AI Is Not Making Your Content Generic. You Are.

Jana Schilling

Aug 20

9 min read

You can usually tell when a piece of content has been generated by AI: It may be perfectly written. The grammar is correct. The structure makes sense. There is a clear introduction, a reasonable argument, and a tidy conclusion. And still, it feels like nobody actually said anything.

The text has no soul. There is no person behind it. No real opinion, no tension, no unusual choice of words, and nothing that makes you think: This could only have come from them.

There is also a lot of it. A 2026 analysis by Graphite examined 55,400 English-language articles from Common Crawl and estimated that 49.9% of articles published in the first quarter of 2026 were primarily AI-generated. That does not mean half of the entire internet was written by ChatGPT. The study looked specifically at newly published articles and used AI-detection tools, which always come with limitations. But it gives us a pretty good idea of how quickly AI-generated content has become normal.¹

It is easy to blame AI for the resulting sea of generic content. But AI is not necessarily the reason your content sounds generic. You might be.

The problem starts before AI writes a single word

A few days ago, I wanted to develop topics for articles I could publish on the Change Strategies blog. I asked AI to look at our existing articles, consider my role in the company, and suggest topics that would fit my profile.

The suggestions were strategically plausible. They suited the Change Strategies brand. They were relevant to the work we do. There was just one problem: they were not my opinions.

They were topics AI had decided I could reasonably have an opinion about. I could have selected one and asked it to write the article. The result probably would have sounded intelligent. It might even have been useful. But I would have relied on AI not only to structure and write the article, but to decide what I believed in the first place.

That is the wrong starting point. If you ask AI to fill a space where your own thinking should be, it will. It is very good at producing a convincing answer. But a convincing answer is not automatically an authentic one.

AI can write the article. It cannot give you a point of view.

The thinking needs to remain human

My basic position is simple: the thinking should remain human. The execution can be supported by AI.

That does not mean AI cannot participate in the thinking process. It can ask questions, identify gaps, challenge assumptions, suggest connections, and develop a concept further. In fact, those are some of the things I find it most useful for. But there is a difference between allowing AI to challenge your opinion and allowing it to invent one for you.

You still need to bring the idea. You need to know what bothers you, what excites you, what you disagree with, and what you have learned through your work. You need to decide which suggestions are relevant, which arguments you believe, and which conclusions you are prepared to defend.

AI can help you sharpen your thinking. It should not remove the need to think.

AI makes it very easy to get lazy

This is where I think the real risk lies. AI makes creating content so fast that it becomes tempting to skip the difficult part. You can type a broad topic into a chat window and receive a complete article seconds later. You no longer have to sit in front of a blank page, struggle with the structure, or spend hours finding the right words.

That is incredibly useful. It also makes intellectual laziness incredibly convenient.

When you let AI do the heavy lifting before you have formed your own opinion, it has no choice but to generate the most likely version of that opinion. It draws on common arguments, familiar phrases, and patterns that have worked in similar content. The result is often polished, reasonable, and completely interchangeable.

This does not mean we should return to writing every article manually from a blank page. I certainly do not intend to. That would ignore the biggest advantages AI gives us. In a controlled experiment involving 453 professionals completing writing tasks, researchers found that access to ChatGPT reduced the time participants needed by 40% while increasing the independently evaluated quality of their work by 18%.²

AI genuinely can make writing faster and better. But removing the blank page should not mean removing the person.

Start with a voice, not a prompt

My preferred process is to speak before I ask AI to write.

I record a voice note, answer questions, or simply talk through whatever is on my mind. I do not have to worry about finding the perfect introduction, organizing every argument, or phrasing each sentence correctly. I can contradict myself, correct something halfway through, and follow an idea before I know exactly where it leads.

That gives AI something real to work with. It can identify the central argument, create a structure, remove repetition, develop underexplored points, and turn the raw material into a coherent article. The speed and structure come from AI. The opinion and source material come from me.

For this article, I did not begin by asking AI to write about authentic content. I first explained why its original suggestions for me had missed the point. It then asked me questions, and I answered them in a long, unstructured voice note.

This article is the result of that process. I would not have written it this quickly without AI. But it would not exist without my opinion either.

Research from the Content Marketing Institute reflects that tension. In its 2025 survey, 68% of marketers said generative AI saved them time, while only 44% used it to write drafts. At the same time, 69% described the writing produced by AI as mediocre. One respondent called it “soulless.”³

The issue is not that AI cannot write. It is what we give it to write with.

AI needs two kinds of context

Having an opinion is only one part of producing authentic content. AI also needs to understand how you communicate.

