drafting, and the three replies below are what it asks for instead.
"Someone asked you an honest question. You dumped it into a chatbot, copied the answer, and sent it back."／"It felt fast. It felt helpful. It wasn't."／"The world is full of people who don't want to read or think things through. Don't be one of them."（dontpastetheai.com・the opening and the What happened section）View official source →
The other person has the same tools
The page's reason is about access. The person who asked has the same tools you do. If all they wanted was the generic answer, they could have it in a couple of seconds without you.
"has the same tools you do"／"If they wanted the generic answer, they'd have it in a couple of seconds."（dontpastetheai.com・two fragments from the What happened section）View official source →
The three replies it offers instead
The page does not tell anyone to stop using AI. It says AI is a great tool for drafting, and it puts the line at what you send. Under "Do this instead" it lists three replies, in this order.
The three replies, as the page words them
| Option | What the page says to do |
|---|---|
| 1. Rewrite | Read what the AI gave you, then write your own version or polish the text |
| 2. Quote and explain | Quote the useful part of the answer and say why it is useful |
| 3. Say nothing | If you have nothing to add, say so; silence is also an option |
Rewrite it in your own words
The first option is to read the AI's answer and then write your own version, or polish the text. The page puts it as not being a middleman between the model and the answer.
Its bar is low: keep the part that answers the question, drop the rest, and three sentences is all it takes. Even if the sentences are copied, the page says, at least you read them. A reply that short is easy to check before sending.
"You can use AI. Seriously! It's a great tool for drafting."／"Don't be a middleman between it and the answer."／"Take the part that actually answers the question and ignore the rest. Three sentences is all it takes, and even if they're copied, at least you read them."（dontpastetheai.com・the Do this instead section）View official source →
Quote the useful part and say why
The second option keeps the AI's words but marks them as someone else's. If one part of the model's answer is useful, quote it and explain why it is useful. The page's own example reads "I asked Claude and this bit here makes sense:", which names Claude as the source.
That way it is clear which words came from the model and which are yours. Related reading on this site: Claude Text Watermark: What a Hit Actually Proves and AI Agent Detection: Binary Bot Filters Miss 39%.
"If some part of the model's answer is genuinely useful, quote it and explain"／"I asked Claude and this bit here makes sense:"（dontpastetheai.com・the second fragment is the example phrasing the page gives）View official source →
Say nothing: the page lists silence as an option
The third option is the shortest. If you have nothing to add, say so. The page adds that silence is also an option.
"If you have nothing to add, just say so."／"Silence is also an option."（dontpastetheai.com・the Do this instead section）View official source →
Where dontpastetheai came from, and how far it spread
Nobody is named as the author. Its Hacker News record is three submissions that scored 4, 1,040 and 8 points.
The three Hacker News submissions (2026, times in UTC)
Satire, written by some random guy
The footer describes the author as some random guy and the site as a spiritual cousin of nohello.net and dontasktoask.com. nohello.net asks people not to send a bare greeting. dontasktoask.com asks people not to ask whether they may ask a question. The footer calls the site satire, allows copying, changing and translating, and says pull requests are welcome on GitHub.
The page links a version with stronger feelings and has a line for teachers whose students send in AI-generated work. It also says how to use it: send the link to anyone who has just dumped a long block of LLM text into your DMs, Slack or a code review.
"written by some random guy"／"spiritual cousin of"／"This site is satire (in case you hadn't noticed) — Feel free to copy, change, translate, etc."／"PRs welcome on GitHub"／"If your students send you an AI-generated dissertation or project, you can send them this link:"／"If someone just committed a wall of LLM text in your DMs, Slack or code review, you can send them this link."／"If someone sent you this link: they want you to think about it. Read, breathe, and write the next answer yourself."（dontpastetheai.com・the footer, the variant links and the Want to send this to someone? section; the nohello.net and dontasktoask.com links sit between the second and third fragments）View official source →
1,040 points on Hacker News versus 4 for the earlier post
A Hacker News search for dontpastetheai returns three submissions. The main page's submission on August 20, 2026 had 1,040 points and 578 comments when read on August 23, 2026. Both numbers keep changing, so they are tied to that date.
The same URL was submitted on August 18, 2026 and had 4 points; the 1,040 on its August 20 submission is 260 times that. The angry variant was submitted on August 20, 2026, about two hours after the main one, and had 8 points.
"2026-08-18T11:40:43Z"／"2026-08-20T08:20:44Z"／"2026-08-20T10:35:03Z"（Hacker News Algolia Search API・the created_at values of the three submissions; the point and comment counts are live values and are given in the body with the date they were read）View official source →
Reply in your own words instead of pasting AI output
Don't paste the AI makes one claim: the person who asked has the same tools you do, so they could have the generic answer in a couple of seconds themselves. It offers three replies instead, which are to rewrite the answer in your own words, to quote the useful part and say why, or to say you have nothing to add.
It is anonymous satire that calls itself a spiritual cousin of nohello.net and dontasktoask.com. The score behind it is 1,040 points on one Hacker News submission, read on August 23, 2026.
FAQ
- What is dontpastetheai.com?
- A single page asking people not to answer a question by pasting a chatbot's output. It calls itself a spiritual cousin of nohello.net and dontasktoask.com, says at the foot of the page that it is satire, and describes its author only as some random guy.
- What does it want you to do instead?
- It lists three replies. Read what the AI gave you and write your own version, keeping only the part that answers the question. Or quote the useful part and explain why it is useful. Or say you have nothing to add, with silence also listed as an option.
- How popular did it get?
- The main page's submission on August 20, 2026 had 1,040 points and 578 comments on Hacker News when read on August 23, 2026. The same URL was submitted on August 18, 2026 and had 4 points.
Related articles
Claude Fable 5.2: 20% price cut, up to 30% fasterAnthropic released Claude Fable 5.2 on November 9, 2026. API prices fall 20% to $8/$40 per million tokens, and output is up to 30% faster than Fable 5.1.
PayPay security and passkeys in 10 Japan pay appsPayPay security compared with 9 other Japanese payment apps: passkeys, extra checks to send or pay, and fraud compensation deadlines.
AI fact check: only 40% of cited stats hold upAI fact check: Trakkr traced 1,285 stats cited by ChatGPT, Perplexity, Gemini, Claude and AI Overviews. Only 40% led to a matching study. How to check one.
Akamai: online scams double and overtake phishingAkamai's fraud report says online scams doubled to about 6.4 million and passed phishing in Q1 2026, while phishing sites now last 22 days instead of 48.