My AI Writing Process: How I Use Different Tools at Each Stage

AI is now part of my everyday writing life but it doesn’t replace me. Instead, I use it as a set of smart helpers at specific points in my workflow.

In this piece I’ll walk step by step through my AI writing process, where I use AI, where I deliberately don’t, and how all of this fits together into a clear workflow.

What is an AI Writing Process?

An AI writing process is a repeatable workflow that maps out where artificial intelligence is used, where human creativity leads, and how the two interact to produce strong, authentic writing. Creating an AI writing process lets you use AI tools intentionally at the right points.

Understand when to rely on AI tools and trust your own judgement. For me, this means I still guide the ideas and structure. AI, for its part, helps me generate possibilities, summarise information, and refine language.

One approach is to see your AI suite as a small team of assistants. For example:

One is great at brainstorming,
Another excels at research summaries,
One helps me tighten my writing, and
Another keeps an eye on grammar and style.

Together, they make the process faster and more focused but the final decisions always come from me. I stay in control of the ideas and final product but AI just builds on the thinking and speeds up the polishing.

My AI Writing Workflow (Step by Step)

This is exactly how I use AI to research, draft, and refine my content.

1. Brainstorm: Generating Ideas with ChatGPT

I start most projects by brainstorming with ChatGPT. This works best when I write my ideas out as fully as possible first. Instead of typing:

“Ideas for article about AI and writing?”

I’ll say something like:

“I’m a digital content specialist writing an article about my AI writing process. My audience is marketers and writers who are curious but slightly sceptical about AI. I want to show my step-by-step workflow with specific tools. Can you help me brainstorm angles, section headings, and a few hooks?”

Because I give context (who I am, who I’m writing for, what I want), the ideas I get back are much closer to what I’d actually write. During this phase, ChatGPT helps me:

Expand half-formed ideas into clearer angles
Come up with titles, hooks, and subtopics
Spot missing angles or questions readers may have

Think of this as a whiteboard session. Pick the ideas you like and throw away the rest.

2. Outline: Human-Only

My outline is where I usually don’t use AI. I was afraid of outsourcing too much of my reasoning and thinking to AI, but I’ve come to see there’s a better, more balanced way to do it.

When you’re outlining, you want to get clear on the promise of the piece and the order in which the reader should be guided through concepts. Consider which sections stand on their own and which are supporting points. Decide where to add stories, case studies, or personal experience.

During my time at the content marketing agency Animalz, I got a masterclass in outlining from one of my editors. I learned how to set up outlines that clearly articulated the value and purpose of the piece. One key takeaway was to have two outlines, a shorter one that you created right at the start of writing the piece and a longer one that included your key ideas and supporting evidence.

Often I’ll sketch this out in a simple list:

Introduction: why an AI process matters
Definition: what is an AI writing process?
My workflow overview
Step-by-step breakdown (brainstorm, outline, research, draft, refine, edit)
Why I still edit everything myself
Final thoughts and practical tips

You could do this with or without AI. I built an AI outline generator to quickly create an outline to guide the piece, based on your writing goals.

3. Research: Google, Google Notebooks, Perplexity, ChatGPT, and Google Scholar

With an outline in place, I switch into research mode. I use a mix of tools here, each for a different purpose.

I start with Google and Google Scholar to find primary sources such as reports, official documentation, and reputable publications. Books and podcasts also feature and the goal is to consume as much information on the topic as possible.

a ) Google Notebooks

I save these into Google Notebooks, adding links, short notes, and key statistics or definitions. Once I’ve gathered enough material, I use Google Notebooks’ built-in AI features to generate an article or podcast-style summary of my collected sources. This allows me to listen to or read a concise overview of the research while I’m walking or doing other tasks — a great way to absorb ideas and spot useful connections between sources.

b) Perplexity & ChatGPT for Synthesis

Once I have my main sources, I turn to tools like Perplexity or ChatGPT to help synthesise the information. I might ask them to summarise long reports, compare different viewpoints, or explain complex ideas in plain language. I always check any references and click through to the original sources to verify the information. AI summaries are a shortcut to understanding, not a replacement for reading.

This helps turn dense, technical research into something conversational and easy to digest. I then rewrite those explanations in my own tone, always double-checking against the original paper to ensure accuracy.

4. Draft: Human-First, AI-Assisted

At drafting stage, I have options:

Option 1: I write a messy human draft myself.
This is usually my default for anything personal, strategic, or opinion-heavy. I follow my outline and just get words out, without worrying about perfection.
Option 2: I co-write with AI (ChatGPT or WriterZen).
I might:
- Give the outline and ask for a rough draft of a specific section, or
- Ask for a few variations of an intro or conclusion.

If I use ChatGPT or WriterZen to generate a draft, I always:

Rewrite passages in my own voice
Add my own stories, examples, or data
Cut any fluff or generic phrasing

The goal is not “AI wrote this article,” but “AI helped me get from blank page to something I can shape much faster.

5. Refine: Using ChatGPT as a Thoughtful Editor

Once I have a full draft, I use ChatGPT again and this time as a refining tool.

Here’s how:

I paste a section and ask:

“Can you suggest ways to tighten this section without losing my voice? Audience: [describe]. Tone: [describe]. Keep it under 300 words.”
I ask for:
- Clearer subheadings
- Smoother transitions between sections
- Suggestions for examples, metaphors, or comparisons that could help readers

Sometimes I’ll also ask:

“What questions might a sceptical reader still have after reading this?”

This is a great way to catch gaps or objections I haven’t addressed.

Again, these are suggestions, not instructions. I pick what I like and ignore the rest.

6. Edit: Grammarly + ChatGPT + Final Human Pass

The last stage is editing and here’s where we’ve been using AI for some time now.

I use a combination of tools here. Grammarly is good for catching typos and grammar issues. ChatGPT can assist with higher-level editing, I might ask:

“Is anything unclear or repetitive in this section?”
“Can you simplify this paragraph for a non-expert reader?”
“Does this conclusion feel satisfying and complete?”

Finally, I read the piece as a human:

Does this sound like me?
Is it accurate and fair?
Would I be happy to publish this under my name?

Only after that do I hit publish.

Why I Share My AI Writing Process

I’m open about my AI use for a few reasons:

Transparency: Readers (and search engines) deserve to know how content is created.
Quality: A clear AI workflow helps me keep quality high. I know exactly where my judgement is needed.
Repeatability: Following the same process makes my work more consistent and easier to improve.

AI doesn’t remove the need for real thinking, experience, and craft. It just gives me better tools to do that work faster and with more confidence.