The Marble and the Chisel: Why Bad AI Content Usually Isn't AI's Fault
NielsenIQ (NIQ) studied more than 2,000 people watching AI-generated video ads. Consumers spotted most of them as AI and rated them more annoying, boring, and confusing than traditional ads (NIQ, 2024). That's the number people usually quote, and it's real.
Here's the part almost nobody quotes. In NIQ's study, AI ads as a group still scored worse, but the only one viewers did not immediately flag as AI had been heavily edited by an advertising professional (NIQ, 2024). Same technology. Different hands.
When AI content looks fake, the cause is usually how it was used, not the tool itself.
Why People Say AI Content Looks Fake
If you spend any time on LinkedIn, you've seen it. Don't use AI for your ads, they look fake. Don't use AI for your writing, it's full of em dashes and sounds canned. I keep seeing people who are afraid of AI, and I think much of that fear comes from what they've actually seen.
They're not imagining it. Bad AI content is everywhere, and the NIQ study shows people react to it on instinct. But look at the conclusion they draw. They watched a tool used carelessly and decided the tool was the problem.
Why the Same AI Tool Gives Different Results in Different Hands
In the 1500s, Michelangelo wrote a sonnet about this. In my own words: every figure the best artist could imagine is already hidden in the marble, and only a hand that obeys the mind can reach it.
Put a block of marble, a hammer, and a chisel in front of a beginner and you get a pile of chips. Put the same three things in front of Michelangelo and you get something people still travel across the world to see. The marble didn't change. The chisel didn't change.
AI works the same way. The tool doesn't decide the outcome. The hands do. I'm not telling you the tool is magic. Even NIQ's polished AI ads left weaker memory activation than traditional ads, and its researchers urge brands to be cautious. I'm telling you the gap between careless use and careful use is a large part of the story.
What the Backlash Gets Right, and What It Gets Wrong
Take the em dash. People hunt for it as proof that a machine wrote something. A 2023 study from Stanford and Cornell, published in the Proceedings of the National Academy of Sciences, didn't test em dashes, but it tested something more basic: whether people can tell AI-written short self-descriptions from human-written ones. They identified the AI writing with only 50 to 52% accuracy, about the same as a coin flip, and the cues they relied on turned out to be unreliable (Jakesch, Hancock, and Naaman, 2023). That study used the AI models of its time, so treat it as a baseline, not the final word.
We still keep em dashes out of everything we publish. Many readers have decided they're a tell, and that's reason enough. But punctuation was never the real problem. A post that could sit on any business's page without changing a word is the problem. You can scrub every em dash out of a generic post and it's still generic.
What This Series Will Show You
This is the first piece in a short series called “AI Isn't the Problem. How You Use It Is.” It's about using AI the way a trained hand uses a chisel. Here's where we're going:
The Em Dash Problem: why the most famous AI giveaway is no longer reliable, what really gives lazy AI writing away instead, and what fixes it.
Why so many AI ads look fake, and what actually fixes it.
Garbage in, garbage out: how what you feed AI decides what you get back.
What disciplined AI use actually looks like, day to day.
Two pieces already published cover related ground:
How to use AI to stay consistent on social media without the content sounding canned or fake.
One Thing to Check Right Now
Open the last thing you made with AI help: a post, an email, an ad image. Ask two questions. Could this sit on any other business's page without changing a word? And does it contain one detail only your business would know? If the first answer is yes and the second is no, the tool didn't fail you. It just never got a pass from someone who knows your business.
Frequently Asked Questions
What is AI slop?
“AI slop” is the popular name for low-quality, mass-produced content made with AI tools. The label describes what got published, not which tool was used. Content that a person shaped, checked, and made specific to a real business is a different thing from content that was generated and posted untouched.
Is AI-generated content actually bad?
Not by default, but it isn't automatically good either. In NIQ's study, AI-generated ads as a group were rated more annoying, boring, and confusing than traditional ads, and even the polished ones left weaker memory activation. The one ad viewers did not immediately flag as AI had been heavily edited by a professional, and NIQ's research notes that AI ads need to be high quality to land their message. The tool sets the starting point. The editing sets the result.
Can people really tell when something was written by AI?
Less reliably than they think. In a 2023 study published in the Proceedings of the National Academy of Sciences, people picked out AI-written short self-descriptions with only 50 to 52% accuracy, about a coin flip, and the cues they trusted were unreliable. That study used the AI models of its time, so it's a baseline, not a final answer. What readers do respond to is how content feels, and flat, generic content feels flat no matter who made it. That part is within your control.
Didn't Michelangelo say “I saw the angel in the marble and carved until I set him free”?
That line is everywhere and usually credited to him, but its source is uncertain. Wikiquote lists it as disputed. The earliest versions it traces to are anecdotes from the late 1800s and early 1900s, and one of them credits only “a sculptor.” The marble idea in this piece comes from his own sonnet, which we can trace.
Should I just avoid AI to be safe?
Avoiding AI doesn't make your content better. It just takes a tool off the table. The risk people worry about comes from unreviewed output, and that's fixable with a short, honest review step before anything goes out.
Here's the Plan
You don't have to become an AI expert. You need a person who knows your business to look at what the tool produced before it goes out.
If you'd rather have someone set this up properly and keep it sharp: in NIQ's research on AI ads, the warning was that weak AI creative could drag down how people feel about the brand behind it. Book a free, zero-pressure strategy session with me today. If you want us to build it for you, great. If not, you walk away with a free roadmap.
Sources & Further Reading
Snopestionary: AI Slop, Explained, Snopes, the source for the definition of AI slop
Michelangelo, Wikiquote, the source on the disputed “angel in the marble” line
