Most AI-generated work has a smell.
Event posters that look okay, but feel emotionally hollow, and suddenly you notice the person has a third hand. Cartoonish visuals that feel oddly out of place, even if you can’t quite put your finger on what’s wrong. Paragraphs that are technically competent, but are actually using many words to say nothing. When this content gets mass-produced at scale, the word “slop” becomes perfect to describe it.
But it doesn’t have to be this way. It is possible to leverage the acceleration that AI brings, without dooming your output quality.
I was on a flight to Seoul recently and watched a 45-minute long AI-generated documentary on Korean cuisine. As an AI advocate, I held hope that it would at least be half-decent, but truth be told I was ready to be disappointed with the constant exposure to mediocre AI-generated content on social media.
I’m happy to report my pleasant surprise at how much it didn’t suck.
Let’s be clear, this film did not have the cinematographic chops of 1917, or the mind-blowing storyline of The Prestige. I probably wouldn’t watch it again if I scrolled across it on a documentaries page. But despite clear signs of AI usage, the overall quality cleanly surpassed the standards of your typical Instagram AI slop, and that’s important because it proves it is possible to create content of a reasonable quality using AI.
The question is, how?
Stop Blaming AI for Bad Direction
It is easy to look at bad AI work and conclude that the issue is AI itself.
That reaction is understandable, but it is too simple. If you handed someone a $60,000 professional cinema-grade RED camera and asked them to shoot a short clip, the output will probably be bad, but you wouldn’t be blaming the camera.
The tool matters, but the tool is not the full explanation.
This is especially important with AI because people often misdiagnose the problem as a model problem.
“The model isn’t good enough,” or “Gemini makes better visuals than ChatGPT.”
That is sometimes true; better models do produce better raw material. But often the failure mode is largely outside of the model. Outputs come out weak because nobody gave the model enough direction, context, taste, or review. The person expected the AI to be the tool, the operator, the editor, and the quality control system all at once.
That’s not a model problem. That’s a process problem.
AI Sounds Generic Because Generic Is the Default
The familiar “AI feel” does not come from one specific phrase or structure. It comes from the absence of direction.
Left alone, AI tends to optimize for the safe middle. It gives you the answer that is most likely to be acceptable, coherent, and broadly correct. That’s great when you need a quick summary or a first draft, but it is also why so much AI output feels similar. In many ways, it is designed to be this way.
In creative fields, this shows up as visual styles that look the same across generations, or music that follows the same overused rhythmic pattern. In professional writing, obvious language patterns start to emerge too.
That is why the output often sounds generic even when it is technically competent. The tone is clean, but flat. The argument is orderly, but weak. The examples are relevant, but predictable. Nothing is badly broken, but nothing really lands either.
And if you’re trying to stand out, this averaged output just doesn’t cut it.
This is also why better models alone do not solve the slop problem. A better model may produce sharper images, smoother video, and more fluent writing. But if nobody is directing the work, the output still drifts toward the safest average version of the thing.
Here, the problem is not that AI was involved, it’s that we tried to get it to “one-shot” the problem. The mistake is treating AI as the author, director, editor, and quality control department all at once.
The First Output Is Just Raw Footage
AI can generate raw material extremely quickly. That is beyond question and should be recognised as a real breakthrough. For anyone starting from a blank page, that is a really useful tool to have.
But raw material is not the same as finished work.
Creation includes deciding what the work is for, crafting your message, cutting unnecessary parts, improving the sequence, sharpening the language, and checking whether the final thing actually does its job.
Anyone can shoot with a camera, but that alone does not make them a filmmaker. Extra thought and effort has to be put into the creative process to come up with an intended effect, composition, storyboarding, and post-production. Only then do we really consider the final product to be complete.
The same is true with AI-assisted content creation. AI may be great at rapidly producing material, but that can’t be the entire creative process. All AI has done is shifted your focus: instead of spending all your time producing the first draft, you spend more time deciding what the draft should become. You become less like a typist and more like a director.
That is where many people mistakenly use AI. They treat the first output as the final work. In reality, the first output is just the raw footage.
AI Makes Output Cheap, But Taste Is Still Expensive
The best AI users do not simply “ask better prompts.” Prompting matters, but the deeper skill is direction.
Instead, the best AI users also bring intent and taste. They know what the work is supposed to achieve, and what good looks like. They can tell when something just looks good, or actually has substance.
This matters more now because AI has made production cheap. Anyone can generate ten different drafts, scripts, or reports at the press of a button, so the scarce skill is no longer in the act of production. The differentiator now lies in knowing what is worth making, and what makes it good.
The average user asks AI to “write something professional” and accepts the first output they receive. The skilled user explains the audience, the context, the desired reaction, the constraints, the examples to learn from, the mistakes to avoid, and the standard the output needs to meet. The output is then treated as a draft and iteratively improved upon.
AI doesn’t remove the need for taste. It punishes the lack of it.
AI Slop Is Already In Your Business
The slop problem is not limited to creative work. If your team has been using AI, you’ve probably come across some version of it: an overly verbose report saying a lot but not communicating much, or a slide deck that pitches the company in very generic ways.
Businesses make the same mistake artists do when they treat AI as a vending machine for outputs. They ask for AI-generated slide decks, AI-written emails, AI-produced summaries, AI-built dashboards, or AI-handled customer replies. At no point does the user slow down to provide appropriate context, goals, or direction for the AI.
The end result is the same AI slop we see in creative fields, just dressed in a nicely formatted PDF.
For example, an AI-generated procurement summary sounds useful, but only if it knows the vendor history, past pricing, contract terms, budget owner, urgency, and approval policy. Until then, all it can do is give generic smart-sounding business advice, and the user is back in the same spot as before the report.
The naive goal here is to use AI to produce more output and reduce operating costs.
But the better goal is to design better processes where AI handles the parts it is good at, and humans remain responsible for judgment, exceptions, accountability, and taste.
The Winners Will Not Be the Ones Who Generate the Most
As AI gets better, producing more will become less impressive. The advantage will belong to the people and companies that have figured out how to turn cheap material into important and meaningful work.
For creative work, that means taste, direction, editing, and a clear point of view. For business work, it means context, process, review, and accountability. The surface looks different, but the underlying problem is the same.
AI only gives you the raw material. Someone still has to own the outcome.
Slop happens when AI is treated as the whole process.