You only notice bad AI
I see half a dozen posts on LinkedIn every day from people complaining about content that was “obviously” generated by AI. They’ve cracked the code. They can spot it every time. The telltale signs, the giveaway phrases, that distinctive AI “voice.” They’re frustrated that people are flooding their feeds with this garbage. If only everyone would just stop using AI and go back to creating authentic, human content.
Here’s the thing, though. I don’t think AI-generated content looks bad. I think bad AI-generated content looks bad.
This is exactly the same logical fallacy people fall into with CGI in movies. They see bad CGI and conclude that CGI always looks bad and practical effects are always better. But they’re not counting all the times they watched CGI and never noticed it. This video breaks down why you only notice bad CGI, and it’s the perfect parallel to what’s happening with AI-generated content right now.
Amazing, well-executed AI is everywhere. You just don’t notice it. When done well and paired with actual human judgment, you’ll rarely realize that what you’re reading or seeing was churned out by an AI model. Great AI-assisted work serves the goal and the audience. And in doing so, it’s by definition invisible.
The way you get great results has nothing to do with avoiding AI altogether, although you certainly can create great work without it. It has to do with understanding the strengths and weaknesses of what AI can do and playing to those strengths while supplementing the weaknesses with human judgment, creativity, and editing.
What AI actually does well
Right now, AI is really good at certain things. It handles structure fantastically. Outlines, frameworks, boilerplate code, standard formats. You use this stuff all the time and don’t even realize it.
Like writing assistance. That email from your coworker that was well-structured with clear sections and good flow? Probably got a pass through Claude or ChatGPT. The product description that perfectly captured the benefits in a scannable format? Same deal. The meeting notes that somehow captured everything important despite the rambling discussion? AI likely helped organize that.
AI code is everywhere now. Any developer who tells you they’re not using AI assistance is either lying or falling behind. They’re using it to generate boilerplate, debug issues, write tests, refactor legacy code, and explain complex functions. Sometimes it’s entire implementations. You use the software every day and never know which parts were AI-assisted. The code works, it’s maintainable, it passed review. So does it matter that AI was involved?
Technical documentation too. A lot of that clear, well-organized documentation you rely on? It started as an AI draft that a human refined. The AI is good at consistent structure and covering all the bases. The human adds the context, the gotchas, the wisdom from experience.
When it comes to images and visual content, obviously the cartoonish AI portraits stand out. The ones with six fingers or melting faces or that weird AI sheen. But product photos with cleaned-up backgrounds? Marketing images with perfect lighting that never existed in the source photos? Textures and patterns in video games? Concept art that gets refined by human artists? That’s AI too, and you’d never guess.
Customer support responses are increasingly AI-assisted. Many of the helpful replies you get started as AI drafts. The really good ones blend the AI’s structure and comprehensiveness with a human’s judgment, empathy, and ability to read between the lines. The human knows when to override the AI, when to add a personal touch, when to escalate. You can’t tell the difference because the end result is genuinely helpful.
Translation and localization work. AI handles the heavy lifting of initial translation, and human translators focus on cultural nuance, idioms, and making sure nothing sounds weird. It’s faster and often better because the human can spend more time on the hard problems instead of grinding through straightforward translation work.
Research and analysis. People use AI to process large amounts of information, identify patterns, summarize findings, and draft initial analyses. Then they apply their expertise to interpret, critique, and draw conclusions. The AI handles the grunt work. The human handles the thinking.
The invisible line
Here’s what’s interesting. When you read a great article, do you wonder if the writer used AI to help with research, outlining, or drafting? Probably not. When you see a polished image, do you check if parts were AI-generated or AI-enhanced? No, because it looks good and serves its purpose.
When someone sends you a thoughtful, well-organized email, you don’t think “I wonder if they used AI.” You just think “this is helpful.” When documentation is clear and comprehensive, you don’t investigate its provenance. You use it and move on.
We only notice AI when it fails. When the hands are wrong. When the writing sounds stilted and generic. When the logic doesn’t track. When there’s obvious repetition or that telltale “as a large language model” kind of phrasing. When there’s that uncanny feeling that something’s off.
But we don’t count all the times AI was involved and we never knew. That’s selection bias. We’re sampling from “AI content I noticed” and concluding “all AI content looks like this.” We’re ignoring the base rate. It’s like looking at a handful of plane crashes and concluding that flying is dangerous while ignoring the millions of safe flights.
The same thing happened with CGI in movies. People pointed at bad CGI and said “practical effects are always better.” But they weren’t noticing all the invisible CGI. The backgrounds that were completely digital. The cars in chase scenes. The crowds. The environments. Even actors’ costumes sometimes. When it’s done well, integrated properly, and serves the story, you never see it.
