Ask ChatGPT for twenty blog ideas and it will hand you twenty. That is the problem. They will fit any business in your field, which means they are twenty ideas nobody needed you to write.
You know the moment. It is Tuesday, you promised yourself you would post something this week, and the cursor is blinking. So you open ChatGPT and type “give me content ideas for a massage practice”. Out comes “5 Benefits of Massage”, “How to Relax After Work”, “Self-Care Tips for Busy Professionals”. You read them, feel nothing, and close the tab. Nothing is wrong with the model. The fix is not a better AI or a cleverer prompt template. It is one piece of information that only you have, and almost nobody thinks to hand over.
Why ChatGPT gives you generic content ideas
A language model can only be as specific as the thing you give it. When you type “content ideas for a massage practice”, you have named a category, not a business. So the model does the only sensible thing: it answers for the average of every massage practice it has ever read about. The average is generic by definition. That is not a flaw in the tool, it is arithmetic.
This is also why switching models does not rescue you. Claude, Gemini and ChatGPT will all produce a competent, forgettable list from a category prompt, because the limit is the input and not the engine. We went into that trade-off in detail when we compared which AI models are actually reliable for copywriting and email, and the conclusion was the same: what you feed it decides more than which one you pick.
What should you feed it instead?
Every guide on this subject tells you to add context, give the model a role, and be more specific. That advice is true and almost useless, because none of them tell you which context. Your brand voice does not fix idea generation. Neither does your industry, your tone of voice document, or a list of your services. Those describe you. They do not describe the thing readers care about.
The context that works is the one thing a model can never have: what actually happens to a person who works with you. Not what you sell. The change. Where somebody stood the week before they found you, and where they stand three months later. Fill that in honestly and you are no longer asking a machine to invent topics. You are asking it to read a map you drew.
Map your customer before and after
Pick one real client. Not a persona, not an “ideal client avatar” with a stock photo and an invented name. Someone whose actual name you know, ideally a recent one. Then write two short columns.
Before, meaning the week before they contacted you. What did they have? How did they feel? What did a normal day look like? What had they already tried that did not work? What were they afraid of? And what exactly did they say in that first conversation?
After, three months in. What do they have now? How do they feel? What does a normal day look like? And what do they say about it now, in their words?
The only rule that matters here is to stay concrete. Not “she was stressed”, but “she checked her phone at six in the morning to see whether anyone had booked”. Not “he felt more confident”, but “he stopped apologising for his prices”. Vague notes produce vague topics, and you will have gained nothing.
The two-column idea itself is old, it comes straight out of direct response marketing. What makes it work here is that you fill it in about a real human being, so it ends up full of details no model could ever guess.
The gap between them is your content calendar
Now look at the space between the two columns. Everything in there is an obstacle, and every obstacle is a topic. You are looking for four kinds in particular.
- What they did not know. Information they were missing, which kept them stuck longer than necessary.
- What they believed that turned out to be wrong. These make the strongest posts, because your reader probably believes it too.
- What they were afraid of. Usually money, usually wasting it, usually having been burned before.
- What they tried that failed. The cheaper route they attempted first, and why it did not hold.
Each of those converts into more than one piece. Here is what that looks like in practice for a coaching or therapy practice.
| Obstacle from the map | Blog post | Social post |
|---|---|---|
| “I thought I just needed more discipline” | Why discipline is the wrong diagnosis for burnout | The sentence clients say in week one, and what it usually hides |
| Tried a cheap online course first | What a self-paced course cannot do for you | Short story: what happened after course number three |
| Afraid it would be another wasted six months | How to tell in session two whether this is working | What progress actually looks like at four weeks |
| Did not know what a session involved | What actually happens in a first session, step by step | A walkthrough of the room and the first ten minutes |
Notice what these have in common. None of them could be written by a competitor, because they come from your client rather than from your category. And they are phrased in the words a real person used, which happens to be the same words the next person will type into Google or into ChatGPT when they ask about your field.
