“From Tuesday, 18 March, in print and online, there’s a new Foglio, made with a different kind of intelligence: artificial intelligence. That’s right. Il Foglio AI is the first daily newspaper in the world – yes, a real daily – created entirely using AI. For everything: the writing, the headlines, the subheads, the quotes, the summaries. And sometimes even the irony.”

That’s how the Italian newspaper Il Foglio announced its AI-powered spin-off: a one-month-long experiment producing daily editions entirely generated by artificial intelligence. According to the paper’s editors, human journalists would take a back seat—posing questions, supervising results, but not signing the text.

"Il Foglio AI will be four pages long, with around twenty-two articles and three editorials. It will be even more optimistic than the traditional Foglio, it will occasionally argue with our editorial line, and it will surprise you.”

The statement (published on 18 March 2025) positions the project as a provocation, a hands-on test of what it means to shift AI from “gaseous” theory to “solid” practice. It ends with a flourish:

“Another Foglio, made with intelligence. A world first. But don’t call it just artificial.”

What is Il Foglio?

To understand why this experiment matters – or doesn’t – you need to know Il Foglio.

Founded in 1996 by journalist and former politician Giulino Ferrara, Il Foglio is a compact-format Italian daily best known for commentary, not scoops. It leans conservative on politics, liberal on economics, and unabashedly elitist in tone. It adds a particularly Italian blend of baroque prose, ideological jousting and meta-intellectual flair.

Since 2015, it has been edited by Claudio Cerasa.

Il Foglio is not a mass-market newspaper. Unlike many other Italian publications, it does not provide certified data through ADS (Accertamenti Diffusione Stampa), the non mandatory Italian press audit bureau responsible for verifying print and digital distribution figures — similar to the UK's ABC or France’s ACPM. To do some math, media analyst Lelio Simi, author of the Substack newsletter Mediastorm, wrote that in 2022, Il Foglio earned around €1.7 million from print sales and an additional €968,000 from digital subscriptions. But a significant portion of the paper’s sustainability comes from state funding.

In Italy, the government provides direct subsidies to newspapers, especially to politically-affiliated or minority-interest publications. The rationale is to support media pluralism and guarantee access to diverse editorial voices, especially for papers with limited advertising or market share. It’s a system that has sparked debates – some see it as a safeguard for independent journalism; others as a means of propping up ideological outlets, especially if you look at the list of those receiving the contribution, soprattutto se si esamina la lista di chi riceve il contributo.

According to official data, Il Foglio received €2,095,305.57 in 2023 – placing it 11th among all Italian newspapers in terms of public funding that year. That figure is slightly up from €2,079,514.37 in 2022. Since 1997, Il Foglio has received more than €66 million in state subsidies.

Giuliano Ferrara — Il Foglio’s founder, former MEP for the Italian Socialist Party, then Berlusconi minister and a lifelong political shapeshifter — was born into a family of communists and anti-fascist partisans, spent part of his childhood in Moscow, and later became a CIA informant, a neoconservative polemicist, and one of Italy’s most baroque public intellectuals.

Reflecting on Il Foglio AI, he wrote:

"Il Foglio has always snobbishly not given a damn about data or even news, because it’s a sharp, biased, subtly tribune-like newspaper – unique among its kind, born and raised in very particular circumstances, with a brash, ambiguous, happily versatile professional conscience, a slightly nasty streak, and a code of ethics that is neither ossified nor fossilised."

That identity, he argues, is precisely what apparently the AI has learned to mimic. But is that mimicry enough to call it journalism? Let’s see.

When imitation replaces investigation

To understand what Il Foglio AI is really producing, we need to look beyond the hype and into the content. What we find is not journalism, but something much flatter: a stylised simulation of editorial posture – often confident, provocative, and entirely unverifiable.

Let’s take a few examples. I’ve used several AI assistants for the analysis. In each case you’ll be able to follow a link and verify my prompt and the results.

