ChatGPT will write your newspaper article. It won't set the page.

Short answer: ChatGPT can write a convincing newspaper article in about fifteen seconds. It cannot give you a front page. Text models output text, and image models output pictures of text — neither one produces a typeset, editable, shareable newspaper page. To get that, you need something that runs a layout engine after the writing is done.

This post covers three things: the prompt that gets the best article out of ChatGPT, the technical reason the layout never comes with it, and the three real routes to an actual front page — including the free one.

What ChatGPT actually gives you when you ask for a newspaper article

Type "write a newspaper article about my roommate eating my leftovers" into ChatGPT and you get something genuinely good: a headline, a dateline, a lede, and three tight paragraphs in inverted-pyramid order. The writing is the easy part, and AI is good at it.

Then you look at it. It's a chat bubble. Grey background, sans-serif system font, a copy button in the corner. Nothing about it reads as a newspaper.

Ask for the layout too, and you'll get one of two things:

1. ASCII art. The model draws a "front page" out of equals signs and pipe characters. It falls apart the moment the text wraps, and it looks like a 1994 terminal, not a broadsheet.
2. A generated image. Ask for a picture of a front page and modern image models will render the big headline convincingly. Then look at the body columns: repeated words, invented letterforms, text that dissolves into texture halfway down. And even when it looks right, it's a picture. You can't fix a typo, can't change the date, can't export a PDF, can't add tomorrow's story to it.

Neither is a newspaper. Both are a picture of one, and not a good picture.

The prompt that gets the best article out of ChatGPT

If all you need is the writing, the fix is specificity. Generic prompts produce generic copy because a real newspaper article has six distinct parts and the model won't infer them. Name all six.

Copy this:

Write a single newspaper article in the style of a serious daily broadsheet.

EVENT: [one line about what happened]
TONE: deadpan and completely straight — never wink at the reader,
      never acknowledge the subject is trivial

Give me, in this order and labeled:
1. MASTHEAD — an invented paper name, 2-3 words
2. DATELINE — city, state, and today's date
3. HEADLINE — 6 to 9 words, present tense, no articles ("a", "the")
4. DECK — one subheading of 12 to 18 words that adds new information
5. BYLINE — an invented reporter name and a beat title
6. BODY — exactly 180 words, inverted pyramid, past tense, with one
   invented quote from a named eyewitness and one from an official source

Rules: no exclamation marks. No adjectives in the headline. Do not explain
the joke. Write as if this were the most important thing that happened today.

Three things in there are doing the real work:

• "Exactly 180 words." Language models are famously careless about length unless you give them a number. "Short" means nothing; 180 means 180-ish.
• "Never wink at the reader." This is the whole comedy engine. A trivial event set in the register of a serious newspaper is funny because nothing in the writing admits it. The second the model adds "in a hilarious turn of events," it's dead.
• "No articles in the headline." Real headlines drop them. "Router Surrenders After 45-Second Siege," not "The Router Surrendered After a 45-Second Siege." This single instruction does more for authenticity than any amount of "make it sound like a real newspaper."

Run that and the writing will be genuinely good. You will still be looking at it in a chat window.

Why the page is the part AI can't do

This isn't a limitation that goes away with the next model release. It's a category difference.

Typesetting is not writing. Setting a front page means solving a layout problem: balancing columns of unequal length, hyphenating and justifying so the right margin stays clean, sizing a headline to the exact column measure it has to fill, hanging a drop cap into the first three lines of the lede, deciding where a story jumps to page two. That is a layout engine doing geometry, and it runs after the words exist.

A text model has no canvas. An image model has a canvas but no text engine — it's painting pixels that resemble letters, which is why the body copy on any AI-generated newspaper image turns to soup under a zoom. The two capabilities live in different kinds of software, and chatting with an AI gives you access to neither.

So the honest answer to "is there an AI that makes a newspaper layout?" is: the AI writes, and a separate typesetting engine sets. You want a product where both are wired together.

Three ways to actually get a front page

The first two are the pre-AI world: you either do the writing or you do the design. Newzly's whole reason to exist is that both halves are now solvable at once.