The ChatGPT prompt for content ideas, step by step
Fill in the map first, on paper or in a note, and give it fifteen honest minutes. Then paste the whole thing into this prompt. It is written to refuse to guess, which is the part that keeps the output specific.
You are helping me find content topics for my business. Do not invent anything. Everything you produce must come from the notes below. Here is one real client, before and after working with me. BEFORE (the week before they contacted me) - What they had: - How they felt: - What a normal day looked like: - What they had already tried: - What they were afraid of: - What they said to me in the first conversation, in their words: AFTER (three months later) - What they have now: - How they feel: - What a normal day looks like: - What they say about it now, in their words: Step 1. List every obstacle that stood between BEFORE and AFTER. Include what they did not know, what they believed that turned out to be wrong, what they were afraid of, and what they tried that did not work. Use their wording, not marketing language. Step 2. For each obstacle give me three things: - one blog title that answers it directly - one social post angle that opens with the obstacle in their own words - the question someone would type into Google or ChatGPT to find it Rules: nothing that would fit any other business in my field. If a detail in my notes is too vague to build on, ask me for the missing piece instead of guessing.
That last rule is doing a lot of work. A model asked to be specific from thin notes will quietly fill the gaps with plausible invention, and you end up back where you started. Told to ask instead, it hands the vagueness back to you, which is where it belongs.
How do you never run out of content ideas after that?
One map usually yields between fifteen and thirty obstacles, which at a sane publishing rhythm is months of material. The question stops being what to write and becomes what to write first.
Start with the obstacles closest to the moment they said yes. The last doubt somebody had before booking is the doubt your next reader is sitting with right now, and content that removes it earns money rather than applause. Work backwards from there toward the early confusion, which pulls in people who are still orienting.
Then stretch each one. A single obstacle is comfortably a blog post, three social posts and one email, because you are approaching the same problem from different distances rather than repeating yourself. If turning that into an actual week of posts is the part you dread, we wrote a separate piece on batching a month of content in one morning.
Redo the map after every three or four new clients. Your customers change, their obstacles change, and a map from two years ago will quietly drift back toward generic. If you would rather have someone build this into your site and your calendar instead of doing it alone, that is the kind of thing we set up for clients, and a free website check is a reasonable place to start the conversation.
Frequently asked questions
Why does ChatGPT give such generic answers?
Because the prompt described a category rather than a situation. Asked about “a massage practice” the model answers for the average of every massage practice in its training data, and an average is generic by definition. Give it details that are true of one real client and nobody else, and the output narrows immediately.
What should I feed ChatGPT so it gives content ideas only my business could publish?
Your customer’s before and after state. Write down where one real client stood the week before they contacted you and where they stood three months later, in concrete detail and in their own words. The obstacles between those two points are topics no competitor can copy, because they did not have that client.
What should I post if I have no idea what to write about?
Answer the last question a client asked you before they booked. It is already proven to matter to somebody who was about to spend money, and you have already answered it once out loud. That single habit produces better posts than any list of prompt templates.
How many content ideas do I actually need?
Far fewer than the idea lists suggest. At one or two posts a week, fifteen to thirty obstacles from one customer map covers several months. Going faster is rarely the constraint, and Google treats mass-produced content as a quality problem regardless of who or what wrote it.
Will AI-written content hurt my Google rankings?
Not by itself. Google judges the page, not the tool that produced it. What does hurt you is publishing the generic output described above, at volume, because it is indistinguishable from everyone else’s. Content built on details only you have is the opposite of that problem.
The bottom line
Not knowing what to write is almost never a creativity problem. It is a symptom of never having written down, in plain language, what actually changes for the people who work with you. The moment you do, the topics are simply sitting there, and the AI turns into what it is good at: a fast pair of hands for a map you drew yourself.
So take one client, two columns and fifteen minutes before you next open a blank post. If you would like a second opinion on whether your site is set up to turn that content into enquiries rather than just traffic, start with a free website check. We will tell you honestly what is working and what is decoration.