If you have a certain way of speaking, that should remain visible in the final piece. The goal should not always be to turn your words into the supposedly best or most professional version of them. Sometimes the sentence you would naturally say is better precisely because it sounds like you.

AI cannot learn that from a tone-of-voice label such as “warm, professional, and confident.” Almost every company uses some version of those words.

It needs real material: things you have written, voice notes and transcripts, previous content you genuinely like, examples of phrases you would never use, and enough feedback to understand what sounds like you and what does not. The more you work with AI, the more clearly you understand what kind of output you want. You learn how much context it needs, what instructions produce better answers, and where it tends to flatten your personality into generic professional language.

For company content, the individual voice is not the only layer AI needs to understand. It also needs a clear strategic foundation. What is the company actually about? What does it believe? What does it offer? Who is it speaking to? What does the brand sound like? What should it never sound like? What role does the person creating the content play within that larger brand?

A personal voice and a company voice are connected, but they are not identical. At Change Strategies, for example, an article written by me should still belong unmistakably to Change Strategies. It should reflect our directness, our strategic principles, and our refusal to hide behind marketing language. But it should also sound different from an article written by Fabian because we have different expertise, experiences, and ways of expressing ourselves.

For me, those are the two foundations of authentic AI-supported content:

A clear company strategy and brand
A real human voice and point of view

Without the first, the content may not belong to the company. Without the second, it may belong to the company but could have been written by anyone.

Authentic input still needs human judgment

At the beginning of 2025, I worked on a brand strategy for a coach. As part of the process, she completed a guided vision exercise, and I used the transcript to help develop the strategy. She later said that the result sounded AI-generated.

What made the situation interesting was that the ideas in the document were based on what she had actually said. AI had not simply invented a random brand. It had worked from her own words. The tools were already useful then, but they have improved significantly since, and I would expect the same process to produce a more nuanced result today.

I still do not have a completely tidy conclusion about that experience. It is possible that knowing AI was involved influenced how she perceived the result. It is also possible that something was lost when her spoken thoughts were turned into a polished strategic document.

Research suggests that the label itself can affect how we judge content. In two experiments published in PNAS Nexus, people were less likely to believe and share headlines labeled as AI-generated, even when those headlines were true or had actually been written by humans. The researchers found that people tended to assume “AI-generated” meant AI had written the entire article without human supervision. When participants were told that a human had selected the topic and AI had only improved the clarity and style, the negative effect disappeared.⁴

But using a person’s words does not automatically mean you have captured what feels most important or distinctive about them. A transcript is source material. It is not a strategy. A collection of personal phrases is not automatically a voice. Someone still has to decide what is meaningful, what is merely incidental, and what accurately represents the person or brand.

That is precisely why human judgment still matters. If something does not sound like you, say so. If an argument goes too far, pull it back. If AI has invented a conclusion you never made, remove it. If it simply agrees with everything you say, instruct it to challenge you.

The back and forth is not evidence that the process has failed. It is part of the process.

Proper prompting is not a magic formula

There is a lot of advice about finding the perfect prompt. I do think learning how to communicate with AI matters. Better instructions, stronger context, and clearer feedback produce better work.

But proper prompting is not about discovering one clever sentence that makes AI generate a perfect article. It is about learning how to direct the process. More importantly, the strongest context does not come from one clever sentence. It comes from the body of real material AI can work with: brand documents, meeting notes, previous writing, project files, and the decisions a company has made over time. No prompt hack can replace that.

You need to know what information to provide, which questions to ask, when to push back, and how to recognize an answer that sounds convincing but does not actually reflect your position. You also need to tell AI what kind of relationship you want it to have with your thinking.

If you only reward agreement, it will become a very efficient machine for telling you that all your ideas are excellent.

That is not particularly useful. Ask it to identify weaknesses. Ask it where your argument contradicts itself. Ask what a skeptical reader might say. Ask whether your example actually supports your conclusion. Then decide which challenges are valid.

AI can make your thinking stronger, but only if you allow it to do more than flatter you.

The Change Strategies position

We use AI across almost every part of our work. It helps us move faster, organize complex information, explore ideas, develop creative concepts, and produce work that would otherwise require significantly more manual effort.

I still have late nights. But I no longer spend them moving boxes around PowerPoint and manually building slide after slide. I spend that time thinking about the strategy, developing the argument, and making the work better. That is where the time AI saves us should go.

But speed is not the same as quality, and output is not the same as thinking. We do not ask AI to decide what a brand believes. We use it to help uncover, sharpen, translate, and execute what is already there.

That requires a clear strategy, real source material, active direction, and a person who remains responsible for the result.

AI is not making your content generic. You are, when you ask it to speak before you have decided what you want to say.