Since autumn 2025, I’ve been using generative AI on my French blog, Le Carnet de Seb. The result: more frequent articles, more consistent and refreshed illustrations, and sometimes more topical content. It’s no accident that I started writing with ChatGPT. Or rather, ChatGPT has become a useful and productive writing assistant. It gave me back the urge to give some love and fresh content to a blog that has now been around for almost 15 years. It’s also what made this English edition, GateA1, possible.
In this article, I’ll show you in five steps how ChatGPT helped me get back into a regular rhythm, with the prompt I use for each one, and how it might do the same for you.
In short
- ChatGPT doesn’t replace the author: it’s an assistant that proofreads, helps with conclusions, checks facts, suggests titles and imagines illustrations.
- Each step relies on a simple prompt, included in the article, ready to copy and paste.
- Cross-checking with a second tool, such as Mistral’s Le Chat, limits errors and hallucinations.
- The final text is still yours: keep what sounds right, ditch the rest.
1. Proofreading: correcting your writing with ChatGPT
Proofreading your own writing is a bit like trying to tickle yourself: it doesn’t work. You think you’re spotting the mistakes, but mostly you read what you think you wrote. After years of publishing articles, I’ve come to accept the obvious: my brain betrays me.
I can go over a text ten times, hunt down typos, polish the wording… and a silly mistake will still slip in between two commas. Worst of all, you never see it until you’ve clicked “Publish”. It’s practically a universal law for solo bloggers.
Years later, I still find myself correcting published articles, a bit like scratching at the paint on a wall you’ve just repainted. Word could have been my safety net, but even its spellchecker has its limits.
ChatGPT, on the other hand, proofreads tirelessly, without indulgence and, above all, without ego. It never gets tired of pointing out a silly mistake (like ChatGTP instead of ChatGPT), a clumsy agreement (French grammar is the devil’s own work) or a sentence that goes round in circles. It isn’t a magic corrector, but a fresh pair of eyes, available at any hour. And that changes everything.
This little safety net gave me the confidence to publish more often, without fearing the eagle-eyed reader who will kindly point out an “its” that should have been “it’s”.
Just paste your article and add the prompt below (you can complete it with a pre-prompt to set the tone you expect):
Here is my article. Correct any spelling, grammar or wording mistakes. Put the corrected parts in bold, and below the text, list detailed explanations of your changes.
That way, proofreading also becomes a chance to learn from your mistakes.
On a Mac, there’s an alternative: Apple Intelligence’s Writing Tools. In the right-click menu of most apps, the Proofread option gives you a corrected text, but without annotations or explanations. It’ll be fixed, full stop.
2. Writing the conclusion: the agony of the final paragraph
The end of an article has always been my blogging nightmare. The beginning comes easily: you have an idea, you run with it. But the ending? Pure agony, torture, a wart at the bottom of the page.
For years, I had this feeling of a rushed ending. As if my articles just stopped dead, without a goodbye, without that little sentence that makes you want to come back. A bit like a TV series that never got the ending it deserved.
ChatGPT helps me tidy all that up. When I can’t find the way out, it opens a door: an idea for a summary, an opening line, or simply a conclusion that sounds right. It doesn’t write in my place, but in many cases, it saves the day.
Here’s the magic prompt. In my humble opinion, it gives acceptable results. It certainly does better than me.
Write a simple, natural conclusion that encourages readers to leave a comment.
What used to be torture is gone. Now it’s a relief, and an encouragement to keep going.
3. Fact-checking with AI: does my article hold up?
Writing with ChatGPT isn’t only about checking how you write. Publishing an article online means accepting a risk: being contradicted. And nothing is more unpleasant than that comment, sometimes kind but always a little prickly, that starts with: “Actually, that’s not quite right…”
For a long time, I did my checks the old-fashioned way: rereading my sources and what I’d written, even redoing the steps or the formulas… But let’s be honest: you tend to validate whatever confirms your initial idea. That’s where ChatGPT became useful, not as the “absolute truth”, but as a critical mirror.
When I finish an article, I ask it to check whether everything is consistent, whether a detail is off, or whether there’s an angle I’ve missed. It sometimes sends back extra information, or even sources I wouldn’t have found on my own. It’s like having a meticulous reader before publication, except that it doesn’t post a public comment the next day.
In practice, my method is simple: I give ChatGPT the Word version of my article and send a prompt along these lines:
Read this article carefully and do a full fact-check.
List every claim or concrete fact stated in the text.
For each one, say whether it is correct, approximate or wrong.
Give a short explanation and, if possible, one or two recent or official sources to check it.
If a claim is debatable, explain the nuances or known points of disagreement.
End with an overall summary of how reliable the text is (as a percentage or an assessment).
This prompt looks complicated, but its precision makes all the difference: ChatGPT goes through every claim instead of giving a general opinion. Before GPT-5, I used a much simpler prompt, but it isn’t enough any more.
And to compare points of view, I’ve picked up a little habit: I ask Le Chat, the assistant from the French company Mistral AI, to do the same check. Writing with AI doesn’t mean relying on a single tool.
This second opinion saves me from plenty of approximations. Above all, it spares me from having to answer a comment like: “Nice article, but that hasn’t worked for ages.”
4. Titles, subheadings and the rest: a prompt to find them
A blog post is never just a post. It also needs a catchy title, an excerpt that makes people want to click on the home page, and a description for Google that sums it all up without giving everything away. These little bits of text, usually written at the very end, take a disproportionate amount of energy.
Before ChatGPT, I sometimes spent more time finding a title than writing the introduction. Too short, too long, too bland, too “SEO”… nothing ever worked. Now I ask it to help me fine-tune things based on the full text. My prompt is simple:
Suggest five possible titles for this article, clear and natural, suited to a personal blog. The main keyword is [keyword]. Keep a human tone, not a marketing one.
It usually gets my style, although I sometimes have to remind it that I’m not trying to “sell” an article, just make it readable and honest. And often, the good idea isn’t the one it writes, but the one its suggestion sparks in my head.
As with all the other prompts in this article, keep in mind that it’s only a tool. The final text will always be written by you.
5. Image generation: visuals that finally tell the same story
For years, the most frustrating part wasn’t writing, it was illustrating. Finding a royalty-free image that wasn’t too ugly and was vaguely related to the topic.
ChatGPT changed that. Not by picking a generic image from a stock library, but by first helping me imagine it, putting into words what I want to show. In a few lines, I can then get an image that really fits my article and extends its tone, rather than just filling a gap.
It isn’t a question of pure aesthetics. It’s a question of consistency. The illustration becomes an extension of the text, not a generic backdrop added at the end. For a personal blog, that makes all the difference: the visual is part of the story, just like the title or the writing style.
Today, I no longer search for images. I design them. And that simple change has given me an unexpected pleasure: seeing my articles look like what they’re about.
Is writing with ChatGPT a good thing for you?
ChatGPT hasn’t changed the way I write; it has simply made it lighter. Fewer blocks, less hesitation, more flow. And above all, a bit more pleasure in publishing without fearing a typo, a clumsy sentence or a failed visual. A far cry from my first experiment, back in 2023, when ChatGPT tried to write an article for my blog on its own.
It isn’t a revolution, just one more tool in the blogger’s kit. But it’s a tool that, used well, makes you want to keep writing after all these years.
What about you? Has AI ever helped you get past a block, improve a text or publish with more peace of mind? Have you taken the step of writing with ChatGPT too? Tell me in the comments. I’m curious to see how everyone tames it in their own way.
Leave a Reply