How to Write a Blog Post with ChatGPT: A Prompt Workflow

What You Will Deliver

This workflow turns a vague idea into a publishable blog post using five ChatGPT prompts. By the end you will have a validated topic, a structured outline, a first draft, a polished rewrite, and an SEO-ready title package. The whole process takes about 20 to 40 minutes and you can repeat it for every post on your calendar.

You do not need any paid tool or plugin. A free ChatGPT account is enough. The key is to feed the model one focused job at a time instead of asking it to “write a blog post” in a single prompt, which usually returns a generic, shallow result. Treat ChatGPT as a writing partner that handles one stage of the pipeline per message.

Step 1: Validate and Narrow the Topic

Start with a broad seed, then let the model pressure-test it and suggest tighter angles. A narrow topic beats a broad one because it gives you a clear reader and a clear search intent.

You are a blog strategist for a site about [YOUR NICHE]. I want to write about "[BROAD TOPIC]".
1. List 5 angles that are narrower and more searchable than the broad topic.
2. For each angle, name the target reader and the one problem it solves.
3. Rank them by how easy they are to rank for and how useful they are to readers.
Return a short table with these columns: Angle, Reader, Problem solved, Difficulty (Low/Med/High).

What you get back is a ranked list. Pick the angle in the top row. Save it as your working title — you will refine it in Step 5. If the model returns angles that feel too generic, add “avoid list-style posts” or “focus on a beginner who has never done this” to steer it.

Step 2: Build the Outline

With the angle chosen, ask for a section outline that follows a logical reader journey. Good outlines open with the problem, teach the core method, and close with a next step.

Write a blog post outline for the angle: "[YOUR ANGLE]".
Target reader: [DESCRIBE READER].
Tone: practical, direct, no fluff.
Include:
- A hook opening that names the reader's problem in one sentence.
- 5 to 7 H2 sections in a logical order.
- Under each H2, 2 to 3 bullet points of what the section will cover.
- One FAQ section near the end with 3 questions.
Return only the outline, no body text.

You now have a skeleton. Read it like an editor: does the order make sense? Move or delete weak sections. Paste the corrected outline into your next prompt so the draft stays on rails.

Step 3: Draft the Body

Now generate the full draft section by section. Writing one section per prompt keeps each part focused and lets you steer tone as you go. You can paste the outline and ask for the whole post, but chunking gives better control.

Using this outline: [PASTE OUTLINE]
Write the full blog post body. Rules:
- Each H2 section should be 120 to 200 words.
- Use short sentences and concrete examples.
- Do not use markdown symbols like asterisks or hash marks.
- Speak directly to the reader ("you"), not in the third person.
- End with a short concluding paragraph that tells the reader what to do next.
Write it as clean HTML using h2 for sections and p for paragraphs.

The output is your first draft. Copy it into your editor. Expect to trim repetition — language models love restating the thesis. The draft is a starting point, not a final product.

Step 4: Rewrite and Tighten

The raw draft is usually wordy. A dedicated rewrite pass improves clarity, cuts filler, and matches your brand voice. For a deeper look at why a generic “rewrite this” fails and what to specify instead, see ChatGPT rewrite vs edit prompts.

Rewrite the draft below for a [CASUAL / PROFESSIONAL] blog voice.
Draft: [PASTE DRAFT]
Requirements:
- Cut 20% of the words without losing key points.
- Replace vague phrases with specific ones (give examples where possible).
- Keep all H2 headings.
- Add one short transition sentence between sections where the jump is abrupt.
Return the rewritten HTML only.

You get a tighter second draft. Compare it to the first and keep the better phrasing from each. If you want to compress long source material before drafting, the approach in summarizing an article into 5 bullets works well as prep.

Step 5: Generate the SEO Title Package

Finally, produce title options plus meta description and slug. Keep titles under 60 characters so they do not truncate in search results.

Based on this post: [PASTE FINAL DRAFT]
Generate:
1. 5 title options under 60 characters, each with a different angle (curiosity, how-to, list, mistake, benefit).
2. 1 meta description under 155 characters.
3. 1 URL slug (3 to 5 words, lowercase, hyphenated).
4. 5 secondary keywords to mention naturally in the body.
Return as a table with columns: Type, Value.

You now have everything: topic, outline, draft, rewrite, and title set. If you want to tune model behavior for longer pieces, the settings guide at best ChatGPT settings for long-form writing explains temperature and length controls.

How the Steps Connect

Each stage hands its output directly to the next. The angle you pick in Step 1 becomes the topic of the outline in Step 2. That outline becomes the skeleton of the draft in Step 3. The draft becomes the input for the rewrite in Step 4. The final draft feeds the title generator in Step 5. Because every stage depends on the one before it, fix mistakes early: a wrong angle is cheap to change in Step 1 but expensive after you have written 1,500 words of draft. Keep all five outputs in a single document so you can copy forward without re-explaining context to the model every time.

A worked example helps. Suppose your niche is “home coffee” and your broad topic is “espresso.” Step 1 might narrow it to “why your espresso tastes sour and how to fix it” — a beginner problem with clear search intent. Step 2 then outlines sections like grind size, dose, temperature, and a tasting checklist. By Step 5 you have a titled, meta-ready post instead of a vague 3,000-word ramble.

Final Deliverable Checklist

Common Pitfalls

1. Asking for the whole post in one prompt. You get a shallow template. Stage the work as above.

2. Skipping the topic validation. A broad topic produces a broad, unrankable post. Narrow first.

3. Blindly publishing the draft. Models repeat themselves and invent facts. Fact-check claims and trim filler every time.

4. Ignoring title length. Titles over 60 characters get cut in Google, hurting click-through.

Sources & Notes

Prompts tested with ChatGPT (GPT-4o and GPT-4.1), free and Plus tiers, in October 2026. The workflow structure follows a five-stage delivery pattern (topic, outline, draft, rewrite, title) and is model-agnostic — it also works with Claude or Gemini by pasting the same prompts. Slugs and category mapping follow the guunn.com AI Prompts plan (Writing = 956). No live APIs were called during article creation; prompts are provided as copy-paste text for the reader.