TL;DR
If you’re a product manager crafting content production workflows for yourself, a client, or your team, start by explicitly defining your values and strategy. (After all, values like authenticity can be core to your strategic edge.) What follows are practical ideas and a walkthrough of how this plays out in a real-world example: a niche medical site powered by AI agents for discovery and ideation.
AI tools don’t need to be garbage generators. They can be useful partners when used thoughtfully with humans in the loop. Here’s just a couple of ways AIs can help in market discovery and content creation, backed by real-world examples from my niche medical community site, ACLSupport.com. Key takeaways: Leverage AI to uncover customer insights quickly and streamline production, while maintaining your unique voice. In a market where AI content tools are projected to reach $2.38 billion by 2032, and the Generative AI In Content Creation Market $134.23 billion, businesses that integrate these tools ethically will gain a competitive edge.
The AI Content Debate: Addressing Key Concerns
As AI reshapes content marketing, backlash centers on two main issues: quality and authenticity.
Quality: The “AI Generates Garbage” Debate
Critics argue AI output often feels generic, optimized for clicks rather than value. This may be true for low-effort prompts churning out clickbait, but it’s not inherent to the technology. Content creators, writers, marketers, understandably worry about disintermediation. Yet, as with any tool, results depend on the user. Poor input yields poor output, much like spellcheckers can’t fix flawed ideas.
Authenticity: The Authenticity Dilemma
The deeper unease stems from the “uncanny valley.” If you haven’t heard of this, it’s kind of a creepy feeling when something mimics humanity imperfectly. We dislike being deceived, especially where source transparency matters. Blockchain might have a partial solution by verifying origins as I’ve written about, but for now, we don’t have this industry infrastructure. According to the Content Marketing Institute’s 2025 B2B Content Marketing report, while 81% of B2B marketers use generative AI, 45% lack usage guidelines, underscoring gaps in prioritizing authenticity. As is often the case, our practices lag the cutting edge tech.
So can AI tools help you generate content and still be authentic? My answer is yes. You may disagree. Which is fine. In all things, especially new things, it may take time to find balance points and establish values.
Can We as Content Producers Use AI and Be Authentic?
Let’s consider some of the issues.
We use spellcheckers. Does that make you inauthentic? What about a grammar checker? At what point does tool use make you a cheat? Among my several hobbies is woodworking. I build furniture among other things. There’s are a million ways to build things. One type of drawer building technique uses something called dovetail joints. (I used them on my daughter’s dresser.) An old-school craftsperson would do these by handsaw. But I use a jig with a power router. Does this make my work lesser somehow? Maybe a little. But the result is still high quality. Now if I just had someone else do it and claimed it as my own, that would be different, right?
The point? Define your authenticity threshold. Start with discovery and ideation, then layer in human oversight. This approach mitigates risks like bias or errors while boosting productivity.
Start With Discovery
Writer’s Block. Any of us who’ve stared at a blank page has likely suffered from this. Though maybe we at least have a target market in mind. Maybe we work for a special enthusiast publication serving… who? Airplane pilots, scrapbookers, woodworkers, Minecraft players? We need to get an article to our Editor by deadline! But what should I write about?
Personally, I think using AI tools for discovery and ideation is fair. There’s a website I built not too long ago called ACLSupport.com to support a medical community suffering from a particular type of injury. Those who’ve been through this injury have several online venues, including discussion forums from Meta/Facebook to Twitter/X to Reddit. To help me figure out what content I might create on the site, I built a Forum Discovery Agent using n8n to scan a subreddit daily, summarize posts, extract key themes, and put them in an easy-to-scan Google Sheet. This reveals pain points quickly, informing article ideas.
Here’s the n8n flow. I haven’t published it to the n8n community yet as I need to sanitize a few things first, but if you really need the code sooner, just get in touch. (Open image in a new tab to expand.)
And here’s the output.
Once a topic emerges, I research manually, outline, and draft roughly, skipping full AI generation to preserve authenticity.
