The guide on turning "write me a blog post" into a brand-ready prompt
These prompts are based on the How AI Maturity Levels Transform Your Marketing Work Week webinar, published on the AI Lab by ActiveCampaign.
Get the quick-start guide
What will I accomplish with this guide? By the end, you’ll have rewritten a generic blog-post prompt into a specific, file-grounded prompt that produces on-brand copy and then validated and optimized that draft for AI search. The flow is the one Pam Didner walked through in this webinar: the same four-rung climb that takes “write a blog post for me” from a passable draft to a piece that reflects your brand, your competitors, and your search intent.
Before you start, you’ll need:
- A working ChatGPT, Claude, or Active Intelligence account that accepts file uploads
- A blog-post topic you actually need to write
- Reference files for upload: brand guidelines, a keyword list, a buyer persona, a couple of your own past blog posts on the topic, and 2–3 competitor blog posts on the same topic
- A short-form description of your product and what it does for the target reader
Quick reference
- Total time: 45–60 minutes for a 1,000-word post (compared to several hours from scratch)
- Tools needed: ChatGPT, Claude, or Active Intelligence (anything that accepts file uploads)
- Key output: A 1,000-word blog post that reflects your brand voice, differentiates from competitors, and is optimized for AI search
Watch this section
For full context on the following topics, watch these sections of the webinar:
- Pam’s prompting ladder, from generic to specific — [22:51]–[26:00]
- The intermediate-level prompt (file uploads + competitor scan) — [24:31]–[26:05]
- Validation and AI search optimization (GEO) — [26:05]–[27:55]
The workflow
Phase 1: Start with the generic prompt and feel its limits
After this phase, you’ll have: a baseline draft you can compare every later iteration against.
- Write the simplest version of your prompt: open ChatGPT and type something like “write a blog post for me about the benefit of email marketing for small business.”
- Read the output as if you were the target reader: note where it’s vague, where the voice is wrong, and where it could be about any company in your category.
- Keep this draft for comparison: the whole point of climbing the ladder is to see the difference. Save the generic draft so you can compare it against the version you build over the next three phases.
Phase 2: Specify the audience, tone, length, and call to action
After this phase, you’ll have: a more usable draft that at least sounds like it was written for your reader.
- Set a word count: “Write a 1,000-word blog post about the benefit of email marketing for small business.”
- Name the tone and tie it to your product: “Use a friendly, educational tone that fits our product, [PRODUCT NAME].”
- Specify the audience: name the vertical and region. “Small business owners in [VERTICAL] in [REGION].” Generic personas produce generic copy.
- Surface the product benefit: “Emphasize how [PRODUCT NAME] helps them [SPECIFIC OUTCOME].”
- End with a real call to action: “End with a strong call to action to start a free trial.”
- Compare against the generic draft from Phase 1: the difference between rung one and rung two is the floor. This is the level most marketers stop at.
It will write the blog post, but that’s very generic.
Phase 3: Upload your files and let the model brief itself
After this phase, you’ll have: a draft that reads like your brand wrote it, not like a generic AI did.
- Upload your brand guidelines: voice, tone, banned phrases. The model uses this to keep the draft on-brand.
- Upload your keyword analysis: the actual keywords you want to rank for, not the ones the model would guess.
- Upload your buyer persona: so the draft speaks to the person who actually buys.
- Upload 2–3 of your own past blog posts: the model picks up your sentence rhythm and structure from real examples.
- Upload 2–3 competitor blog posts on the same topic: this is the rung most marketers skip, and it’s the one Camille flagged as the standout.
- Tell the model to scan your website and the competitors’ websites: for messaging on your side, for differentiating opportunities on theirs.
- Re-run the prompt: “Analyze the uploaded files and the websites I named, then write a 1,000-word blog post on [TOPIC] that reflects my brand voice and finds an angle the competitors haven’t covered.”
Phase 4: Validate and optimize for AI search
After this phase, you’ll have: a final draft scored against AI search, with a refined CTA, sharper headline options, and a keyword list cross-checked against your own.
- Ask the model to score your draft: “Score this blog post for generative engine optimization (GEO) on a scale of 1 to 10.”
- Ask for specific steps, not vague suggestions: “Give me specific steps (not ‘identify ways’) to improve clarity, structure, and flow.”
- Cross-check keywords: with your keyword file already uploaded, ask “Suggest keyword optimization opportunities,” then compare the model’s list against your own to find gaps.
- Strengthen the call to action: “Strengthen my call to action and explain why your version is stronger.”
- Generate headline options: “Give me three headline options for this post and tell me which one will perform best for [AUDIENCE].”
- Apply the changes you agree with, keep your point of view: Pam’s rule is that AI is a constructive critic, not the final editor. You don’t have to take everything the model says.
From my perspective, the validation part is something marketers fall a little short on. We need to do that a bit more.
Related
More data from the AI Lab.