The short version: ChatGPT is brilliant for content creation when you feed it your own material, your own voice, and your own opinions, and it's rubbish when you ask it to invent an angle from scratch. The tool doesn't do the thinking for you, it does the typing, and once you accept that, everything else falls into place. Below is the exact setup and workflow I use every week for blog posts, LinkedIn, email, and video scripts.
Related reading: The Content Creation Tools That Earn Their Keep in a Small Business.
Worth reading next: How AI Tools Can Improve Your Resume Writing (Without Making It Sound .
Worth reading next: How to Get Ten Pieces of Content Out of One Interview.
Why most ChatGPT content still sounds like ChatGPT
I read a lot of AI-generated content because clients send it to me for feedback. I can spot it in the first sentence, usually because of the same three words: "in today's fast-paced world" or "in the ever-evolving landscape of." Nobody talks like that. Nobody thinks like that. The reason so much ChatGPT content sounds identical isn't the model, it's the prompt. If you type "write a blog post about email marketing tips" you'll get the same generic output that 40,000 other people got that week, because you gave it nothing to work with.
The fix is boring but it works: feed the model your own facts, your own numbers, your own opinions, and ask it to arrange them. Don't ask it to have the ideas. You have the ideas. It has the keyboard.
My setup before I type a single prompt
Before I ask ChatGPT to write anything, I do three things that take about fifteen minutes total and save me an hour later.
- I write custom instructions once. In Settings, under "Custom Instructions," I've told ChatGPT that I'm a 53-year-old UK-based marketing consultant, that I write in first person, that I hate corporate jargon, that I use British spelling, and that I want short sentences and no filler intros. It remembers this every single time I open a new chat.
- I paste in three examples of my own writing at the start of a new project. Not as an instruction, just as raw material, followed by "this is roughly my tone, don't copy the sentences, match the rhythm."
- I give it the boring specifics first. The client's actual pricing, the actual case study numbers, the actual objection a prospect raised on a call last Tuesday. Specifics are what separate content that reads like it came from a person versus content that reads like it came from a search result.
This is roughly the same groundwork I cover in more depth in my guide to B2B prompts for AI-driven content, if you want the longer version with more prompt templates.
A story from an actual Tuesday
Two months ago I had a client, a small accountancy firm in Reading with eleven staff, who needed a blog post explaining Making Tax Digital changes to their small business clients. Their previous attempt, written by a junior team member using ChatGPT with a one-line prompt, read like a government leaflet. Technically correct, entirely lifeless, zero personality.
I sat with the practice's senior partner for twenty minutes on a call. I asked him what actual clients had asked him that month. He told me a story about a client who nearly missed a deadline because she thought the rules didn't apply to her because her turnover was "only" £42,000. That one detail became the opening of the post. I fed ChatGPT the transcript of that twenty-minute call, the specific turnover threshold, and the partner's actual phrasing ("only" in quotes, because that's how people talk), and asked it to draft an 800-word explainer using that story as the hook.
The first draft took four minutes to generate. It took me another 35 minutes to edit: tightening sentences, cutting three paragraphs that repeated the same point, and adding one more real example the partner had mentioned but ChatGPT had left out because I hadn't flagged it clearly enough in the prompt. Total time from call to published post: about an hour. Without ChatGPT, that post would have taken the junior team member half a day, and it still wouldn't have had the client story in it, because nobody would have thought to ask for it.
That's the actual value. Not "AI writes your content for you." It's "AI turns your raw material into a draft fast enough that you'll bother doing the interview in the first place."
The prompt structure that works
I use roughly the same skeleton for every piece of content, whether it's a blog post, an email, or a LinkedIn post. It has five parts:
- Who it's for. Not "small business owners," but "small business owners with 5 to 20 staff who've tried and failed to use AI before and are sceptical."
- What they already believe, wrongly. This is the part everyone skips. Give ChatGPT the misconception you're correcting and the output improves dramatically, because now it has an argument to make instead of a topic to summarise.
- The specific facts, numbers, or stories to include. Paste them in as bullet points, don't make the model guess.
- The format and length. "800 words, three subheads, one list, no conclusion that starts with 'in conclusion.'"
- The tone reference. Point back to the examples you pasted earlier in the chat.
A real prompt I used last week: "Write a LinkedIn post for solo consultants who think ChatGPT will replace their expertise. They're wrong, but the fear is reasonable given how AI is marketed. Use this fact: 73% of consultants I surveyed in my own newsletter list said they'd tried ChatGPT once and never gone back, because their first prompt was too vague and the output disappointed them. Make the point that the tool didn't fail, the prompt did. 150 words, no hashtags, end on a question."
Step by step: my actual weekly workflow
Here's the process, start to finish, for a single blog post:
- Step 1: Brain dump, 10 minutes. I talk into my phone's voice memo app about what I think on the topic, no structure, just talking.
- Step 2: Transcribe and paste. I use ChatGPT's voice or a separate transcription tool, then paste the raw transcript into the chat.