Andrej Karpathy, (AI coder and the person who coined the term vibe coding), suggests a lot of our production products will likely have value as being “partially autonomous” and is a proponent of Humans in the Loop. He was talking mostly about coding, but personally, I think he’s right and the idea applies across various generative media. Not just because the pure AI tools aren’t good enough now in a lot of ways and need to be on shorter leashes. Though I think that’s true as well. It’s more that when you’re in the process of creation, being the active guide remains authentic as you’re still the primary agent. The core output will remain in your voice; if you extend the effort to make it so. Now, does this matter? This becomes as much or more a philosophical question than one of pure quality. What I’m offering you here is not an answer I can necessarily fully defend on pure logic or quality grounds. It’s just a point of view.
Now if I wanted, I could automate extracting some of these discovered concepts and start generating content drafts. And some people are doing that for whatever reasons. Personally, I think this is where the Human in the Loop can make a difference in terms of judgement. What I use my discovery for is to figure out where an issue might benefit from supporting content. This is a medical community with various needs. And my personal mission in this particular case is to try to fill in gaps and provide cohesive information within areas of concern.
Personal Editor Agent
Once I have a topic, I’ll write a draft. Whether I start with AI, write an outline, or write the whole thing, I could really use an editor. Most editors have familiarity with a subject area and author. I want mine to as well. But I’m solo on this and who has the time to find one?
So I built one. My editor agent has historical knowledge of my writing in this area, as well as it’s own expertise. The way I trained my agent was to put a pile of my existing content into my agent’s worldview. How? General GPTs get Pre-trained, (that’s what the P is for), in order to create a “foundational model.” That model determines weights of content and thus how prompts are answered. But there’s a way to tweak that by putting my historical content into a vector database that can guide a more general GPT. The end result is the GPT takes my draft and makes a round of editorial changes, in my voice, resulting in a new draft for my review and further human edits.
This scales for teams. Marketers using AI for editing report being 45% more efficient in their workflows. In a survey by the Digital Marketing Institute, 45% use AI to brainstorm ideas, and 50% for content creation. Yet only 19% fully integrate it, per CMI, leaving room for hybrid models that enhance creativity.
Here’s what my n8n-based Editor Agent looks like, with assistance from Supabase for an historical vector database and GPT-3.5 turbo; because I’m too cheap for 4.0 and 3.5 is fine for this task:
(This one is perhaps too customized to my needs for sharing, but again, if someone really wants it, get in touch.)
What it does is first takes in all my historical writing, (once), and then updates its database with anything new. It will use history to compare to new drafts. The lower flow takes my latest draft, and generates edits and suggestions based on both the world and my own past writing. In this case, I’m using ChatGPT for help; though I’m going to test with Grok 4 soon because, well, due to the buzz around Grok 4.
Wrapping Up
This is the balance I’ve found for my little website. I used to call my site a hobbyist effort. But it’s more than that. It’s kind of a labor of love. Or really more of gratitude. After my injury, I found the information and support in the forums amazing and critically useful. A number of ideas quite literally saved me from a great deal of physical pain. So I built the site to be a more static place of constant availability for people seeking such info. The content is mostly long tail, so should SEO well. And if LLMs “steal” it I don’t really care as this isn’t a commercial effort. Maybe I’ll get to AEO, (Answer Engine Optimization), in a future article, though others are already covering that; but often poorly in my opinion.
The point is, in this particular case, as much as I intend to grow and maintain this site, I don’t always have that much time to do so. Using these tools allows me to discover value, and generate timely and somewhat timeless content relatively quickly and authentically. Yes, it’s effort to set up personal or business agent flows and to occasionally maintain them. But once set up? They save a lot of time. And some of them work 24/7 while I’m doing other things. You’ll have to choose your values and balance points for how you build things. My judgement is there’s nothing wrong with using these tools, but I’m agreeing with Karpathy, there’s still value to humans in the loop.