The most common mistake when working with Claude on content is asking for the finished result straight away: "write me 30 posts". The model doesn't know who it's writing for or what counts as a good post, so it produces generalities. Below is a chain of four prompts in one conversation that first assembles the context and only then turns it into concrete copy. Each step uses the previous one's answer — no need to open a new chat.
Why it's a chain, not a single prompt
A good content plan is built top down: first an understanding of the audience, then topics that hit their pain points, then those topics distributed across formats, and only at the end the finished copy. Skip the first steps and ask for copy right away and Claude has to invent the audience itself — and it invents an averaged-out one. The chain forces the model to gather facts first and write afterwards.
Step 1 — putting the audience into words
Don't ask for topics — ask for insights about the people you're writing for.
You are a marketing analyst. Target audience: [describe: who these people are,
what they do, what the product or blog means to them].
List their 10 deepest fears, their 10 main desires
and their 10 hidden objections before buying [product/service name].
The output of this step is raw material, not content. Those 30 items are what the topics will be built from next.
Step 2 — hook topics from pains and desires
Now we turn the list of insights into headlines that grab attention rather than sounding like ads.
Based on these fears, desires and objections, write 30 headlines
for posts. Use intrigue, paradoxes and concrete numbers where
appropriate. The headlines must not sound like advertising — they
should land precisely on one of those items.
The measure of a good result at this step: you can read a headline and immediately see which pain or desire from step 1 it addresses.
Step 3 — distributing across formats
Thirty topics with no structure is just a list. We lay them out so the feed works as a system rather than a set of random posts.
Split these 30 topics across formats: some for Reels (aimed
at broad reach), some for carousels (deep expertise
and saves), some for Stories (warming up and engagement).
Put it all in a table with the columns: topic, format, goal.
The resulting table is already a working content plan for the coming weeks, even before any copy is written.
Step 4 — finished copy for the first carousels
The final step expands specific topics into finished text you can put straight into design.
Take 2-3 carousel topics from the table and write dense
copy for them: a hard hook on the first slide, 4-5 slides
of concentrated value with no filler, a clear call to action
at the end. Short sentences, no bureaucratic language.
From there you can ask Claude to make the hook stronger, cut a particular slide or change the tone — the model holds the context of the whole chain, because it's all one conversation.
The complete chain
- Audience → 10 fears, 10 desires, 10 objections.
- 30 headlines based on that list.
- The 30 topics distributed across formats in a table.
- Finished copy for 2-3 priority topics.
You can run all four steps back to back in one sitting or spread them over several days — what matters is not opening a new chat between steps, otherwise Claude loses the audience context it has assembled.
Next level: turn the chain into a skill
Four prompts that always run in the same order are a candidate for a skill of their own: one command and Claude walks you through all four steps, asking the clarifying questions in between. How to build skills like that is covered in the skills section of the knowledge base.
How we do it
We use a similar prompt-chain principle in our own content production: not a single "make me a carousel" request but a sequence of steps where each one refines the last. That's exactly how the pipeline feeding the content hub on the qvib engine is built.
Related links
- The "Prompts" section: qvib.pro/arsenal/prompts/
- The "Claude skills" section: qvib.pro/arsenal/skills/
- How our content pipeline works: qvib.pro/