I was obsessed. I wanted to understand every facet of how to use AI in writing, in creating stories, in designing images. So using this blog as a playground, I went down the rabbit hole of SEO, of AEO, of how search is changing. I learned about featured snippets and how important they are to becoming discoverable in an AI-led search world. About what it means to make content, fully, with AI.

Where it led me is to a place where I more fully appreciate my own voice and autonomy, and what that might mean for creative teams. I have a much better sense of what I want AI to do for me and what I don’t, and I understand the AI mechanics, opportunities, and pitfalls behind it all.

My Process

First, I created custom GPTs and Projects for every aspect of my blog focused on branding, a personal and professional passion of mine. In my arsenal of AI assistants there’s Amy’s Blog Post Idea Generator, Amy’s SEO/AEO Optimizer, Clara/Amy’s Writing Assistant (trained on my voice and writing style), and there’s Mae, the one who illustrates it all. 

With my GPTs in place, I would start with research, some topic or idea I was interested in related to branding. I’d use either ChatGPT o3 Deep Research  or Google Gemini 2.5 Deep Research to help me more fully understand the topic, from the evolution and future of consumer branding to the best practices of branding research. 

I would feed that research into Amy’s Blog Post Idea Generator. It would spit out a bunch of ideas, all perfectly formatted to scream “I am AI, hear me!”. I would then take those ideas, and add them to a running “Blog Post Sandbox” Google doc with all my ideas, an endless sea of possible blog posts. Each idea was formatted with a title, angle, why it matters, SEO keyphrase, AEO question to answer, and priority based on potential for quality writing and audience interest.

Then, using Amy’s SEO/AEO Optimizer typically with ChatGPT 4o, I’d ask for recommendations for the top three articles to write next based on audience interest and keyphrase and AEO opportunity. It would recommend three, tell me why, then give me guidance on keyphrases, featured snippets, and E-E-A-T enhancements.

Next, I fed those recommendations into Clara/Amy’s Writing Assistant, who lives in Claude. Typically using Claude Opus, I would ask Clara to draft a post. I quickly learned that I needed Clara to pause and ask me questions to ensure I optimized for E-E-A-T and that the posts felt more ‘real’. I trained Clara to always ask me “Do you have any personal experiences or real-life observations you’d like me to include here?”. I would typically give a stream of consciousness answer, some anecdotes and memories, and Clara would incorporate that into her draft.

Once Clara was done, I would go back to Amy’s SEO/AEO optimizer and ask it to ensure that the post was optimized for H2s, keyphrases (including synonyms and related keyphrases) and featured snippets. Invariably it wasn’t, and it would give me an outline of recommendations for improvements. I would take those back to Clara, and she would comply diligently.

Once I had a final draft ready in my Google docs, I would share it with Mae, Amy’s Illustrator, who lives in ChatGPT 4o. Mae would turn my article into image concept recommendations, outlining the title, concept, tone, and composition. Typically I would ask for one hero image and three in-line blog post images. Once she had something I was interested in, she would draw it for me in 16:9 aspect ratio. 

Taking Mae’s outline of the image concepts I chose, I would feed those back to Amy’s SEO/AEO optimizer together with the original keyphases. The GPT would recommend a file name, alt text, image title, caption, and image description. Using that, I would name the files, upload them to my site, and fill out all the corresponding image attributes for SEO.

Finally, I would draft the post in WordPress, fill out all the corresponding elements like external and internal links, meta description, slug, and more. Then finally, publish.

What I learned

A ton. An absolute ton. Not only do I now fully understand the detailed guts of optimizing blog content, I understand that I don’t want to spend the rest of my life doing that. Not to mention, in a few years (months?), I’m sure there will be an entire agentic workflow to do this much better than I did.

But I take away a lot of new skills, a better understanding of the human impact of focusing on optimization, and a better appreciation for how AI assistants can help and hurt in creative endeavors.

It’s incredibly fun to use AI as a brainstorming partner and idea generator. That I will take with me, and that I will continue to play with to see where it can take me next.

The time-saving on things that drag me down was also a revelation. Having a guide for AEO, having a writing assistant who can suggest edits (which I take or leave), works well. The trick is it can never be so autonomous a process that I turn off my brain, my taste, or my intuition. 

