How Not to Get Labeled AI Slop

You finally built your content machine. You fed it all the necessary information: product, brand, target audience, competitor info. Its instructions are rooted in a clear positioning and messaging framework (I hope. If not, go back a step). 

Now you’re ready to hit ‘start’. Your very own automated AI content engine is just waiting to rustle up anything from research articles to promotional videos.

I have one question for you: Once it has produced that content, who reviews and approves it? 

To ensure information is accurate and instructions are met, someone needs to do that. But you know that already. If you don’t want your content labeled as AI slop, you need to give it the human touch of a writer or editor.

What is Editing Anyway?

In human writing, editing is the second most important phase in the entire process. The first is research. The actual “writing’ makes up only 20% of the process. 

It’s nothing new. On the contrary, newspapers and magazines have an Editor-in-Chief. This person makes the final decisions about content, style, voice, and message and they also determine whether a piece of content meets its goal. 

Authors write books, but editors give the story its final character. Yes, even the likes of Ernest Hemingway (his editor was Maxwell Perkins) and Leo Tolstoy (his wife Sofia Tolstoya edited his work).

Today, with AI-generated marketing content, it’s the product manager or founder who reviews the writer’s work, not a language professional. Or, you are letting AI write your content and then review it yourself. Why? Because AI knows what it’s doing? You don’t think human authors did? 

AI Is a Digital Brain 

Much is being written about how LLMs put together language, so I won’t go into it. You know it’s got to do with probability. But one thing is for sure: it’s not how the human brain processes language. 

The human brain generates an idea. To express the thought, it looks for the most appropriate vocabulary and semantic meaning. The writer writes it down. The editor’s job is to ensure that others understand the meaning as it was intended by the writer. 

Those are two different things: a) productive language processing: how the writer frames ideas b) receptive language processing: how the reader perceives the idea. 

If you have ever written anything, you know that at the time of writing it’s all crystal clear in your mind. You can’t imagine why it may not be obvious to others. A month down the line, when you haven’t looked at the text, you suddenly can’t grasp what you were trying to say at the time of writing. 

But I’m getting carried away. 

My point is: editing is important and even more so for AI-generated texts.

Edit AI-Generated Texts With AI?

Traditional editing has five phases: content edit, structure edit, style edit, presentation edit, and proofreading. The editor works through a text five times after reading the whole thing once in full.

For AI-generated texts we need to rethink the process. And I’m going to say something that will surprise most of my readers: AI can do most of the work. The only condition is this: a human language professional must make the decisions. 

I wrote about the reasons in my previous article and how the best outcomes are achieved, when each party (human professional and AI) do the thing they do better than the other. 

AI can analyse, verify, and generate ideas faster and at a greater scale. The human professional leads and decides based on expertise. 

The emphasis is on ‘expertise’, because it explains why many non-writers are convinced there’s no difference between AI writing and professional human writing. It’s a bit like believing there’s no difference between a hand-tailored suit and one from Woolworth, just because they look the same at first sight. You are not the expert.

Edit out AI Slop using AI?

Of course not every LinkedIn post has to go through a five-stage editing process. AI gives people the words to finally share ideas in writing they previously couldn’t. Also, some content is short-lived or intended for search visibility, so the priority isn’t deep human understanding or engagement. 

You should still double-check the information and sources. I recommend running it through another tool to analyse for SEO/AEO. And it’s a good idea to read with a critical eye for anything that doesn’t make sense in the real world or simply reads weird. 

But websites, long-form informational pieces, and thought leadership pieces define your brand; sales copy, promotional texts, and marketing content bring you business. These texts serve a critical purpose. And they are intended for humans, not bots. Let a professional plan the strategy and make the decisions.

Edit AI-Generated Content in 3 Stages

Here’s a basic framework on how to get started editing AI generated texts:

Stage 1: Run it through another tool

Have a tool like Perplexity fact-check and verify sources. Ask Claude to flag missing, duplicate, or contradictory information, as well as misalignment with language instructions and style guide. 

Stage 2: The human touch 

Ask yourself: Does it have a strong hook? Are the transitions smooth? Can you find stronger verbs, add specificity, or remove filler words? Is there a conclusive ending? Does it speak to human emotion? Does it keep the reader curious throughout?

Stage 3: Teamwork human & AI

Ask whatever LLM you use to generate alternative headlines. Give it direction, i.e. 5 provocative headlines, 5 informative headlines and 5 audience-focused headlines. 

Have AI proofread. You decide which corrections to accept and which not. 

2 Quick Checks Before Publishing 

1. Before you publish, read your text out loud. You may hear something you didn’t see. 
   

2. Give it to someone who’s not interested in the subject. Ask them the next day what they remember. If you can’t wait, ask them for immediate feedback. 

Get an Editor

I’ve learned the theory and also edited content for years. Editing AI content isn’t exactly the same because - as I said - language is produced in a different way. 

All content automation tools that I’ve tried show the same flaws. You know how people from the same country who learn a second language all sound the same and make the same mistakes? Same with AI trying to learn our language. 

AI-generated language also lacks emotional and real-world associations. But purchasing decisions are emotional. Much of our thinking is associative. 

On the other hand, AI processes a lot more information into content. And the data it analyses and evaluates in no time creates a solid basis for decisions. Decisions that must be made by someone with expertise.

I’ll offer a free editing course soon, so keep following me.