- Step 3: Ask for a structure, not a draft. "Based on this transcript, suggest three possible headline angles and an outline for each. Don't write the post yet."
- Step 4: Pick the strongest angle and ask for the first draft, using the prompt skeleton above.
- Step 5: Edit on paper or read it aloud. If a sentence makes me wince when I say it out loud, it goes. This step is not optional and it's where most people cut corners.
- Step 6: Run it past one more prompt: "What's the most obvious objection a sceptical reader would have to this, and have I addressed it?" This catches gaps I've missed because I'm too close to my own argument.
- Step 7: Publish, then repurpose. One blog post becomes a LinkedIn post, three tweets, and an email in the next twenty minutes, using the same source material.
If you want the fuller repurposing side of this, I've written up my exact repurposing workflow that turns one recording into seven assets, which slots in directly after step 7 above.
The bit that makes people uncomfortable
Here's the uncomfortable truth: most people using ChatGPT for content are trying to skip the part where they have an opinion. They want the tool to generate the point of view as well as the sentences, and that's the one thing it can't do well, because it has no experience, no clients, no failed campaigns, no Tuesday afternoon call with a nervous accountant in Reading. It can only rearrange what you give it. If you give it nothing but a topic, you get nothing but a summary of the top ten Google results on that topic, which is exactly why so much AI content reads the same. It's not copying each other. It's copying the same average.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The people getting real results from ChatGPT right now aren't better prompters, they're better thinkers who happen to be using a fast typist. That's a harder skill to build than learning prompt syntax, and it's why "AI will replace content creators" has turned out to be only half true. It's replaced the ones who had nothing to say in the first place.
Which content types this works best for
Not every format benefits equally.
- Blog posts and articles: excellent, especially first drafts and repurposing, as long as you supply the specifics.
- LinkedIn and social posts: very good for tightening and shortening something you've already drafted or dictated.
- Email sequences: strong, because tone consistency across five or six emails is exactly what the model is good at once you've set the voice.
- Video scripts: decent for structure, weak for delivery, since spoken rhythm and written rhythm aren't the same thing.
- Case studies: only as good as the interview you feed it. Garbage transcript in, garbage case study out.
If you're still working out what "content creation" even means for your business versus what it should mean, I've gone into that at length in what content creation is about after 15 years of doing it, and the free content question specifically in what free content means once you strip out the marketing textbook definition.
Mistakes I still see people make
Three come up constantly when I audit client content:
- Publishing the first draft. The first draft is a starting point, not a finished piece. Every single time I've published without editing, I've regretted a line within a week.
- Asking for opinions the model doesn't have. "What does the market think about X" gets you a plausible-sounding average, not real market intelligence. Ask your own clients instead, then feed those answers in.
- Using one giant prompt for everything. Long, over-engineered mega-prompts with fifteen instructions tend to produce muddled output. Short, specific, sequential prompts in the same conversation beat one enormous instruction every time.
If you're building this out at team or agency scale rather than solo, the workflow changes a fair bit, and I've written separately about building an AI content team that can outperform competitors, plus a broader look at using ChatGPT across the full content marketing function in 2026, not just individual posts.
How much time this saves
I timed myself over four weeks last autumn: 22 pieces of content, blog posts and LinkedIn combined. Average time without ChatGPT, based on my old habits, was roughly 2 hours 15 minutes per blog post and 20 minutes per LinkedIn post. With the workflow above, blog posts averaged 55 minutes including editing, and LinkedIn posts averaged 8 minutes. That's not "AI does it instantly," it's roughly a 55 to 60% time saving, and every minute of that saving came from not staring at a blank page, not from skipping the thinking.
Frequently asked questions
Can ChatGPT write an entire blog post on its own with no input from me?
It can, but it will read like every other AI blog post because it has no specific facts, stories, or opinions to draw on. Feed it your own material first and edit the output; skipping both steps is the main reason AI content gets flagged as generic.
Does Google penalise content written with ChatGPT?
Google has said publicly it cares about quality and helpfulness, not whether a human or a tool typed the words. The problem isn't AI detection, it's thin, unedited, opinion-free content, which ranks badly whether a person or a model wrote it.
What's the single biggest mistake people make using ChatGPT for content creation?
Publishing the first draft without reading it aloud first. Anything that makes you wince when spoken out loud will make a reader stop trusting you, and that step gets skipped constantly because it feels slower than it is.
Is it worth paying for ChatGPT Plus for content creation, or is the free version enough?
The free version is fine for occasional short pieces. If you're producing content weekly, the paid version's longer memory, faster responses, and ability to hold a longer back-and-forth in one conversation make the workflow above considerably smoother, and the cost is small compared to the hours it saves.
Related reading: The Best AI Productivity Tools for Entrepreneurs in 2026 (What I Use, Daily) and The Best AI Tools for Small Business Owners (What I Pay For).
This builds on my main AI marketing guide, my main guide on the topic.
Sources worth reading
Pair it with 50 free content creation tools worth trying.