How Newzly works

You type one line about something that actually happened. Newzly's AI writes it as a news story — headline, deck, byline, dateline, body — and then a real typesetting engine sets that story on a dated newspaper page, with a masthead, newsprint typography, columns, and a drop cap.

From there it's an object you can move:

• A share link anyone can open and read without an account.
• A 1080×1920 story card sized for Instagram, TikTok, and Snapchat.
• A PDF export.

Multiple stories compose into a dated edition — a multi-page paper for that day rather than a single loose article. Editions are dated by your midnight, not a server's, which matters more than it sounds like it does if you write at 1am.

You can try it without making an account: the box on newzly.app takes one line and gives you a front page back. It's free to start and there's no card. Premium is $6.99 a month or $60 a year, which works out to $5 a month.

One thing Newzly is not: a news source. It's a newspaper about your life — your Tuesday, your dog, your group chat. The mastheads are invented. Nothing in it is real journalism and it isn't pretending to be.

If you're using ChatGPT as a journal, this is the same problem

A large number of people have quietly turned ChatGPT into a diary. It's good at it: it asks decent follow-up questions, it never judges you, and it's already open. AI journaling has become one of the fastest-growing things people do with a chatbot: US search interest in the term is up more than sevenfold year over year, according to Google Trends.

It has one structural flaw, and it's the same one as the newspaper article: at the end of the year you have nothing.

• A chat log is not an archive. There's no calendar of days, no way to flip through last March. Search only works if you remember a phrase you used.
• It isn't durable. In 2026, OpenAI users reported saved memories vanishing, a library bug wiped conversation history for some accounts, and one professor lost two years of work after changing a training-data setting. Chat threads are not built as a system of record.
• You can't hand it to anyone. There is no version of a ChatGPT thread you'd send to your mother.

Dedicated journaling apps solve the first two — Day One has added an AI chat that turns a conversation into an entry, and a wave of AI journals now compete on mood tracking and encryption. If what you want is a private, searchable, long-term record of your inner life, use one of those. That's genuinely what they're for.

Newzly solves the third one, which nobody else is really trying to solve. The output isn't an entry in a database you'll never open. It's a page — dated, typeset, and built to be sent. Your friends subscribe and read your paper the way they'd read anyone's. The reason people keep coming back to it isn't discipline; it's that somebody replied to yesterday's edition.

And if the feed you are getting away from is the other kind — the one that keeps showing you people you have never met — that is a different problem. Here are five private social networks compared, including what you can actually fix inside Instagram first.

FAQ

Can ChatGPT make a newspaper front page? No. ChatGPT can write the article — headline, byline, and body — but it outputs text into a chat window. It has no typesetting engine, so it can't produce a laid-out page. Asking for an image of one produces a picture whose body text is unreadable and which can't be edited afterward.

How do I get ChatGPT to format an article like a newspaper? Name the six parts explicitly in your prompt — masthead, dateline, headline, deck, byline, and body — and give the body an exact word count. Formatting in the chat window will still be plain text; the prompt improves the writing, not the layout. There's a full copy-paste prompt above.

Is there a free AI newspaper generator? Newzly is free to start and needs no card, and you can generate a front page from the box on the home page without making an account. Free template editors like Fodey and 101Planners will lay out a page but won't write anything for you.

What's the difference between this and a fake newspaper generator? Fake newspaper generators are template editors — you type your own headline into a fixed design. Newzly writes the story from a one-line prompt and sets it, then keeps it as a dated edition you can build on. It's also explicitly about your own life rather than imitating a real publication.

Can I use ChatGPT as a journal? You can, and plenty of people do. The limitation is that a chat thread has no calendar, no reliable long-term archive, and nothing you'd ever show another person. If you want a private searchable record, use a dedicated journaling app. If you want your day to end up as something you'd actually send someone, that's what Newzly is for.

Does Newzly work on mobile? It's a web app and it works in a mobile browser. There is no native iOS or Android app.

Try it with one line

Think of the most boring thing that happened to you today. The dishwasher. The group chat. The parking ticket. That's the line.

Set it in print at newzly.app →