The Guardian's Peter Lewis Certifies His Columns as Human-Written

Guardian columnist Peter Lewis has publicly confirmed that he writes his own pieces, without having a neural network generate paragraphs. The trigger was the growing volume of AI-made text and the launch of Proudly Human, an initiative that invites authors to certify human authorship.

Why this became an issue

Lewis writes that opinion pages have started filling up with what he calls "slop-inion": columns and essays in which the author's voice is replaced by machine smoothness. As symptoms, he cites recent cases in Australian media: Crikey had to remove a series of articles on leadership, and a Capital Brief editor complained that 80–90% of submitted texts look AI-generated.

For Lewis, the problem is no longer simply plagiarism, where someone passes off a ChatGPT answer as their own work. In his view, the real risk zone lies much earlier: at the stage of conception, research and assembling the argument. Hence the central question of his piece: if a model does more than fix spelling, helping to find material, suggesting an angle and pointing out logical holes, where does human authorship end?

Lewis calls this a cultural problem rather than a technical one. Checking the final text for "bot style" is not enough, because the finished column can read as lively even if the system has already suggested the author's important decisions. That is why, despite his long-standing skepticism toward generative AI, he deliberately spent a month with Anthropic's Claude to find the boundary in practice.

How the certification works

In this experiment Lewis relied on the approach of Alan Finkel, Australia's former Chief Scientist, who launched the international Proudly Human initiative. Its aim is not just to say "I wrote this myself" but to set out clear rules under which a text still counts as human-made.

It rests on the de minimis principle: assistance is acceptable if it does not undermine the author's right to call the work their own. In other words, this is not a total ban on AI, but a ban on its substantive co-authorship.

• Spelling and grammar checks are allowed
• Generating ideas at a general level is acceptable
• Automating the basic design of the story and the structure of the piece is discouraged
• Delegating the writing of text passages to a model is not allowed
• Using content that significantly affects the final work is not allowed

For Lewis, the key word here is provenance, meaning verifiable origin of a text. Readers should know they are hearing from a specific author, not a well-tuned assistant hidden behind a personal brand. In the article he compares this to the practice of authenticating Indigenous art, where authorship is directly tied to protection against appropriation and forgery. In his logic, such a label is both a promise from the author and a new trust mechanism for the audience.

"I don't want to make my work easier at all."

What his experience showed

Lewis describes his usual column cycle in detail: two weeks before the deadline he reviews the political agenda, formulates questions to probe public sentiment, looks for cultural associations, and then gradually builds the argument. Next come early poll results, drafts, discussion with close readers and editors, a final proofread and submission to the desk.

In this workflow AI can indeed be useful: it can quickly stress-test a thesis, find connections in a mass of data, highlight logical gaps or suggest a non-obvious, if slightly weak, metaphor. But the outcome of the experiment was, for him, rather negative. Yes, such support speeds up the process and lowers cognitive load, but it also removes what he sees as the essence of writing: doubt, trying out options, the painful assembly of structure, dropping beloved but superfluous phrases, several failed versions before a good one.

Lewis writes that this "intellectual friction" is what makes a text alive. Remove the hard part, and a column may become more efficient to produce but poorer in tone and thought. So human certification is for him not only a public promise to readers but also protection against the temptation to take the too-easy path.

What this means

Lewis's story suggests that the next argument over AI in media will be not only about detectors and formal plagiarism, but about the origin of an idea, an argument and a voice. For newsrooms this may mean new rules for disclosing AI involvement, and for readers, labels that work as a trust signal.

If such practices take hold, the question will no longer be "did AI touch this text at all," but "at what point did the machine become a co-author."