Well, you don’t have to be fooled, we all went to IA or almost. When I say all, I’m talking about people like me who do things on the internet & mldr; I even see who shit all day long on the generative AI and who use it. It feels in their articles. Or is it because they pump content made with AI? I don’t know.

But you have to recognize it, it’s the brothel!

For my part, I do a lot of tests here of editorial staff assisted by IA. I say assisted because it is out of the question for me to go into mode “Site generated automatically with AI“Like some media. On the other hand, it is true that I use it, I hope in the most intelligent way possible. Thus, when I have to analyze a subject for an article, I first ask the AI ​​to make a complete analysis.

SCR 20250410 JBDM

I then have a working document which allows me to fully understand the concepts linked to the subject, and above all to have angles of analysis of which I would not have necessarily thought. I also have some intro examples which I can be inspired by if I am blocked by the blank page and an example of a plan that I am free or not to follow (in general, I am not really because lazy & mldr;).

From that, I can start writing old -fashioned as usual. And when I have finished, I submit the article to my favorite LLM (namely Claude) which then criticizes my article and gives me advice to make it better, even if I ask, slightly modify my article to integrate the info that I could have forgotten and its criticisms.

SCR 20250410 JKMA 1

I am therefore a blogger assisted by AI and I assume because I use it as a tool to accompany me, in the same way that I use Wikipedia, a spelling corrector, or sometimes a SEO tool. And obviously, I reread and valid everything. I am the only master on board and I am 100% responsible for what is published under my name. Besides, a small parenthesis, I saw that in 2026, the sites that use AI will have to mention it clearly. I dug the subject a little and obviously, the way I use it is not subject to this warning Because precisely, there is a human look and a human validation on what is published.

Now for the images, this is another subject. In the past, I used free right images (and my own images) but that earned me concerns because there are unscrupulous companies that put free right -of -law images on the net for a few years, the time that pigeons like me use them and then withdraw their status CC0 (or other) to then send lawyers that claim the money for theft of photos. So today, I generate them with AI and it suits me very well.

Now as a reader, I am used to it and I notice immediately when there has been an LLM in the generation of an article (hence my intro) and at a time when Gemini, Chatgpt and others flood the web from automatically generated texts, not easy to separate the good grain from the tares. Fortunately, Australian researchers have discovered the invisible signature who betrays artificial writing.

With an accuracy of 98.3%, their algorithm can distinguish a text written by a professional journalist from a text laid by an AI, even when your eyes see no difference. And it’s not just a subjective impression like mine. These researchers from Charles Darwin University conducted a rigorous scientific study where they compared 150 articles written by award -winning journalists with 150 articles generated by Gemini (the new Google toy which replaced Bard) on the same subjects.

And it turns out that humans and AI have different “stylistic imprints” that can be measured and quantify. A bit like recognizing a robot in his approach too regular & mldr; And yes, the AI ​​writes with a metronomic cadence that betrays it in the eyes of the algorithms.

The 5 most revealing characteristics are:

  • There Phenious length range. We humans sometimes write very short sentences and sometimes with endless sentences that go in all directions before returning to the initial subject after a detour by our disorderly thoughts. On the contrary, AI is more constant.
  • THE paragraph variation coefficient. Our paragraphs vary enormously in length, while those of AI are more regular.
  • THE Verbs ratio. Humans use about 16% verbs against 13% for AI, which is not a huge difference, but that matters.
  • There grammatical complexity. Our phrase structures are less predictable (well, it depends for whom & mldr; loool).
  • There variability of paragraphs. And again, AI loves order and regularity too much, which betrayed it & mldr;

The algorithm that detects all this is called Random (literally “random forest”), which allows you to count each comma and analyze each structure of the sentence to then see what is repeated the most. The most crazy is that this method reaches 98.3% precision!

Another interesting discovery, AI prefers names to verbs. Indeed, it describes the world rather than putting it in motion. It is a bit of a beginner sports commentator syndrome that says “heated Mbappé pass“Instead of saying”Alalalala, Mbappé dribble and pass the ball !!”. It’s less dynamic, much more static.

Now, now if AI produces more standardized, more “boring” but just as readable and informative content, is it really a problem for readers? I don’t believe it.

In any case, me it doesn’t bother me, as long as it’s cool and pleasant to read. Especially since certain types of basic journalistic content without “artistic” value such as financial reports, sports summaries, coverage of predictable events, can, I find, be entirely delegated to AI. This frees time for humans who can then focus on their added value: investigation, deep analysis, emotion and creativity.

And above all, what I explained to you about my “collaborative” use, it also saves time while bringing this touch of creativity that machines cannot (yet) reproduce.

In short, one thing is certain, detection tools such as that developed by Charles Darwin University will be essential to maintain transparency in this evolving ecosystem. But be careful not to fall into a ridiculous witch hunt because everyone remains free to read what they like, no matter who writes it (IA or human), as long as the info is reliable and published in a responsible manner.

Source


Source link

Categorized in:

Technology