Tech journalists are using AI agents to write and edit their stories

Independent tech journalists are increasingly building AI agents into their workflows, from gathering information to the final edit. Wired examined how this is changing the profession and what is left to the human in an era when a machine can write a draft in a minute.

A few years ago, the conversation about AI in journalism boiled down to fear: "will neural networks take our jobs?" Today it sounds different: not "will they take them," but "how do we best use them." According to Wired, independent writers and freelancers are adopting AI tools especially actively. They have no editorial staff, but they do have tight deadlines and must handle the whole cycle themselves, from idea to publication.

A typical workflow looks like this: transcribing interviews with Otter.ai or Whisper, an initial analysis of the topic with Perplexity or Claude, generating a draft outline with ChatGPT, and a final edit and style polish, again with Claude.

One reporter interviewed by Wired describes a technique he calls the "blind reviewer": he sends a draft to an AI and asks it to find logical holes and unasked questions. Another journalist uses an agent as a "second editor," asking it to pick apart the structure before the piece goes to a real editor at the publication.

This raises an uncomfortable question about authorship. If AI transcribed the interview, suggested the angle, wrote the draft and edited the final version, what exactly did the journalist create? Supporters of the approach answer simply: what the machine cannot do, which is build trust with sources, make editorial choices and bear moral responsibility for the facts that get published. Critics counter that when a human name sits on a piece but machine work is hidden behind it, honesty is in question.

While major newsrooms work out policies on AI tools, independent writers are acting pragmatically. Some disclose their use of AI openly at the end of a piece, much as people once noted "text prepared with the help of a translator." Others consider that unnecessary, by analogy with the fact that nobody says which search engine they used to research a topic. The industry still has no single standard.

Accuracy remains a sharp problem. AI models hallucinate systematically: they insert nonexistent quotes, mix up dates and confidently assert unverified claims. Journalists who have built agents into their workflow have developed personal verification systems: never trust AI-supplied facts without independent checking, and use the agent only for structure and style, not for assertions. This adds an extra step, but the time savings, according to the reporters, are still noticeable.

The main takeaway: AI does not replace the journalist, but it changes what a journalist has to be good at. Routine work, meaning transcription, structuring and initial synthesis, goes to agents. The core remains: sources, judgment, accountability. Those who master this partnership with the machine before others will gain a competitive edge. Those who pretend the tools do not exist risk becoming slower and more expensive, not because they lose to a robot, but because they lose to a colleague who uses the available tools more smartly.