What separates good from bad
You can usually boil down bad AI content to a few specific problems.
First, there’s the prompt-and-paste approach. Someone types a quick prompt, copies the output, and ships it with zero editing or judgment. No refinement. No checking if it actually makes sense. No adding the context that makes it valuable. This is like a filmmaker pointing a camera at a green screen and calling it done without any consideration of lighting, composition, or post-production. Of course it looks bad.
Second, there’s using AI for the wrong task. AI is great at structure and synthesis, but it can’t create genuinely novel insights or deeply personal perspective. It can’t tell your specific story. It doesn’t have your expertise or experience. When people try to use AI as a complete replacement for human creativity and judgment rather than as an assistant, the results feel hollow.
Third, there’s lack of oversight. The AI makes mistakes. It hallucinates. It misses context. It doesn’t understand your specific audience or situation. Without a human reviewing and correcting, you get plausible-sounding nonsense or tone-deaf content.
Fourth, there’s mismatched expectations. People expect AI to be magic. They think it should read their mind and produce exactly what they want on the first try. When it doesn’t, they either give up or blame the tool instead of recognizing that good results require iteration, refinement, and clear communication.
The AI industry has real challenges. There’s pressure to ship fast and cut costs. Some people want to automate everything without human involvement. There are razor-thin margins and competitive pressure. Companies promise more than they can deliver. These problems are real.
But when someone combines AI’s strengths with human creativity, judgment, and editing, the results can be excellent. They understand what AI does well: generating options, handling repetitive tasks, processing information, maintaining consistency, covering bases. And they understand what humans do well: making judgment calls, adding nuance, ensuring quality, understanding context, bringing genuine expertise.
They spend time refining the output. They don’t accept the first draft. They edit ruthlessly. They add the details and context that matter. They make sure it actually makes sense for the audience and purpose. If they do their job well, you never even notice the AI was involved.
The real pattern
Here’s something worth considering. Think about the best content you’ve consumed recently. Articles that taught you something. Documentation that solved your problem. Images that caught your eye. Code that worked perfectly. Emails that were clear and helpful.
Now ask yourself: do you actually know whether AI was involved in creating any of that? Probably not. You didn’t check. It didn’t matter. The content was good, so you used it.
You ever wonder why you almost never see great content that happens to use AI and think “this is terrible because it used AI”? It’s because when content is great, we don’t care about the tools. The quality speaks for itself. We’re not in the back of our heads looking for an easy scapegoat.
When a movie is great, we forgive imperfect visual effects. When the storytelling is compelling and the characters are engaging, we’re not sitting there critiquing whether that explosion looked realistic enough. We’re invested in the story.
When content is bad, it’s easy to blame the AI. But plenty of bad content existed before AI. Bad writing, bad images, bad ideas, bad execution. Content that was generic, unhelpful, poorly structured, or just plain wrong. The tool didn’t create those problems. Those are human failures.
There’s a reason corporate speak existed long before ChatGPT. There’s a reason stock photos looked soulless before DALL-E. There’s a reason documentation was incomplete or confusing before Claude. AI didn’t invent low-effort content. It just made it faster to produce.
What actually matters
Maybe the reason people think AI is ruining content isn’t a problem with AI. Maybe it’s just a problem with the content itself.
AI, like every innovation in creative work, is simply a tool. Photoshop is a tool. Spell check is a tool. Search engines are tools. They all changed how we work, and they all faced backlash from people who worried they’d ruin everything.
When the end result is bad, maybe it’s not the tool’s fault. Maybe it’s on the creator to use the tool wisely.
Human judgment still matters. Creativity still matters. Quality still matters. Expertise still matters. Taste still matters. The ability to understand your audience, to know what’s valuable, to recognize quality, to make things clear and useful. AI just changes how we get there.
The writers who succeed with AI aren’t the ones who let it do all the work. They’re the ones who use it to handle the grunt work so they can focus on the parts that require genuine human insight. The developers who benefit most aren’t the ones who blindly copy-paste code. They’re the ones who use AI to speed up the boring parts so they can spend more time on architecture and problem-solving.
The question isn’t whether to use AI. For most knowledge work, that ship has sailed. The question is how to use it well. How to maintain quality. How to keep the human elements that make work valuable.
And maybe the first step is recognizing that you’re already consuming AI-assisted content all the time. You just don’t notice it when it’s good. Which kind of proves the point.
-–
If you’d like, I can help you create a LinkedIn post to promote this article, develop some discussion questions for social media engagement, or write a follow-up piece exploring specific industries where AI integration has been particularly successful.