In an age where artificial intelligence is transforming the way news is produced and consumed, Il Foglio has launched a groundbreaking experiment: an AI-generated version of its journal. It is important to understand the ideology behind such a project in order to grasp the dynamics of these AI-generated articles and provide a well-structured analysis capable of uncovering the messages and editorial agendas they may carry. To delve deeper into this innovative editorial shift, we conducted an interview with the director of Il Foglio, Mr Claudio Cerasa.
Through this conversation, we explored the motives behind the launch of the AI version, the ideological framework guiding it, and how the journal’s editorial line is maintained—or redefined—within this new technological context. Based on the director’s responses, we are now better equipped to interpret the key messages conveyed in the AI-generated articles. His insights provide essential context that allows us to trace recurring themes, ideological positioning, and discursive strategies employed by the AI. This background is crucial not only to understand what the AI is writing, but also to analyse why and how certain narratives are emphasised over others. Ultimately, this interview serves as a foundation for a critical reading of the content produced by Il Foglio’s AI, helping to reveal the broader agenda embedded in this editorial experiment.
General Questions
Are you satisfied with this experiment?
Yes, absolutely. It has exceeded our expectations both in terms of impact and clarity. The Foglio AI project was born as a provocation, but it quickly turned into a concrete and constructive example of how artificial intelligence can be integrated into the most human of crafts: journalism.
What is the main outcome in your opinion? How will you make use of that?
The main outcome is that AI can become a tool for editorial experimentation—not a replacement for journalists, but a partner that expands the range of formats, voices, and ideas we can offer. The experiment taught us that clear human guidance is what determines the quality of AI output. That insight will shape how we continue to innovate, using AI to support but not dilute the editorial process.
I. Origins and Motives
What inspired the creation of the AI version of Il Foglio? Was there a specific challenge or opportunity that led to this innovation?
The inspiration was both playful and strategic. We wanted to explore whether a traditional newspaper, printed on paper, could work hand in hand with artificial intelligence. The challenge was to test whether it was possible to create not just content, but editorial judgment with the help of AI—and to see if the public debate on AI could be shaped not by fear, but by experimentation.
II. Development and Implementation
What kind of AI model or technology is used to generate the articles?
We used OpenAI’s large language models—mainly GPT-4—through the ChatGPT platform. We did not use a custom-built model, but we treated the prompting process as a form of human-driven editorial training.
How is the AI trained to reflect Il Foglio‘s editorial line and writing style? What kind of data or articles was it trained on?
The AI was not retrained with new data. Instead, it was guided through detailed prompts, designed daily by the editor-in-chief. These prompts included tone, ideological direction, rhetorical approach, key references, and style suggestions. Essentially, we taught the AI how to write like us each day, through prompt engineering.
III. Theme Selection and Article Generation
How are the themes or topics of AI-generated articles selected?
They are selected exactly as we would choose themes for a traditional issue of the paper: editorial judgment, current events, ongoing debates, and cultural or political relevance. The AI does not suggest topics—the editors do.
Who decides what the AI should write about—is it a journalist, an editor, or the AI itself based on prompts or algorithms?
Every article begins with a decision from the editor-in-chief. No autonomy is given to the AI in content selection.
What is the workflow once a theme is selected? Is the AI given a specific angle or just a general topic?
The AI is always given a very precise prompt: topic, tone, ideological framing, writing style, rhetorical goal. We don’t say “Write about taxes,” we say “Write a pro-business critique of Italy’s recent tax reform using irony and economic data.” That level of direction is essential.
Are multiple drafts or versions generated for one article? If yes, who chooses the final version?
Sometimes we ask for rewrites if the first result is unsatisfactory. But usually, with a good prompt, one draft is enough. The editor-in-chief reviews every output and approves or requests modifications.
IV. Review, Editing, and Approval
What is the editorial process for reviewing AI-generated content before it is published? Is there a team dedicated to verifying and editing this content?
There is no separate team. Every AI-generated article is reviewed by the editor-in-chief or a senior editor. The review focuses on factual accuracy, stylistic coherence, and editorial alignment.
What criteria must an AI-generated article meet before it’s considered fit for publication?
It must sound like Il Foglio. That means: sharpness of thought, originality, clarity, cultural depth, and stylistic confidence. It also must be factually correct and aligned with the requested editorial tone.
Do AI-generated pieces go through the same fact-checking process as traditional articles?
Yes, and sometimes even more carefully. We are very aware of the potential for hallucination or fabricated details, so every quote, name, and reference is checked manually before publication.
V. Ideology and Editorial Voice
How do you define the editorial line or ideological orientation of the AI version of Il Foglio?
It reflects the same orientation as the human-written paper: liberal, pro-European, pro-technology, often contrarian, with a taste for cultural irony and intellectual debate.
Can an AI truly reflect the nuances of political and ideological discourse in journalism?
Not by itself. But with good prompts and clear human supervision, it can approximate complexity quite well. What it lacks is intention—but it can simulate nuance, argument, and dialectics with impressive agility.
VI. The Role of AI in Journalism
Do you think AI could eventually replace certain types of journalism? If so, which ones?
Yes, some forms of journalism are already being automated—sports results, weather, basic finance reports. What AI can also do is assist in more structured formats: summaries, explainers, data digests. But investigative work, cultural insight, human curiosity—these remain irreplaceable.
How has the AI version been received by your readers and your own journalists?
With a mix of surprise and curiosity. Readers appreciated the experiment’s boldness, especially its irony and originality. Journalists were skeptical at first, then intrigued, then inspired. No one felt replaced—many felt provoked in the best sense.
What role do you see AI playing in the future of Italian and European journalism?
AI can become a creative assistant, a sparring partner, a way to accelerate some processes and enrich others. It is not a threat unless we treat it passively. If we govern it, it becomes a collaborator.
VII. Ethical and Social Considerations
Are you concerned about the ethical implications of using AI to produce opinionated content?
Yes—and that’s why every AI opinion piece is explicitly supervised and contextualized. Transparency is essential. We never pretend a machine is a journalist.
Do you think AI in journalism could undermine trust in the media, or help restore it through transparency and innovation?
Both outcomes are possible. It depends on how we use it. If we hide AI, it damages trust. If we explain it, use it clearly, and remain accountable for what we publish, it can reinforce credibility.
Do you track how AI-written articles perform compared to human-written ones in terms of readership or engagement?
We have some internal metrics—and AI-generated pieces, especially the satirical or experimental ones, often perform very well in terms of engagement. But our goal is not competition between human and AI writing—it’s cooperation.