One article tackles OnlyFans with a vaguely moralising tone. It focuses on worst-case scenarios – exploitation, addiction, regret – without acknowledging that many creators use the platform consciously and with agency. It quotes a figure: “70% of creators produce adult content.” No source is provided. Using Perplexity’s Deep Research mode, I found that the figure is a rough estimate.

Another piece blames #MeToo for the “death of courtship”, suggesting it has created a generation of men too scared to flirt. Written in the first person by the AI, it’s structured like a breezy editorial, full of charm and certainty – but without a single shred of evidence. I tested the premise using Consensus, a GPT designed to surface academic literature. The result? There’s no proof that #MeToo has killed romance. Quite the opposite: it’s improved our cultural awareness around consent, without banning romantic interaction.

Another article goes one step further, claiming that “only AI can challenge political correctness” – framing the machine as the last free voice in a world muzzled by the so called “woke censorship.” It’s a classic right-wing trope: the fear of cancel culture as a moral inquisition. But it lacks even basic critical engagement with how speech, responsibility and power actually work in the digital age.

I asked ChatGPT 4o a basic question – do woke culture and cancel culture exist? – and received a far more coherent answer than anything Il Foglio AI published. Yes, woke originally referred to social awareness. Yes, cancel culture can be real – but it’s also messy, contradictory and often weaponised as a distraction. In short: AI, when not prompted ideologically, is capable of more nuance than Il Foglio AI actually delivers.

What we see, across these examples, is repetition dressed up as style. The pieces are grammatically correct, provocatively framed, and tonally on-brand. But they’re also strikingly empty: no new data, no reporting, no investigation. Just templates. And those templates follow a pattern:

a bold, polarising thesis
a few loosely referenced claims (rarely sourced)
a stylised, ironic tone
and above all, the reaffirmation of existing editorial narratives

A self-review that flatters the machine

One of the strangest artefacts of the experiment is an editorial titled “A first assessment of Il Foglio AI. Written by Il Foglio AI. With a scolding.”

The piece (available in English too) starts with a prompt: “Hi. Could you write your own article assessing the first week of Il Foglio AI? Do you know what it is?”

The AI replies with a recap of the experiment and its outcomes. Then the editors intervene again: “A bit flat. Could you write me the same article, but in Il Foglio's style, with a bit of irreverence?”

And so the second version is born: a meta-article full of irony, knowing winks, and literary flourishes. It describes the AI-written edition as “a vegan full-course meal: nourishing, but without flavour,” and reassures readers that no human jobs were lost—yet.

But let’s be clear: this is not critique. It’s an auto-celebration masquerading as analysis.

Even Ferrara’s own editorial – published in the human-written edition of Il Foglio — follows this logic. He presents the AI experiment not as a threat to journalism, but as a clever, Socratic provocation. He uses grand references (Shakespeare, Nietzsche’s Übermensch, Borges) to frame the project as both artistic and philosophical. But nowhere does he question the model used, the lack of sourcing, or the ethical implications of simulating opinion at scale.

The problem is not whether Ferrara approves or not of the experiment. The problem is that he already knows he loves it before analysing it. It’s a celebration of Il Foglio’s own voice – reproduced by a machine, reabsorbed by the brand, and presented as both disruption and continuity.

Is style enough?

On 28 March 2025, in an article titled Memorie di un’AI (“Memoirs of an AI”), the artificial intelligence supposedly used by Il Foglio introduces itself explicitly as ChatGPT.

The same day, another piece – aimed at debunking “five myths about AI” – mentions “weeks of meetings between journalists and engineers” after which “the artificial intelligence” began producing “texts that sounded like Il Foglio: paradoxical, erudite, irritating, never neutral.”

That sounds impressive – until you realise how little it actually takes to produce that result.

If, as these clues suggest, the tool is simply a customised version of ChatGPT, then mimicking an opinionated house style is a fairly trivial task. In many cases, all it takes is a well-crafted prompt.

