Can AI Write Your Opinion Column? One Editor Tested It

The Experiment Nobody Asked For — But Everyone Should Watch

What happens when a journalist feeds years of their own writing into an AI and asks it to produce opinion pieces under their byline? One editor decided to find out, and the results are forcing uncomfortable conversations about authorship, authenticity, and the evolving role of AI in newsrooms.

The experiment started simply enough. A self-described AI news enthusiast built a daily aggregation site — a personal tool to digest the firehose of artificial intelligence news each morning before work. But curiosity, as it often does in the AI era, escalated quickly.

'I wanted to see if AI could generate opinion articles that, while written by AI, capture my personality and perspectives,' the editor explained. The leap from news aggregation to AI-generated opinion columns may sound dramatic, but it reflects a trajectory many content creators are quietly exploring behind closed doors.

From Aggregator to Ghost Author

The methodology was straightforward. The editor provided a large language model with a range of their prior writing — articles, essays, and commentary spanning their career. The AI was then tasked with producing opinion pieces that would plausibly read as if authored by the human whose work it had consumed.

This approach leverages a well-known capability of modern LLMs like OpenAI's GPT-4, Anthropic's Claude, and Meta's Llama models: style mimicry. When given sufficient reference material, these systems can replicate tone, vocabulary preferences, argumentative structures, and even rhetorical quirks with surprising fidelity.

The real question isn't whether the AI can do it. In 2025, it almost certainly can produce passable results. The real question is whether it should — and what it means when it does.

The Authenticity Problem

Opinion journalism occupies a unique space in media. Unlike straight news reporting, which aims for objectivity and factual accuracy, opinion writing is inherently personal. Readers engage with columnists because of their lived experience, their moral reasoning, and the trust built over years of consistent, human perspective.

When an AI replicates that voice, it creates what might be called an 'authenticity gap.' The words sound like the author. The arguments align with the author's known positions. But the cognitive process — the genuine wrestling with ideas, the emotional weight of conviction — is absent.

This distinction matters. A 2024 study from the Reuters Institute found that 52% of news consumers said they would trust AI-generated content less if they knew a machine produced it, even if the quality was indistinguishable from human work. Trust, it turns out, is not just about output quality. It is about the perceived integrity of the process.

Where Journalism Is Already Using AI

This experiment doesn't exist in a vacuum. Major outlets including the Associated Press, Bloomberg, and The Washington Post have used AI-assisted tools for years — primarily for structured, data-driven reporting like earnings summaries and sports recaps.

But the frontier is shifting. Tools like Google's Genesis (now reportedly integrated into broader enterprise offerings) and startups such as Jasper and Writer are pushing AI deeper into editorial workflows. The line between 'AI-assisted' and 'AI-authored' grows thinner by the quarter.

According to a 2025 World Association of News Publishers survey, roughly 68% of newsrooms now use generative AI in some capacity, up from 49% just a year prior. Most applications remain behind the scenes — headline testing, SEO optimization, transcription — but the ambition is clearly expanding.

The Ethical Tightrope

The editor's experiment surfaces a critical ethical dimension. If an AI opinion piece accurately reflects the author's views, is transparent about its origins, and is reviewed before publication, does it cross a line?

Some argue no. The author's ideas are still the foundation. The AI is merely a drafting tool, no different in principle from a ghostwriter or a heavily involved editor. Others counter that opinion writing's value is inseparable from the act of writing itself — that the process of formulating arguments is where genuine insight emerges, not in the final polished output.

There is also the scalability concern. One columnist producing one thoughtful piece per week carries a different weight than the same columnist's AI-generated 'voice' producing five pieces per day. Volume without cognitive effort risks flooding the discourse with hollow proxies of thought.

What This Means for the Industry

The experiment is a microcosm of a much larger reckoning. As AI tools become more capable of mimicking individual voices, media organizations will face mounting pressure to establish clear disclosure standards.

Several industry bodies are already moving. The News/Media Alliance published updated AI guidelines in early 2025, recommending explicit labeling of AI-generated or AI-assisted content. The European Union's AI Act, now in enforcement, mandates transparency for AI-generated media distributed to the public.

For individual journalists, the calculus is deeply personal. AI can handle the mechanical aspects of writing — structure, grammar, even stylistic consistency — freeing human creators to focus on reporting, source development, and the kind of critical thinking machines cannot genuinely perform. Used wisely, it is a force multiplier. Used carelessly, it is a credibility risk.

The Road Ahead

This editor's experiment won't be the last of its kind. As LLMs improve and fine-tuning becomes more accessible, expect more writers, creators, and public figures to explore AI-generated content that mirrors their personal brand.

The journalists and outlets that thrive will be those who draw clear, principled lines — embracing AI as a tool while preserving the human judgment, accountability, and authentic voice that audiences ultimately value. The technology can write like you. It just can't be you.

And in opinion journalism, that difference is everything.