Interview with Fulvio Julita on the Ethical Manifesto for Writing with AI

I discovered the Manifesto a week ago, by chance, while scrolling through social media. It struck me as an opportunity to test the method I've built in recent years, which I describe on AITalk's editorial method page. The comparison was twofold. First manually, reading the ten principles alongside my daily practice. Then using a Skill, a small personal tool I built to cross-reference my methodology and some of my work with the Manifesto's text, because I wanted to see if I had been indulgent with myself.

The result: I was substantially aligned without prior knowledge of the project, and in some areas even more rigorous than I communicated. In all honesty, this is an assessment of myself, so it should be taken with the caution reserved for any self-evaluation. Nonetheless, it helped me clarify two things: when AI generates substantial content versus when it provides marginal assistance, and how I declare human verification of sources. Hence the decision to speak directly with one of the promoters, to understand whether a Manifesto can become practice rather than just a declaration.

Ten Principles, Not a Decalogue

The Ethical Manifesto for Writing with Artificial Intelligence is an Italian initiative launched by a promoter committee of eleven writing professionals, including Fulvio Julita. It proposes ten principles for using AI with transparency, care, and responsibility, covering sources, facts, tone, privacy, bias, accessibility, and authenticity, up to human judgment and signature. The opening sentence says almost everything: AI can assist writing, while responsibility, judgment, and signature remain with the writer.

It is therefore not a technical manual, but an attempt to establish a culture of assisted writing, where signing means checking, verifying, and taking accountability. Nor was it born in a vacuum, as the international debate on authorship, disclosure of use, and textual integrity is already intense.

The World Discusses

The Italian Manifesto is not an isolated case. In the United States, the Authors Guild expanded its Human Authored certification in March 2026 to all book authors published in the country. Texts must be written by human beings, with tolerance for spellcheckers and uses such as research, brainstorming, and indexing, and authors sign a license following identity verification. In academic publishing, COPE rules out AI tools being listed as authors because they cannot be accountable for the work, requiring disclosure of which tool was used and how. The ICMJE (International Committee of Medical Journal Editors) adds that outputs must be reviewed because they can be inaccurate, incomplete, or biased, and that omitting disclosure may constitute misconduct in certain cases.

The common denominator is the central role of the human author, backed by transparency and traceability. The tension between declaring usage and documenting the process—tool, function, extent, human oversight—remains open.

Disclosure Is Not Enough

Many guidelines ask writers to disclose AI usage, but few specify the level of detail required. A study in AI & Society, conducted with 602 participants, indicates that when evaluating an author, what matters most is the degree of assistance received, more than who provides it. However, given equal assistance, a human assistant is perceived as more deserving of authorship and accountability, and disclosure is felt to be more necessary when the aid is human. The authors conclude that it is worth documenting what the AI actually did, beyond merely stating that it was present.

For readers of news and information, the question is not "did you use AI?", but "how did you ensure that what I'm reading is verifiable and responsible?". With this question in mind, I started from the beginning: how the project was born.

Julita Speaks

How and from what concrete need was the Ethical Manifesto for Writing with Artificial Intelligence born? Among the issues you felt (information reliability, opacity around AI use, writing homogenization, skill loss, publisher's lack of accountability), which served as the initial spark, and which turned out to be equally urgent during your discussions?

The Manifesto was born from the need to offer a benchmark for those who choose to write with artificial intelligence and want to do so with awareness and responsibility. I was discussing this with Alessandra Perotti, an editor and writing coach: in the public debate we saw much enthusiasm and much rejection, but little room to reflect on how to use these tools while maintaining one's own judgment and voice.\ My experience, including in training, showed me that AI could stimulate thought and help write better. At the same time, I saw the risk of mistaking a well-formulated text for a reliable one, or accepting an answer without asking if it truly expressed what one wanted to say. The starting point was this: how to embrace AI's possibilities without delegating the writer's decisions?\ With a committee of eleven writing professionals, we discussed this need and identified ten core value principles. Information reliability, transparency, data protection, and attention to bias are different facets of a responsibility that remains human: whoever signs must be able to account for what they publish.

Why did you choose the form of a Manifesto—a statement of principles—instead of a practical guide, a code of ethics, or a certification? What does joining the project imply in practice: a cultural commitment, a professional standard, an organizational policy, or something else?

We chose the form of a Manifesto because we wanted to focus on the values that guide the work of those writing with AI. Tools change rapidly; the need to verify information, protect people, and take responsibility for published work remains. A statement of principles can offer a common reference point for people writing in different contexts and using different tools.\ During committee discussions, the need for practical guidance also emerged. It is a need we share: our intention is to accompany the Manifesto with an operational protocol that helps translate these principles into daily work.\ Joining means recognizing oneself in those values and committing to bringing them into one's way of writing. It is a cultural commitment with concrete consequences: it concerns the sources we select, the checks we perform, and the respect we hold for readers and the people we write about. The Manifesto can also serve as a reference for professionals and organizations wishing to define their practices. However, joining does not certify the quality of their texts: that responsibility must be exercised every time one signs and publishes.

