Disclosure: this post may contain affiliate links and links to our own products.
Most AI blog posts are easy to spot: a vague intro, "in today's fast-paced world", five generic tips, and a conclusion that repeats the intro. Readers leave, and Google has little reason to rank it.
The problem isn't ChatGPT. It's asking it to write the whole post in one go. The workflow below splits the job: AI does research, structure and first drafts; you add experience, facts and judgement. It works in ChatGPT, Claude and Copilot, and it's how we write every post on this site.
Why most AI blog posts fail
- No real information. The AI only knows what's common online, so it writes what everyone else wrote.
- One giant prompt. "Write a 2,000-word post about X" produces padding.
- No experience. No examples, screenshots, numbers or opinions.
- No editing. The first draft gets published as-is.
Fix those four and AI-assisted posts can be better than most human-written ones, in a fraction of the time.
Step 1: Research and outline (10 minutes)
Start with a keyword you can realistically rank for (see how to find low-competition keywords). Then:
Prompt:
> I'm writing a blog post targeting "keyword" for audience. List the 15 most important questions this reader has, in the order they'd ask them. Then suggest an outline with an H1, 6-8 H2s and a FAQ section. Note which sections would benefit from a table, checklist or example.
Then check the actual top 3 Google results. Add anything important they cover that your outline is missing, and note what they don't cover well. That gap is your advantage.
Step 2: Feed it your expertise (10 minutes)
This is the step everyone skips, and it's the one that matters most. Before any drafting, give the AI your raw material:
Prompt:
> Here are my notes for this post. Use them as the main source and don't contradict them: paste your notes: experiences, opinions, real prices you checked, customer questions, examples, mistakes you've seen, tools you actually use.
Bullet points are fine. Voice notes transcribed into text are fine. The more specific, the better the post.
Step 3: Draft section by section (20 minutes)
Never ask for the whole post at once. Draft one H2 at a time:
Prompt:
> Write the section "H2" in about 200-300 words. Use my notes where relevant. Start with the direct answer in the first sentence. Use short paragraphs, plain language and one concrete example. No filler phrases, no "in today's world", no summary at the end of the section.
For the introduction:
> Write a 3-sentence introduction that states the reader's problem, promises the outcome, and gives the short answer immediately.
For comparison or "best X" posts, ask for a table first, then each section.
Step 4: The editing pass (15 minutes)
Read the whole draft as a reader would. Then:
- Cut 20%. Delete repeated points, throat-clearing sentences and anything a reader would skim.
- Kill AI clichés. Search for words like "delve", "landscape", "crucial", "elevate", "unlock", "seamless", "in conclusion" and rewrite those sentences.
- Check it sounds like you. Read a paragraph out loud. Would you say it to a customer?
- Use AI to help, not decide:
Here's my draft: paste. Point out sentences that sound generic or robotic, repeated ideas, and places where a specific example would help. Don't rewrite it; just list the issues.
Then fix them yourself, or ask for rewrites one paragraph at a time.
Step 5: Add proof (10 minutes)
This is what separates content that ranks from content that doesn't:
- Screenshots of the tool, the setting or the result
- Real numbers: prices checked today, your own results, time saved
- Your opinion: "We'd pick X because…", "Skip Y unless…"
- Examples from your own business or customers (with permission)
- A "last updated" date and sources for any statistics
Then fact-check everything the AI wrote: every price, feature, date and claim. AI tools can state wrong facts confidently.
Step 6: The SEO pass (10 minutes)
- Keyword in the H1, first 100 words, one H2, URL slug and meta description, naturally
- A 40 to 60 word direct answer under the main question H2 (featured snippet chance)
- FAQ section with schema (Rank Math or Yoast)
- 3 to 5 internal links to related posts, and links from 2 older posts to this one
- Alt text on every image
- A clear call to action (lead magnet, product or next article)
Prompt for the meta description:
> Write 3 meta descriptions under 155 characters for this post that include "keyword" and give a clear reason to click.
How long it takes
| Step | Time |
|---|---|
| Research and outline | 10 min |
| Feed your expertise | 10 min |
| Section-by-section draft | 20 min |
| Editing pass | 15 min |
| Add proof | 10 min |
| SEO pass | 10 min |
| Total | ~75 minutes for a 1,500-2,000 word post |
That's roughly a third of the time of writing from scratch, and usually a better result than either AI or a rushed human alone.
Tools that help
- ChatGPT, Claude or Copilot for outlines and drafts (compared in our best AI writing tools)
- Grammarly for the final polish 👉 Try Grammarly free (affiliate link)
- Surfer SEO to check you've covered the topic as fully as the top results 👉 Try Surfer (affiliate link)
Copy-and-paste prompt templates
Every prompt in this post, plus 40 more for repurposing each article into a week of social posts, videos and emails, is in our 300 AI Prompts for Small Business pack.
Free: get 50 ChatGPT prompts for your business, sent straight to your inbox. Get the free prompts ()
See the full 300-prompt pack ($19)
Wondering whether Google minds? Read is AI content bad for SEO?
FAQ
Can I use ChatGPT to write blog posts?
Yes. Google judges content on helpfulness, not how it was made. Use ChatGPT for research and drafts, then add your own experience and fact-check everything.
How do I make AI writing sound human?
Give it your own notes and examples, draft one section at a time, cut 20% in editing, remove AI clichés, and read it out loud.
What's the best prompt for writing a blog post?
There isn't one. Use a sequence: research questions, outline, your notes, then one prompt per section with clear rules on length, tone and structure.
How long should an AI-assisted blog post be?
As long as it needs to fully answer the question, often 1,200 to 2,500 words for guides and comparisons. Length without substance doesn't help.
Should I tell readers I use AI?
Many sites add a short note on how they create and review content. Being open about it builds trust, and Google suggests it can be useful where readers would expect it.