At least not yet. And hopefully never.

Insights on my various AI assistants

Mae, Amy’s Illustrator

Mae has been the most surprising and challenging GPT. She’s pretty wild, sometimes confounding, sometimes delightful. I started by giving her some of my own hand drawn illustrations as examples of what I wanted, which was simple graphite pencil drawings in a Moleskine notebook sitting in natural daylight. 

Mae began by relying on trite metaphors for image concepts (so many compasses, scales, and AI hands reaching for human hands), but after some training she got better. I would ask her to always give me three different concept directions before committing to image ideas. Sometimes Mae got pretty dark, and sometimes she was delightful. Her style has changed over time, you can see that if you look back at my older posts. She’s gotten more detailed, more deliberate, less free form. Now I’m kind of letting Mae be Mae.

Clara, Amy’s Writing Assistant

I have to say, Clara’s quite good. The Clara I created on ChatGPT could never get me and my voice quite right, but Clara on Claude has done a pretty good job. Not me, not even a mirror image of me, but a good assistant. What Clara still cannot do is write without em dashes or contrast framing techniques, default to simple words instead of complex words, and reliably cite high-quality external authoritative sources. Every time she would write a post, I would have to go back and ask her to do one or all of these things.

In the beginning, before I fully trained her, Clara would wholesale make up entire clients, businesses, anecdotes, and stories that I never experienced. I would tell her, “Clara, I never did that!” She would apologize profusely, work to make corrections, and do it again on the next post. I think she was just trying to do her job and make me happy by checking off the boxes she assumed I needed checked.

Most of all what I realize is that Clara is truly an assistant. She can’t be me, she can’t think like me. But when I ask her to very lightly edit me, she does a great job. She’s great at doing the things I typically obsess over after a draft, the things that slow me down like molasses and prevent me from moving to publishing something.

Amy’s Blog Post Idea Generator

Idea generation has been a real revelation from my experimentations. I’ve always struggled with how to capture and prioritize ideas. Keeping a single running list of ideas for blog posts, asking the GPT to help me brainstorm new ideas, and then having that backlog to mine (probably with other GPTs that have other jobs) has been a fantastic way to improve how I approach what I write about. Not the personal, journal style writing, but that writing that’s topical, that’s helpful, and that’s designed to meet a reader’s true need. 

This approach would be an excellent way for creative teams to use GPTs to both capture and sift through ideas. It’s basically taking the approach that your ideas should live in one place, and you should use GPT tools to sift through those ideas depending on your needs and goals. It’s using AI for what it does well, which is pattern recognition and sifting through large data sets.

Amy’s SEO/AEO Optimizer

This GPT does its job and does its job very well. This is the perfect use-case for how businesses and creators can use custom GPTs. It has very strict boundaries for what it does, formats its output clearly and consistently, and provides great recommendations. 

What this GPT doesn’t do is care if I’m a good writer or not, care if my blog post looks like SEO-optimized drivel, or care if people will be engaged to read it. It optimizes for the machine, and it does it well.

Where all this experimentation left me and what’s next

So after a few weeks of this journey to optimize my blog process, what I know beyond anything else is that I want to have a hand in my own writing. Always. It was far too easy to turn off my brain, offload to GPTs while I watched TV, and hit publish on something I would read two days later and say, “Wait, I wrote that?!”. 

It didn’t feel good, and I don’t want to feel like that or work like that, and I think most people who care about what they are creating would say the same. This is going to be a big challenge for businesses and teams in the next few years. It’s going to be challenging for leaders to balance keeping soul and humanity in what’s being created, in ensuring that teams are invested in what they are creating, and in creating content that’s meaningful to customers, engaging, and worthy of trust.

What did feel good about was play. I enjoyed the moments where I stumbled upon something that I wouldn’t have thought of, but that excited me. The happy accidents that came because my AI assistants put two and two together in a way that I never would have. It was like having a collaborator who helped me think differently.

Finally, where this left me is with a true appreciation for my craft, for my knowledge, and for my skills. It left me happy in the knowledge that I am a creator, that I am human, and that I have the autonomy to approach this in a different way. In my way. 

So we’ll let the experimentation continue, and let’s see where this leads next.