One of the core nodes of the Manifesto is human judgment: AI can produce a draft, but it cannot assume responsibility. How do you distinguish, in real work, authentic editing from the mere approval of generated text? What minimum checks should be non-negotiable prior to publication: sources, quotes, numbers, attributions, tone, potential biases?

The difference is visible when you interrogate the text rather than merely proofreading it. For every passage, I ask myself: is this what I intend to assert? What is it based on? Are there statements I couldn't explain or justify? A text can sound very good and still fail these questions. Editing also includes the willingness to discard a well-crafted response when it leads in the wrong direction.\ On checks, I start with sources: I open them, verify whether they truly support the statement they are attached to, and evaluate their authority on the topic. The mere presence of a hyperlink is not a guarantee. Quotes, numbers, and attributions require the same effort: I check words in context, the provenance of data, and the accuracy of citations. A model can invent even a title, an author, and a year that appear perfectly plausible.\ Then I consider how the text will speak to readers: is the tone appropriate? Are there generalizations or stereotypes? Are there personal or confidential details that shouldn't be published? Protecting those data, however, begins earlier: when deciding what information to feed into the tool in the first place.

AI can make writing more accessible, but it can also reinforce reliance on proprietary platforms, linguistic stereotypes, and skill inequalities. How do you address this double-edged effect in the Manifesto? And as a trainer, what is the most common pedagogical mistake you see when AI writing is taught to students, professionals, and companies?

AI can help people express themselves, organize thoughts, and make text clearer. This can broaden the opportunity to participate in writing, but having access to a tool does not automatically mean knowing how to use it well. The Manifesto offers guidance to exercise that opportunity mindfully; access inequalities and platform dependence also require answers that go beyond a statement of principles.\ The error I see most often in training is teaching features and providing copy-paste prompts without explaining the underlying logic of the interaction. An instruction works for specific reasons: it clarifies the goal, the reader, the context, and constraints. If I understand those reasons, I can craft my own requests, adapt them, and recognize when an answer is inadequate. If I only learn a formula, I risk becoming dependent on that formula.\ The same applies to stereotypes: teaching people to produce fluent text is not enough. We must train them to recognize which viewpoints it privileges, which experiences it leaves out, and which modes of expression it treats as standard. For me, good training should increase individual autonomy: making a person more capable of choosing, even when choosing means doing without AI.

What is the most serious objection you've received regarding the Manifesto, for instance from those who view ethics as an individual responsibility or fear overly rigid transparency rules, and how do you respond? Is there an objection that caused you to change something in the Manifesto or in how you present it?

One objection concerns the very decision to use AI in writing, which some feel diminishes the value of the author's work. The Manifesto does not urge everyone to use it: it offers a reference for those who choose to do so and wish to maintain awareness and accountability.\ Another question raised by several people was: did you use AI to write the Manifesto? Some even tried to measure its contribution by analyzing the text. We hadn't anticipated this need for clarification during drafting. It led us to update our website communication with a statement on every page and an article detailing the process.\ AI played a significant role primarily in analyzing committee discussions. Working on meeting transcripts, it helped us recover observations that had emerged but had been lost in subsequent debates. Returning to those passages allowed us to reconsider terms and refine concepts when rewriting the text.\ That is why an evaluation based solely on the final text tells only part of the story: it doesn't reflect the work preceding it nor the decisions made after every proposal. The questions we received made us realize that this journey needed to be made visible. We therefore worked on how we present the Manifesto, explaining where AI intervened and how we discussed and evaluated its contribution.

A manifesto ages. Do you foresee a second version, or a transition to more operational tools like guidelines, checklists, or certification pathways? What would need to change, in the Manifesto or in the context, for it to remain relevant?

A manifesto shouldn't change at the same pace as tools. We chose principles precisely so they could endure, but the Manifesto is not set in stone. If experience brings to light an issue we overlooked or the need to better articulate a principle, we can reconvene the committee to discuss it.\ The step we envision today is the operational protocol mentioned earlier. It could also help us understand where applying the principles encounters difficulties. In the committee's intent, it must stem from dialogues with writers and the real problems they face: that is where we can verify whether the Manifesto continues to provide useful guidance.

Signing Is Choosing

The Manifesto is not a final destination, but an invitation to make explicit the criteria that many professionals already apply informally. My initial self-assessment confirms this: I was aligned, and it prompted me to better communicate two aspects of my method. In the international debate the direction seems clear, as transparency, human accountability, and traceability are now quality benchmarks rather than simple labels.

The question is not whether to use AI or not, but how we sign what we publish: with what checks, what sources, and what accountability. Anyone wishing to read the principles or join the initiative can find everything on manifestoetico.it.