These days, AI is everywhere. It helps us write, rewrite, correct grammar, generate ideas, create images, summarize documents and even draft emails. AI has certainly made many things easier. But there is another phrase that has quietly entered the world of writing:
“This looks AI-generated.”
Or simply:
“AI used.” And sometimes, that label seems to be enough. But my question is: Where is the proof?
When writing is judged by a machine
Imagine spending hours writing an article.
You think about the topic, remember something from your own experience, search through old books, speak to people, collect information, rewrite a sentence several times because it doesn't sound right and finally publish it.
And then someone—or some AI detector—says:
“AI detected.”
Just like that. No detailed explanation. No evidence of which sentence was generated. No history of how the article was created. Just a label. That can be quite unsettling for a writer. Because writing isn't only about the final 800 or 1,000 words. There is a whole process behind those words.
So, how to prove_
This is where I think we need to start talking about something more useful than an AI score.
Keep your drafts
Don't throw away your first draft. Keep the messy version. Keep the version with spelling mistakes. Keep the paragraphs you later deleted. Keep the handwritten notes if you made any. A final polished article may look very different from the first draft, but the drafts show the journey.
Keep your research
Save the books, newspaper cuttings, websites, photographs, notes and references you used. For a food or cultural article, for example, perhaps you spoke to your mother, grandmother, neighbour or a relative. Perhaps you referred to an old cookbook. Those sources are part of your writing process.
Keep your notes
Sometimes the most convincing evidence is something as simple as a notebook. A page filled with: ideas, arrows, crossed-out sentences, half-finished thoughts, possible titles, references, questions to investigate. An AI detector cannot show that process. You can.
Use version history
If you write in Google Docs or another platform that maintains revision history, don't underestimate its value. It can show that a document developed over time. A sentence didn't magically appear in its final form. You wrote, changed, deleted and rewrote it. That is a very different story from simply producing a finished document through a prompt.
Keep your correspondence
Emails, interview messages, WhatsApp conversations, photographs from a place you visited, voice notes and other material connected to your research can also establish that the work came from your reporting or personal experience. Of course, not everything needs to be shown publicly. But if your authorship is questioned seriously, having this material preserved can be useful.
AI detectors
This is where I would be cautious. An AI detector saying “90% AI-generated” doesn't automatically establish that AI wrote the piece. A probability or detection score is not the same thing as a recording of what actually happened. Human writing can be polished, predictable, grammatically consistent and formulaic. Those characteristics can sometimes resemble what AI produces. And a human writer can also use AI in completely different ways. There is a huge difference between:
“I asked AI to write this article and published it.”
and
“I wrote this article myself, but used a tool to correct grammar in two sentences.”
There is another category too:
“I used AI to brainstorm ideas, but the research, experiences, opinions and final writing are mine.”
These are not necessarily the same thing. That is why simply stamping something “AI used” can be an inadequate way of judging authorship.
Perhaps writers need a “writing trail”
We already have something similar in photography.
A photographer can show the original image.
An artist can show sketches.
A filmmaker has scripts, storyboards and footage.
Why shouldn't writers have a writing trail?
It could include:
Idea → Notes → Research → First draft → Revisions → Final article
That trail tells us much more about authorship than a detector score.
And perhaps, in the future, writers will routinely preserve this trail.
Not because we should have to prove our innocence every time we write—but because the process of creation matters.
And what if we did use AI?
I don't think the answer is to hide it.
If AI was used substantially, say so.
Transparency is better than pretending.
But the opposite is also important.
Don't accuse a writer merely because a machine says the writing looks like AI.
Ask questions.
What was the writing process? Were there drafts? What were the sources? Were interviews conducted? Who supplied the photographs? Can the writer explain the ideas and arguments?
A genuine writer should be able to talk about their own work. That, to me, is far more meaningful than a percentage displayed by a detector.
The irony of the AI age
Here is the strange part. We have spent years telling writers to improve their grammar, make their writing concise, avoid repetition, organise their thoughts and write clearly. Now, when a writer actually does those things well, someone may say:
“This sounds like AI.”
So should we deliberately write badly to prove that we are human? Should a well-constructed sentence become suspicious? Should originality be measured by how imperfectly we write?
I hope not. Because good writing is not evidence of AI. And neither is bad writing evidence of being human.
Maybe the better question is not “Was AI used?”
Perhaps the more useful question is:
“Who created this work, and what role did AI play in creating it?”
That is a much more nuanced question. Because technology is changing the way we write. There is no going back to a world without digital tools. But writers still have something machines cannot replace:
their lived experiences, memories, observations, conversations, curiosity and individual voice.
A machine can help arrange words. But the reason I want to write about a particular grandmother's recipe, an old temple kitchen, a book that stayed with me, a journey I made, or an everyday moment in my own kitchen—that comes from somewhere else. It comes from me. And perhaps that is what we need to protect in this AI-filled writing world: not just the final words, but the human story behind them.
Over to you
Have you ever had your writing labelled “AI-generated” even though you wrote it yourself?
If someone questioned your authorship today, what would you show as proof—the drafts, notes, research, photographs, conversations, or simply your writing history?
No comments:
Post a Comment