52% of the Internet Is AI Slop Now. Here's How to Not Add to It.

People who study what's actually being published online have started putting a number on something a lot of us already felt in our gut. More than half of new content going up right now didn't have a person driving it. Not fully. A prompt, a generic output, a light pass, published.

I'm not going to pretend I can hand you a perfectly sourced citation for the exact number, because the tracking on this changes by the month and the methodology varies by who's counting. But the direction is not in dispute. A large and growing share of what's online was written by a model, not a person, and that share is going up, not down.

Sit with that for a second. Then let's talk about what it actually means for you.

Scarcity flips the value

For a long time, publishing consistently was the hard part. Show up, write the post, make the video, send the newsletter. Consistency was the moat. If you were the business owner who actually posted three times a week while your competitor posted once a month, you won on volume alone.

That moat is gone. Anyone can generate volume now. Volume is cheap. Volume is, frankly, most of what's flooding the feed you're scrolling through right now.

Which means the thing that used to be table stakes, just showing up with content, isn't the differentiator anymore. What's scarce now is the opposite of volume. It's a specific human being who actually did the thing, saw the thing, made the mistake, learned the lesson, and can describe it in a way no model trained on the general internet could invent on its own.

Your perspective isn't a nice-to-have anymore. It's the actual product.

What generic AI output does to your voice

Here's the trap. Most people don't sit down and decide to sound generic. They open a chat, type a rough prompt, take the first draft, and publish it with maybe one word swapped. The model didn't know their client, their neighborhood, their actual argument with their business partner last Tuesday, so it filled in the gaps with the most statistically likely version of "helpful business advice." Smooth. Confident. Interchangeable with the next ten posts in the feed.

It's not that the tool is bad. It's that it was never given anything to work with beyond the topic.

How to actually not add to the pile

None of this means stop using AI to write. It means stop handing it the wheel with no map.

Use your own real examples, every time. Not "a client I worked with once." The actual client, the actual number, the actual thing they said in the meeting. Specific detail is the fastest way out of generic, and it's also the one thing a model can't manufacture for you honestly.

Don't let the default voice replace yours. If the first draft sounds like it could have been written by anyone in your industry, that's information. It means you handed over a topic and got back a topic-shaped answer. Go back in and tell it what you actually think, not just what you want it to cover.

Save your voice, don't rebuild it every time. This is the part most people skip. If you're working through something like the AI DNA Workflow, this is exactly what the step where you feed Claude your real examples and writing samples is for, and it's exactly what the later step of saving that voice is for. Write down your actual phrases. Note the words you'd never say. Keep a file of posts that sound like you and posts that don't, and why. Then hand that file to the model every time, instead of starting from a blank prompt and hoping today's draft lands closer than yesterday's.

That saved voice file is the difference between fighting the same battle every single time you write, and building something that gets more like you with every use.

The actual advantage

Somewhere out there, a majority of what's published today has no one specific behind it. That's not a reason to panic. It's the clearest advantage available to anyone willing to actually be specific, actually be present, actually let their own weird particular way of seeing things show up on the page instead of getting smoothed out by a default.

Don't learn to write like the algorithm expects. Teach the tool to write like you. That's the whole game now.