ChatGPT Blog Prompts for Every Stage of a Post
The best ChatGPT blog prompts give the model three things: who the post is for, what it should accomplish, and what you already know that it doesn't. Generic prompts produce generic posts. Below are prompts for each stage of writing a blog post, from finding a topic to polishing the final draft, with notes on how to customise them. Swap the bracketed parts for your own details.
Before you start: build a context block
Most weak AI blog drafts fail because the model is guessing about your audience and voice. Write a short context block once and paste it at the start of each session:
I write a blog for [audience, e.g. small bakery owners in the UK]. My readers already know [basics] and want [goal]. My tone is [e.g. friendly, practical, a bit dry, no hype]. I avoid [phrases, topics, claims]. I have personal experience with [relevant background].
This takes five minutes and improves every prompt that follows.
Prompts for finding blog topics
List 25 blog post ideas for [audience] who are struggling with [problem]. Make each one a specific title, not a broad topic. Avoid beginner listicles; my readers have [level of experience].
Here are my five most popular posts: [titles]. Suggest 15 new posts that would appeal to the same readers without repeating these topics.
What questions do [audience] commonly ask about [subject] that are often answered badly or too vaguely? Give me 15, each phrased the way a real person might type it into a search bar.
Give me 10 contrarian or myth-busting post ideas about [subject]. For each, state the common belief and the alternative view in one line.
Treat these lists as a starting point. Check search interest and your own expertise before committing. A random topic generator is handy when you just need a nudge out of a rut.
Prompts for outlines
Create an outline for a [word count]-word blog post titled "[title]" for [audience]. Use H2 and H3 headings. Under each heading, write one sentence describing what that section must cover. Include a section that addresses the most common mistake readers make.
Here is my rough outline: [paste]. What is missing that a reader would expect? What could be cut? Suggest a better order.
Turn this outline into a version where every H2 heading is phrased as a question a reader would ask.
If you'd rather start from a structured draft, Inkwell's outline generator builds one from a title and a few notes.
Prompts for drafting sections
Drafting section by section gives you far more control than asking for a whole post at once.
Using the outline above, write only the section "[heading]". Aim for about [number] words. Include one concrete example involving [specific situation]. Do not write an introduction or conclusion.
Write this section as if explaining it to a friend over coffee: short paragraphs, plain words, one analogy.
Here are my notes for this section: [bullet notes, data, anecdotes]. Turn them into flowing paragraphs. Keep every fact I gave you and don't add new statistics.
That last instruction matters. Models can produce numbers that sound plausible but have no source. Tell it not to invent figures, and check any it includes anyway.
Prompts for introductions
Write three different opening paragraphs for this post: one that starts with a short scenario, one that starts with a direct answer, and one that starts with a surprising but true observation. Each under 80 words.
My intro is too slow. Rewrite it so the reader knows what they will get within the first two sentences: [paste intro].
Write a one-sentence hook for this post that a reader would want to finish reading.
For search-driven posts, the direct-answer style usually serves readers best. They came with a question; answer it first, then expand.
Prompts for conclusions and calls to action
Write a conclusion for this post that summarises the three main takeaways in plain language and ends with one clear next step for the reader. No "In conclusion."
Suggest five calls to action for the end of this post. The goal is [newsletter signups / comments / trying a tool].
Prompts for editing
Edit this draft for clarity. Shorten long sentences, remove filler, and flag any claim that needs a source. Show me a list of changes rather than rewriting silently.
Point out every place where this post sounds generic or could have been written by anyone. Suggest where I could add a personal example or specific detail.
Read this as a sceptical expert in [field]. What would you push back on?
Check this post for repetition. List sentences or ideas that appear more than once.
Asking for a list of suggested changes, rather than a rewritten draft, keeps your voice intact. You decide which edits to accept.
Prompts for SEO elements
Write five title options under 60 characters for this post. Include the phrase "[keyword]" naturally in at least three.
Write a meta description under 155 characters that tells the reader exactly what they'll learn.
Suggest a short, readable URL slug for this post.
List five related questions readers might also have, which I could answer in an FAQ section.
Suggest places in this post where a link to my article on [topic] would be genuinely useful to the reader.
A headline generator is another quick way to get title variations to compare.
Prompts for repurposing
Turn the main points of this post into a five-tweet thread. Keep each under 270 characters.
Write a 150-word LinkedIn post based on this article, opening with the most practical insight.
Turn this post into a short email for my newsletter subscribers with a link to read the full version.
How to keep AI-assisted posts from sounding generic
Prompts help, but the real fix is what you add:
- First-hand experience. What happened when you tried this? What went wrong?
- Specific numbers you can stand behind. Your own prices, timings, results.
- Opinions. Readers return to writers who take a position.
- Local or niche detail. The supplier you use, the regulation in your region, the quirk of your industry.
- Ruthless cuts. Delete any sentence that would fit equally well in a competitor's post.
Also read every factual claim with suspicion. Language models can state wrong things confidently, especially about recent events, product features, prices and laws. Verify against primary sources before you publish.
A sample workflow
- Paste your context block.
- Run a topic prompt and pick one idea.
- Generate and fix the outline yourself.
- Draft one section at a time, feeding in your own notes.
- Write the intro last, once you know what the post actually says.
- Run an editing prompt and accept changes selectively.
- Add personal examples and check every fact.
- Generate title and meta options, then choose.
FAQ
Can I publish a blog post written entirely by ChatGPT?
You can, but fully generated posts tend to be generic and may contain errors. Posts that perform well with readers usually include original experience, accurate facts and a clear point of view, which require your input and editing.
What is the best single prompt for a blog post?
There isn't one. Splitting the work into stages (topic, outline, sections, editing) produces better results than a single "write a blog post about X" prompt, because each stage gets focused instructions.
How do I make ChatGPT match my writing style?
Paste two or three paragraphs you've written and ask it to describe the style: sentence length, vocabulary, tone. Then include that description in your context block. Expect to edit the output anyway.
Will search engines penalise AI-assisted posts?
Search engines generally say they focus on whether content is helpful and original, not how it was produced. Thin, mass-produced pages tend to struggle regardless of method, so focus on usefulness and accuracy.