Building an AI Content Startup in the Age of AI Slop. Here's What That Actually Looks Like.

Every founder puts their best face forward. Product launches get a polished tweet. Milestones get a LinkedIn post. Setbacks get quietly absorbed and mentioned later, once you can frame them as a lesson. I've done all of that. But I want to try something different today, because I think the honest version of this story is actually more useful.
Here's the setup: I'm building Bono AI, a voice-first content platform that turns spoken conversations into published work. The core idea is straightforward. Talk to Bono, share your thinking, and Bono writes on your behalf. No blank page. No wrestling with structure. Just talk the way you think, and the output follows.
We started building at a moment when AI could clearly do things that felt almost magical. The UX we built around the interview-then-write loop was genuinely good. Users got it. The concept landed. And then, almost in real time, the backlash arrived.
Fighting Perception Before You Even Fight the Market
The term "AI slop" caught fire, and it caught fire for a reason. A lot of content being pumped out by AI tools was hollow. Generic. Structurally predictable. You could spot it in about four seconds. And once people developed that instinct, it stuck. LinkedIn even added a flag for it. The gut check on any piece of content shifted from "is this interesting?" to "did a human actually write this?" Those two questions don't always have the same answer, and right now, the second one is drowning out the first.
What nobody fully warned me about when building in this space is that you can fight competitors and you can fight technical limitations, but fighting a mindset is a different kind of problem. It is slower, fuzzier, and it doesn't respond to a product update.
The practical version of this plays out constantly. We've put genuine effort into capturing real voice and real thinking. The Bono experience is built around a 10-minute conversation, not a text prompt. That is an intentional design decision. It is slower than dropping something into ChatGPT or Claude, but the tradeoff is real: a person who would have spent an hour on a post, or skipped it entirely, can now talk their way into something that actually sounds like them. We're cutting 10 steps down to two or three, not replacing the thinking.
But when someone reads the output, the first question they ask is still "did AI write this?" Not "does this sound like someone I'd want to follow?" Not "is this useful?" The label precedes the work. And in a category where your entire value proposition is making AI-generated content feel human, that is a genuinely hard place to build.
The Moat Problem Compounds Everything
There's a second layer to this that I find myself thinking about more and more. In the current landscape, unless you go very deep into a specific vertical or you scale horizontally at a speed that frontier labs and big tech can't match immediately, there isn't much of a moat. The models improve fast. The interfaces commoditize quickly. Something that felt differentiated six months ago can look table stakes today.
This is the part of building in AI that the highlight reel definitely skips. You are always, always racing something. You're racing perception, you're racing capability creep from incumbents, and you're racing the narrowing window for specialized positioning. If you pick the wrong lane or move too slowly, the opportunity closes on you.
For us, the answer has consistently been to go deeper rather than wider. The vision for Bono was never to be a general-purpose AI writing tool. It's to be the best possible version of a specialized marketing agent that runs on voice as its primary interface. That means understanding a person's voice well enough that the output reflects their actual perspective, not a cleaned-up average of everyone else's.
Honestly, we're not fully there yet. The model training on individual voice and tone keeps improving, but the gap between "this sounds like them" and "this is them" is still real. I'm not going to pretend otherwise. We've worked through several versions of this problem and each iteration closes the gap a bit more, but it is still a gap. Naming that isn't a confession of failure. It's just a description of where we are.
Where It's Actually Working, and Where It's Not
Right now, B2B is working better than consumer. That's the honest state of play.
On the B2B side, the use case is clearer. A founder, executive, or subject-matter expert has a point of view worth sharing and no time or infrastructure to do it consistently. Bono solves that in a way that is easy to explain and easy to validate. The business case for consistent thought leadership is understood. The friction of maintaining it is obvious. We slot in cleanly.
Consumer is harder. The "is it AI?" question hits harder on the consumer side, partly because the bar for trust is different and partly because the value of consistent content isn't as viscerally felt. The users who get it love it. The ones who don't aren't coming back for a second conversation.
This is making us think more seriously about where the product goes beyond written content. Voice-first doesn't have to mean text out. There are other ways to bring someone's ideas to life that might carry different perception baggage. We're early in figuring that out, but the direction is clear: the experience and the outcome matter more than the specific format, and leaning harder into the AI-native architecture gives us options that a surface-level content tool doesn't have.
The Hidden Work: Trust, Timing, and Doubt
What I really want to say, underneath all of this, is that building in a category with a reputation problem is a different kind of hard. It is not just a product challenge. It is a positioning challenge and a timing challenge and a trust challenge, all running simultaneously.

You pick up a technology because you can see clearly what it can do. You start building. You ship. The backlash arrives, not necessarily at your product specifically, but at the entire category. Now you're not just building a better product, you're building against a prevailing assumption that the whole thing is a shortcut that produces junk.
The honest answer is to keep building toward the thing that actually works, to stay specific and deep rather than broad and fast, and to be transparent about where you are versus where you're headed. The instinct to call out AI output quality problems is something I've had for a while now, and it hasn't changed. What I keep coming back to is that specialization and transparency are the only real answers to the slop problem. Generic output comes from generic input and generic intent. Building something that cares deeply about a specific problem, with real design attention and honest feedback loops, is the only way through.
That is what we're doing. It's messy and it's slower than the highlight reel would suggest, and there are real open questions about where the consumer side goes from here. But the direction feels right, and I'd rather say that plainly than dress it up into something it isn't.