In Edmond Rostand’s play Cyrano de Bergerac, Christian was young, handsome, and in love with Roxane. But he did not have the words to win her heart.

Cyrano did.

So Cyrano began writing the letters that Christian signed. Roxane fell in love with the intelligence and sensitivity behind those words, believing they belonged to the man delivering them.

More than a century later, artificial intelligence is recreating this story.

Cyrano is AI, hidden behind the scenes and finding the perfect words.

Christian is the user, who publishes, signs the work, and receives the applause.

Roxane is the audience, captivated without knowing exactly whose intelligence it is admiring.

This is what I call the Cyrano Effect.

It happens when someone uses artificial intelligence to win over an audience with thoughts, arguments, and words they would not be able to produce or defend on their own.

I see nothing wrong with using AI to write. Editors and proofreaders have always helped people organize their ideas. Technology has simply made this support faster and more accessible.

However, there is a significant difference between improving an idea and manufacturing intelligence.

When the experience, knowledge, and opinion belong to the author, AI acts as an editor. It organizes, questions, and finds a better way to communicate the person’s thinking.

In this case, Cyrano helps Christian express what he truly feels.

But when the tool also has to create the thought, the argument, and the opinion, Christian begins playing a character written by Cyrano.

The writing starts promising someone the conversation cannot deliver.

This debate gained an interesting new chapter when Anthropic announced that Claude would begin embedding an invisible watermark in the text it produces. By creating statistical patterns through word choices, the system can estimate whether Claude was involved in producing a piece of content. In programming code, the watermark is more limited and may appear only in comments and other sections that do not affect how the software works. The initiative is connected to the transparency requirements of the European AI Act. Anthropic

The announcement has angered some users. Some fear being identified even when they only use AI to review or translate their own writing. Others question whether the watermark could affect the quality of the final text.

It is important to clarify that the system cannot definitively prove that content was created by AI. It indicates a probability, works better with longer texts, and does not identify the user.

Even so, there is something almost poetic about it.

Cyrano himself will begin leaving clues in the letters he writes for Christian.

But perhaps some people are worried about the wrong detector.

The best way to find out whether someone truly understands what they published has never been to search for a hidden mark in the text. You only need to ask a question that was not in the script.

An article can be created in seconds. Knowledge, experience, and the ability to defend an argument still take years to build.

Before publishing something created with the help of artificial intelligence, perhaps the question should not simply be, “Is the text good?”

The more important question is:

Could I personally defend everything I am signing?

If the answer is yes, Cyrano merely helped you find better words.

If the answer is no, be careful.

Sooner or later, Roxane will ask a question.