The AI Content Balancing Act

One of my biggest goals for 2026 when we went into it was to really get better grooved into the use of AI. To learn the ropes and integrate it more fully into my business where it makes sense.

It is absolutely a balancing act.

ESPECIALLY when it comes to content creation.

This last week, I was also deep into actually setting up Claude to do actual work for me. Connecting it up with sites. Connecting it to my tools. And basically setting it up to act as my digital employee.

It is exciting as hell. Not gonna lie. 😎

But, the balancing act is real. Because, as you will see in a moment (in this week’s first feature article of the newsletter), one of my long-time subscribers could kinda tell.

Let’s dive into it. Because, it is quite interesting…

From Writer to Director: The AI Content Balancing Act

I got an email last week from a subscriber who’s followed me since 2013.

She told me she could tell I was using AI to write my content. Not in a mean way — she framed it as a token of appreciation, actually. She knows my voice. And lately, she said, she’s been hearing an “actor” instead. In fact, here’s what she said:

“i totally get using AI for copy. just know that it’s no longer you. it’s well written (as in: it makes sense, it was even a good read – and accurate). but it’s glib. and i may well be the only one noticing: i’m an energy healer. ‘energy’ is my thang.”

I appreciated the honesty. And I’m not hiding anything here. Yes, I wrote the last issue using AI as a tool. This is something I’ve written about. I talk about it openly, and I’m going to keep using it. That’s not changing.

But her note did put a finer point on something I’ve been working through: using AI effectively for content is not the same as using AI to produce content.

There’s a difference. And working out that balancing act is DEFINITELY something you need to do if you’re going to use AI in your content production.

Because otherwise, it looks like AI slop. And it comes off as artificial.

She noticed.

The director shift

I’ve talked before about the idea of moving from doing every task yourself to directing — setting the vision, making the calls, assembling the pieces. AI accelerates that shift dramatically when it comes to content.

But here’s the thing… if you hand AI a topic and walk away, you get competent output. Readable. Even good. But it’s not you. The thinking is the algorithm’s. The choices are the algorithm’s. You’re just the one who hit publish.

This isn’t some kind of secret. I’ve known this. In fact, this is the main reason I was against using AI to “write” content for a long time.

But, we can’t just sit there like a stubborn mule and refuse to use modern technology. 😇

The goal — what I’m actively working toward — is using AI as a magnifier of my thinking, not a replacement for it. That’s a systems problem, not a tool problem.

What that actually looks like in practice

A few things I’m working on, and that you might consider for your own content workflow:

• Give AI your raw thinking first. Before I ask for a draft, I brain-dump. A few sentences of my actual take, my gut reaction, what I actually think about the topic. I usually dictate it, so it is just me talking. The AI transcribes it and gets my raw words.
• Load the context. The more AI knows about how you think, your audience, your voice, your past work — the closer it gets. It’s never going to be perfect, but it gets meaningfully closer when it has real reference points to work from. I dumped a raw CSV of all my past newsletter issues into my Claude project as “memory”. I also archive past issues in a label in Gmail. AI can read a TON of my own history to learn from.
• Flip the direction. Here’s one I’m experimenting with: instead of just asking AI to write for me using prompts, I’m setting it up to prompt me back. Give me a first pass, then ask me questions. Challenge my reasoning. Flag where I sound generic. That back-and-forth pulls my actual opinions to the surface rather than letting the algorithm fill the gaps. This way, I’m still writing the article, but I’ve got a second set of eyes.
• Edit like it matters — because it does. Whatever comes out of the AI workflow still needs your hands on it. Not a quick skim — real editing. The places where you pause, where something sounds slightly off, where you’d never say it that way — those are exactly where your voice goes back in. That’s not a bug in the process. That’s the process. You can “train” AI on common phrases you use. You can add to memory the things it says where you would never say it like that.

The bottom line…

AI is not going to sound like you. Not fully. Not the way someone who’s read you for twelve years would recognize. It’s an algorithm. And algorithms don’t have a point of view, a story, or a history with your audience. They don’t have that “randomness” that makes us humans…. humans.

What AI can do is handle the scaffolding (<– a word AI likes to use, I noticed, but it works 🤣) so your energy goes where it matters most. The opinion. The angle. The thing only you would say.

The balancing act is real, and I don’t think anyone’s fully cracked it yet… including me. But the direction is clear…

You want to be the director, not the typist. Don’t be afraid to use AI as a team member because you may be scared somebody will notice.

Embrace it. Using AI to assist with your content production flow has far too many benefits to refuse out of arrogance. Instead, we need to learn to work it into our systems in a way which preserves who we are.