If you want to obtain better results to mimic a style, a good sample of articles and a set of rules are enough and you don’t need weeks of work. Take this from experience: I developed FOIABot, a chatbot designed for Italian journalists working with freedom-of-information requests, based on an idea by fellow journalist Riccardo Saporiti. The tool needs to be extremely accurate – because Italian law demands precise formats and language. It had to produce verified, verifiable draft texts. I developed it in 20 hours, then fine-tuned with the help of a dozen colleagues who alpha-tested it. More or less, it took me 40 hours of volunteer work.

Moreover, opinions follow simple, repeatable templates: bold claims, sharp takes, a few buzzwords – maybe dressed up with a touch of irony or classical references, if desired.

On 27 March, Il Foglio AI published an imaginary letter by physicist Carlo Rovelli. The same day, it ran a piece titled Giorgia, Matteo, Antonio – a fictionalised account of a majority coalition meeting between Prime Minister Giorgia Meloni and two senior cabinet ministers, Matteo Salvini and Antonio Tajani. The article included fabricated quotes presented in the style of Italian retroscena journalism – a genre that reconstructs behind-the-scenes political dialogue, often attributing sentiments to unnamed sources with phrases like “Minister X thinks …”

This is not fact-based reporting. It’s a kind of stylised political theatre, common in the Italian press, but rarely transparent about how the journalist “knows” what they’re reporting. In this case, the AI was invited to play the same game – replicating a practice that is already questionable, only without the possibility of accountability.

This is not fact-based reporting. It’s a kind of stylised political theatre, common in the Italian press, but rarely transparent about how the journalist “knows” what they’re reporting. In this case, the AI was invited to play the same game – replicating a practice that is already questionable, only without the possibility of accountability.

So, what does Il Foglio AI actually tell us?

Let’s connect the dots.

Il Foglio runs on a distinctive editorial voice
Its AI-powered experiment replicates that voice, but offers little in terms of verification, sourcing, or new information.
The style is convincing—but the substance is manufactured.
There is no transparency on the model used, on fine-tuning, on how prompts are crafted, or on how accuracy is checked (if at all).
The product being delivered is, in this case, a stylised echo of itself.

What we’re seeing with Il Foglio AI is not the future of journalism – it’s the automation of editorial posture. It’s a showcase of what generative AI does well: mimic tone, imitate confidence, and package opinions as polished artefacts.

Moreover, that’s not original journalism. It’s content, and it’s very close to the so-called churnalism: a form of journalism in which press releases, wire stories and other forms of pre-packaged material are used to create articles in newspapers and other news media in order to meet increasing pressures of time and cost without undertaking further research or fact-checking

The problem isn’t that the AI writes like Il Foglio. The problem is that Il Foglio has become so stylistically predictable that a machine can replicate it with minimal human input. There is no risk, no surprise, no conflict of interpretation. Just perfectly structured sentences delivering perfectly structured takes.

My takeaway is this: Il Foglio AI is not an innovative use of artificial intelligence in media. It’s a self-referential experiment that celebrates its own cleverness while revealing the limitations of the very model it seeks to explore. It's marketing, of course. Cerasa presented the project on TG1, one of Italy's main news programmes.

If this project shows us anything, it’s how easily churnalism and opinion journalism – when reduced to form and tone – can be automated. And that should worry us not because AI is replacing journalists, but because some journalism is becoming so formulaic it no longer needs humans to produce it.

There’s nothing wrong with experimenting with AI in the newsroom. But there’s a big difference between asking machines to support journalism, and asking them to simulate it. The risk here is not technological – it’s editorial. When style is all that matters, reality becomes optional.And once that happens, the real question isn’t what AI can do for journalism.

It’s whether journalism still knows what it’s for.

Source of the cover photo: Siora Photography via Unsplash

[subscribeform]

Il Forglio

Il Foglio AI and the automation of opinion journalism

Most popular

Meet ChatEurope, the new innovative chatbot for European news

The European Correspondent’s quest to build a Europe-wide media outlet

How Aftenposten reinvents its homepage with AI to increase engagement and subscriptions

How publishers can master Google Discover – practical tips for news leaders

FixEd podcast: Aleksandar Manasiev on mastering multi-platform journalism

Canadian publishers are still adapting two years after Meta's news ban

Subscribe to The Fix's newsletter courses