Il Foglio AI: Italy now has an AI-generated newspaper

Il Foglio AI is now also included with the daily newspaper Il Foglio in Italy. It is said to be the first newspaper in the world to be completely AI-generated. AI is responsible for both the texts and the headlines. Even the images are generated with Grok. However, the AI edition only has four pages and around 22 articles each, including three editorials. It comes free with the normal edition. After one month, the editorial team wants to evaluate the AI experiment.

In a letter from the editors-in-chief published on the website, it says: “We journalists will only ask the questions, in the AI paper we will read all the answers.” This will help to understand how AI can be brought from a gaseous state, i.e., theory, into a solid state, i.e., practice. The editorial team wants to report on this.

The articles in Il Foglio AI can also be accessed online. There are articles that focus on the production of non-alcoholic wines and others that focus on AI itself: There is an AI-written book review that is immediately answered by the author himself – or by an AI pretending to be the author. One article describes the problem caused by increased rail traffic. The article contains a series of figures and data that are analyzed.

Limits of AI not entirely clear

It is not clear to what extent the editors intervene in the articles. It is said that everything is AI-generated and only the questions are asked by a human – is probably referring to the prompt. Nevertheless, you would think that every article is at least checked for accuracy and hallucinations and amended if necessary. However, this is not the case. There are also no abbreviations or references to those responsible under the articles.

One lead article is entitled “The speech Draghi wanted to give to the Senate, but didn't have the strength”. The accompanying lead picture shows an AI-generated Mario Draghi in a relatively neutral pose, as there are certainly numerous pictures of the politician and former prime minister. The article is really written like a speech, beginning with a welcome to the senators and deputies. Although there is a reference to the fact that all content is AI-generated, the basis on which this article was created remains completely unclear. Another article calls for fact checks to be used to correct the statements of US President Donald Trump.

So far, experiments by other media companies to use a lot of AI have tended to go wrong. There were fictitious authors and survey participants. Neither went down well with readers. The Press Council issued a public reprimand because a cooking magazine relied entirely on AI-generated recipes. In all cases, there was a lack of identification. However, these are probably not the only AI recipes that have been and continue to be published in magazines. Experiments with AI articles at CNET also backfired: more than half of the articles are said to have been incorrect. However, that was two years ago and since then, there have been massive improvements in AI models.

(emw)