AI Writing vs Human Writing: Are AI Detectors The Answer?

Do you use AI detectors?

We are well aware of the scandal in the scientific community being a target for AI-generated papers with AI detectors becoming more refined, there are still questions pending on their accuracy. With the rise of AI usage in our everyday lives, would the purpose of AI detectors be necessary too?

For example, a friend announced the guy she was talking to in Tinder was using ChatGPT to talk to her. I asked her how she knew and she showed me the text. Well-structured, brief, to the point and generic. Also, he forgot to remove the quotation marks. So ok, that was easy.

But another friend told me she received a text from her husband asking her to wire him some money because there was a problem with his bank account and he needed to deal with his employee’s payroll. She didn’t suspect at first, but the question did cross her mind “Is this AI?”.

Would an AI detector have caught that? Probably not.

AI detectors are built to spot patterns in phrasing, not to judge whether a request feels emotionally off.

Patterns of AI-Generated Text

Yes, sure, AI Slop is all over the place.

We are witnessing in text what happened to images a few years back. The dreamlike sceneries with amorphous trees, and portraits with too many teeth and six fingers are a thing of the past. Now we all squint at our phones analyzing which reels are AI and which aren’t. In writing,

Beyond the em dash, the it’s not X, it’s Y structure and the absolute abuse of qualitative adjectives, AI detectors now promise us they can identify “AI patterns” that we are still not able to recognize at first glance. Some writing tools even promise to get rid of these patterns for your generated writing to sound human.

So if AI is being trained to detect it’s own patterns and it is also being trained to remove them.

What does the future of writing look like?

It is true that writing with the sole purpose of clarity can be strategized. It is true that there are prompts out there that help us tailor LLMs to fit the way we write by making a list of instructions that mimic our stylistic choices and call it our voice DNA.

After all, language is a system with a list of rules we are all expected to follow, true. But just as long as the list of rules, there is the list of exceptions. And just like anything human, language, in this case written language is more of a living organisms that evolves and adapts over time not only at the level of society, but at an individual level.

As mentioned in one of the first studies of Pragmatics in AI:

“Short and fragmented sentences produce tension, and longer, more flowing sentences evoke different feelings. AI can certainly write with accuracy, but it cannot feel its own response. It does not know when to use catachresis or break grammatical rules for stylistic effect.”

Why AI Writing Feels So... Average

We must remember AI models are based on statistics.

From all the data fed to it, it will craft responses based on the average. The goal is to get factual technically accurate responses. LLM models are trained on human feedback and general use, which really just compounds the issues, and continues to push the model into statistical average. So if we request writing we will get average writing. Generic, if you will. Not that generic is all that bad. We find comfort in what we know.

Just think about the latest pop start you just don’t get because it’s so generic.

That’s what most people are into! 

Source: Github

Following this logic, it’s no surprise that results of a human writing study that compared the outcomes of written pieces using AI, found the group that used AI to be rated as “more boring,” it had 50% fewer pronouns, and in turn less anecdotes.

In a few words, AI Slop is boring.

The study found that essays about happiness that were created with heavy use of AI did not communicate as much passion as human-written ones.

About the above point, AI detectors will have a hard time trying to catch a moving, statistically 'average' target. Newer models are trained to write with more variation to slip past the very tools built to catch them.

Another interesting question is whether humans will continue to unconsciously repeat the same patterns we believe are AI-telling as we continue to consume AI-generated text daily, in emails, articles, even casual chats.

Could the "tells" we rely on today become a normal convention in written speech tomorrow?

AI Isn't the Enemy 

The same study found that asking the LLMs to revise the human essays based on human feedback from the original human-written dataset, the AI models made much larger edits than a human editor would have done, turning language into a flat surface.

Natasha Jaques, an author of the study, told NBC News:

“Humans care about clarity, relevance, and impact, while AI cares about scalability and reproducibility.” “It’s changing our conclusions in ways that are already affecting our existing institutions.” These are the kinds of gaps AI detectors still can't measure.

But don’t get me wrong. I am not saying AI is not a good writing tool and we shouldn’t use it. AI is great for brainstorming ideas, checking grammar and it saves us tons of time in research when used properly.

More so, as per this research published in 2025 on the impacts of AI in academic writing found that it is precisely because of scalability and reproducibility that they AI tools are increasingly becoming essential for content creation as they are guaranteed to have a positive business impact.

Still, there is a sensitivity that is missing.

As stated in the study mentioned above, AI tools still do not understand nuances of argumentation, discipline-specific writing conventions, and even tone. AI writing tools may give overly generic, or even misleading. AI detectors, in turn, will fail to grasp the pragmatics of it.

For years, writers have tried to keep us entertained, informed and engaged. Emotion is what drives us to identify with a brand, to buy our products, to decorate our house. Is how we build our identity.  And the good news is, we still get to do that!

It is true that AI will continue to refine its outcomes. Will it be possible to continue using AI detectors?

For now, the band aid will hold, but it will not fix the underlying issue. Of course all of these are valid questions, part of the broader debate. I say write because you can do it, because it’s fun. And because AI can sound like you, but language is still ours, and we can continue to do with it as we please.

So yes, create an agent to draft your emails, use AI to generate content ideas, but for the love of God, you don’t need it to flirt in Tinder. 

Author

Claudia Rangel

Claudia brings together language strategy, communications, and AI operations to help global teams communicate clearly, train effectively, and reach multilingual audiences with confidence. From content frameworks to client relationships, she works at the intersection of language and impact.

Let's connect!