How to Create Marketing Images With AI: Prompts and Edits
A repeatable three-step routine for turning a plain idea into a usable marketing image, then keeping every image after it consistent with the first.
How to Use an AI Blog Generator: Six Steps to Posts Worth Reading
The difference between AI blog content that ranks and AI blog content that vanishes is almost never the model. It is whether anyone did the first three steps.
ChatUp GuidesWork
“Write a blog post about email marketing” produces exactly what you would expect: eight hundred words that could have been written about anything, by anyone, at any point in the last decade. It is not wrong. It is just weightless, and weightless content does not rank, convert, or get shared.
The fix is a process. Six steps, and the first three happen before you ask for a single sentence of draft.
Step 1: define who this is for and what they get
Not “small business owners.” A person, a situation, and a specific thing they cannot currently do.
Prompt to try
I sell bookkeeping software to freelancers and one-person businesses. Help me define my audience properly. Ask me about who actually buys, what they were doing before, what they get wrong, and what they've already tried. Then write a one-paragraph reader profile specific enough that I could recognize the person.
Keep that paragraph. It goes into every brief from now on.
Step 2: generate topics from their problems, not from a keyword tool
Keyword tools tell you what is being searched. They do not tell you which of those searches you have any business answering.
Prompt to try
Here's my reader profile: [paste]. Give me 20 blog topics that come from problems this person actually has, phrased the way they'd describe the problem rather than the way a marketer would. For each, tell me what the reader wants at the end, and mark which ones I could answer better than a generic competitor because of what I know.
That last mark is the whole game. Publish the ones where you have an advantage; skip the rest.
Step 3: research before drafting
A draft written from a model’s memory is a draft full of plausible, unattributed, occasionally outdated claims. Research first, with sources.
Prompt to try
Research this topic: [topic]. I want current facts with sources, the points where credible sources disagree, and anything that changed in the last 18 months. Flag anything you're not confident about rather than smoothing it over. Don't write the post.
Read the sources. Not all of them, but the ones carrying the load-bearing claims.
Step 4: write the brief, not the prompt
This is the step that separates usable output from filler. A brief specifies angle, structure, evidence, and voice.
Prompt to try
Turn this research into a content brief for a 1,400-word post. Include: the single argument the post makes, who it's for, what the reader can do differently afterwards, an H2 outline where each section has a job, which facts go where, three things to explicitly not say because every competing article says them, and the two questions the reader will have that most posts leave unanswered.
Step 5: draft section by section
Asking for a whole article gets you an article-shaped average. Asking for one section at a time gets you something you can steer.
Prompt to try
Draft only the section "[H2 title]" from the brief. 250 words. Direct, concrete, no throat-clearing intro sentence, no summary sentence at the end. Use the specific facts from the research rather than general statements. Vary sentence length. Don't use the words "leverage", "delve", "landscape", "robust", or "unlock".
Step 6: edit like the writer, because you are
The draft is raw material. The editing pass is where the post becomes yours.
- Cut every sentence that states something the reader already knows.
- Replace general claims with specific ones — numbers, names, examples from your own work.
- Add the thing only you can add: what you have seen go wrong, what you disagree with, the caveat nobody mentions.
- Read it aloud. AI prose has a rhythm; hearing it is the fastest way to find it.
- Verify every factual claim and every link.
Prompt to try
Here's my edited draft. Point out the three weakest paragraphs and why. Then list every claim that a reader could reasonably challenge and tell me which ones I've supported and which I've just asserted. Don't rewrite anything.
Does Google penalize AI content?
Google’s stated position is that it rewards helpful, original content produced for people, regardless of how it was produced, and that using automation to generate content primarily to manipulate rankings violates its spam policies. In practice that means: mass-produced, undifferentiated content does badly, and useful content does well — which was also true before.
The practical read: the model is not the risk. Publishing at volume without adding anything is.
What this process actually saves
Not the writing. Writing is maybe a third of the time. What it saves is the research consolidation, the blank page, the structural rework, and the second draft — and it makes the good version reachable on a day when you would otherwise have published nothing.
Where this fits in ChatUp
AI Research for step 3, since a topic brief built on stale recall is worse than no brief. The Creative Writer and Marketing Expert assistants for angle and drafting. File Assistant when the source material is a report or a PDF you need to work from. Keep the reader profile and voice notes in the current conversation or paste the reviewed brief into a new one.
Frequently asked questions
Can I use AI to write my blog posts?
Yes. Whether the result is worth publishing depends on the brief and the edit, not the generation.
Does Google punish AI-generated blogs?
Google penalizes unhelpful, mass-produced content. It does not penalize AI as a tool. The distinction is real and it is enforced by content quality, not by detection.
How long should the process take?
For a solid 1,400-word post: roughly 30 minutes on steps 1–4 the first time (steps 1 and 2 are reusable), 15 minutes drafting, and 45 minutes editing. Faster than writing from scratch, considerably slower than “generate.”
Should I disclose AI use?
Follow your publication’s policy. Many outlets require it; most company blogs do not. Either way, the byline still owns the claims.
The brief is the product
If you only take one thing: the quality ceiling is set at step 4. A good brief makes almost any model produce something usable; a bad brief makes the best model produce filler. Spend your time there.
Try it in ChatUp
Turn this guide into a workflow.
Run the prompts above against the model that suits the task, keep the useful context across chats, and pick it back up on any device.
Try for Free