WordPress News & Updates

• 🚨 BuddyBoss Server Compromised — Malicious Updates Pushed to 27,000+ Sites. Attackers compromised BuddyBoss’s update server and pushed credential-stealing malware disguised as legitimate plugin and theme updates, actively compromising hundreds of sites before it was caught. BuddyBoss says the issue is resolved, but affected sites need a close look. Full story and what to check. Honestly, this just goes right to the top of the list of reasons why I think moving away from BuddyBoss is probably a good idea.
• WordPress 7.0 RC1 Delayed — Final Release Still April 9. RC1 was pushed back five days due to concerns around real-time collaboration performance, package bloat (the build was nearly twice the size of 6.9.4), and client-side media processing — which got pulled from 7.0 entirely. Real-time collaboration will ship turned off by default. April 9 release date remains on track. What caused the delay
• AI is Flooding the WordPress Plugin Directory. The Plugins Team is calling for help — weekly plugin submissions have surged from 100–150 a couple of years ago to over 500 now, almost certainly driven by AI-assisted development tools. The backlog is growing faster than volunteers can review. Read the full story. Wonder how many of these plugins were “vibe coded”. 😜
• FluentCommunity 2.3 Released. Solid update for FluentCommunity users. New features include custom member profile fields (which is great to see), CSV export for course and quiz data, and the ability to trigger community messages directly from FluentCRM automations. See what’s new in FluentCommunity 2.3. Switching from BuddyBoss is pretty easy, FYI.
• WordPress Playground Now Supports MCP. The browser-based WordPress environment now supports MCP, which means AI agents like Claude and Gemini can install plugins, run PHP, and manage sites directly in the browser. Interesting territory for anyone building AI-assisted WordPress workflows. More on Playground’s MCP support
• Featured Plugins Experiment Delivers 26,000 New Installs in Two Weeks. The rotating Featured Plugins tab — new this month — is already working. The first cohort of lesser-known plugins picked up 26,000 combined new installs in the first two weeks. Good news for plugin discovery on WordPress.org. Full results from the first cohort
• FluentCart Hits 5,000 Active Stores. WPManageNinja’s eCommerce plugin crossed 5,000 active installations this week. Still a young product but growing steadily — worth watching if you’re keeping an eye on Fluent ecosystem tools. FluentCart on WordPress.org
• WooCommerce 10.6 Released. Performance-focused release with faster checkout and admin pages (fewer SQL queries), refined Cart and Checkout blocks, expanded Product Collection options, and new EU tax-inclusive shipping pricing support. What’s new in WooCommerce 10.6

Behind the Curtain: How I Set Up An AI Newsletter Assistant

OK, let’s get tactical…. Because “set up your AI well” is advice that doesn’t mean much without seeing what it actually looks like. It is like saying “you should really climb Mount Everest”. 🤣 Feels a little…. tough. 😏

So here’s a behind-the-scenes look at the actual system I use to produce this newsletter with AI assistance — including the things I’ve built specifically to keep it sounding like me.

And keep in mind, this is always a work in progress. As I learn more, it gets better.

Also, while I’m using my newsletter as the example, the same thinking applies to whatever content you produce. Blog posts, YouTube scripts, podcast outlines, social content — the principles are the same. Context is what makes AI output actually useful.

The Claude Project

I use Claude as an assistant for newsletter production. It is one of the many roles in my business that I am building AI systems for.

I do it inside a dedicated Claude “Project” — which is basically a persistent workspace that maintains context across multiple conversations. Each issue of The WP Edge gets its own chat inside that project. And the project itself has instructions that act as a persistent memory – telling it what to do and HOW to do it.

This matters because a generic AI conversation starts from zero every time. A Project starts from your foundation every time.

You could set up a separate project for each type of content you produce. One for your newsletter. One for blog posts. One for YouTube scripts. Each one tuned specifically for that context and format.

Project Instructions: Teaching AI how you think

The single most important thing I’ve done is write detailed project instructions that live inside the project permanently. Think of it as an onboarding document for a new team member — except the team member is an algorithm.

Mine cover things like:

• Who my audience is (solopreneurs using WordPress — coaches, course creators, consultants)
• My voice and style notes (conversational, opinionated, short paragraphs, self-deprecating humor, emojis used naturally)
• The newsletter’s structure, section by section
• Specific formatting details that are easy to get wrong if AI doesn’t know them

The more specific, the better. Vague instructions produce vague output. When you tell AI exactly how you think, who you’re talking to, and what your non-negotiables are — it gets meaningfully closer.

This isn’t one huge “magic prompt” that I had to be gifted enough to think up all at once. Instead, I work out these instructions with the help of AI. Over time, I correct it, tell it what I want. Each time, I give it a new rule and tell it to update the project instructions. Those project instructions are actually AI-generated. I’d HATE writing those! But, through the iterative process of me dictating what I want and correcting it over time, we dial it in. Together.

Memory: Feeding it your history

I’ve also loaded a CSV export of past newsletter issues into the project as a reference file. Hundreds of past issues. That’s not just style reference — that’s pattern recognition at scale. Topics I’ve covered, angles I’ve taken, phrases I actually use.

I also archive every sent issue in Gmail, which Claude can access directly (through a connector I’ve set up that allows Claude to access my email). So when it needs to re-calibrate my voice or check what I’ve already covered, it’s reading my actual published work instead of just making things up.

For your own setup: feed AI your best existing content. Your top blog posts. Past emails. Anything that represents how you actually think and write. The more real history it has, the less it has to invent. You can literally export things out of WordPress and feed it right into AI and it can chomp on all of that to formulate your history.

The brain-dump first

Before I ask for any draft, I give Claude my raw thinking first. Usually I just dictate — talk through my actual take on the topic, unfiltered. The Claude app has a nice voice interface and I just talk to it like I would talk to a regular person. Claude works from that as the foundation rather than starting cold from a topic alone.

This one change makes a bigger difference than anything else on this list. When AI is organizing your thinking, the output sounds like you. When it’s supplying its own thinking, it sounds like… an algorithm.

The edit pass still matters (a lot)

None of this replaces a real editing pass. Wherever you pause reading, wherever something sounds slightly off, wherever you’d never actually say it that way — that’s where your voice goes back in.

Don’t skim it. That edit pass is where you go from director to author. And that distinction is still worth making. AI will get you a long ways to the finish line, but you need to take that baton, add your magic touch, and ensure that article is exactly what you would say. Edit manually until it does.

You can even add to your AI’s memory over time — flag phrases it uses that you’d never say, correct patterns that feel off, reinforce the ones that land. It gets better the more you work it.

The bottom line

The investment is front-loaded — writing good instructions, loading your history, establishing the workflow. But it pays off every single time you sit down to create something.

AI isn’t going to sound exactly like you without training. But with the right setup, it gets close enough that you can close the gap in the edit.

That’s the goal. Not AI instead of you. AI that makes you faster.