AI and Journalists: What's Actually Happening and What to Do

The honest assessment

Journalism was struggling before AI arrived. Falling ad revenues, shrinking newsrooms, the slow collapse of the local press. AI isn't killing journalism. It's arriving at the bedside of a patient that was already seriously ill.

Here's what AI can do in journalism right now. Write earnings reports, sports match recaps, and weather summaries. The Associated Press has been using AI to generate corporate earnings stories since 2014. That's over a decade. It can summarise council meeting transcripts. Generate SEO-optimised rewrites of press releases. Produce aggregation pieces from multiple sources. Draft interview questions. Transcribe audio and video. Translate stories between languages. Create social media versions of published articles. For the commodity end of journalism... the reporting that involves taking publicly available information and reformatting it... AI is already there.

What AI can't do is journalism. Real journalism. The kind where you sit in a pub with a whistleblower for two hours and earn their trust enough that they hand over the documents that bring down a council leader. The investigation that takes six months and involves reading 10,000 pages of leaked files. The war correspondent filing from a basement in a city under siege. The long-form profile that captures a person so precisely that readers feel they've met them. The editorial judgement that decides this story matters and that one doesn't, and the courage to publish when powerful people don't want you to. AI doesn't have sources. It doesn't have courage. It doesn't have the instinct that something smells wrong even though the data looks clean.

But the uncomfortable truth is that a shrinking percentage of journalism jobs involve that kind of work. Most journalism jobs involve producing content to fill space, generate clicks, and keep the ad server happy. That content is what AI produces cheaply and endlessly. The economic foundation of most journalism jobs is the very type of content AI generates best. That's the problem.

Your exposure level: High

High exposure. Journalism has structural vulnerabilities that AI exploits directly. The business model is already broken, and AI both reduces the cost of content production and increases the volume of content competing for attention. That combination is painful.

The news organisations that have already adopted AI are numerous. The AP, Bloomberg, Reuters, and many others use AI for routine reporting. BuzzFeed used AI for quizzes and listicles before shutting its news division entirely. CNET published AI-generated articles (poorly, as it turned out, which is its own lesson). The trend is clear: any reporting that involves reformatting publicly available information is being automated.

For working journalists, the exposure depends on what you do. Beat reporters covering niche topics with deep source networks and original reporting are relatively safe, because what they produce can't be generated from existing data. General assignment reporters, content producers, and aggregation journalists are highly exposed. Copy editors and sub-editors are also at risk, as AI handles spelling, grammar, style, and headline writing well.

The freelance market is particularly brutal. Content rates have fallen as publications can generate basic articles with AI. The gap between "journalism" and "content" has become an economic chasm. Publications will pay for original reporting, analysis, and distinctive writing. They won't pay journalist rates for content that AI can produce for effectively nothing.

The 90-day action plan

1. This week: test AI against your own work. Take your last three published pieces. Give ChatGPT the same briefs and see what it produces. Be ruthlessly honest about where your work is better and where the AI is "good enough." If the AI version is good enough for your editor... that's the information you need, even though it stings.

2. Week two: build an AI research workflow. Use Perplexity for backgrounding stories. It searches current sources with citations, which saves hours on desk research. Use Claude to summarise lengthy documents, reports, and datasets. A FOIA response that would take a day to read can be summarised in minutes, letting you focus on the newsworthy elements.

3. By day 30: develop a transcription workflow. If you're still manually transcribing interviews, stop. Use Otter AI. It's not perfect, but it's fast and the accuracy is good enough for a first pass. The time you save on transcription is time you can spend on the next interview or the actual writing.

4. By day 45: invest in original reporting. The type of journalism AI can't do is the type that requires going somewhere, talking to someone, and finding out something that wasn't publicly known. If you're not already doing original reporting, start. Even small scoops, local investigations, or first-person pieces create content that AI literally cannot produce because the information doesn't exist in its training data.

5. By day 60: build a direct audience. Start a newsletter, a blog, or a social media presence around your beat. Audiences that come directly to you for your specific perspective and expertise are the foundation of a sustainable journalism career. AI produces content. It doesn't produce trust. Readers who trust you will follow you regardless of which publication you're at.

6. By day 75: learn data journalism. If you can analyse datasets, create visualisations, and write stories from data, you combine a skill AI has (data analysis) with skills it doesn't (editorial judgement, storytelling, source development). Data journalists who can do all three are in demand and relatively rare.

7. By day 90: diversify your income. The single-employer journalism career is increasingly precarious. Consider consulting (media training for organisations, communications strategy), teaching (journalism schools need practitioners who understand AI), content strategy for organisations, or writing books. Having multiple revenue streams isn't abandoning journalism. It's making it sustainable.

AI tools you should be using this week

• Perplexity for Research — This should be your default research tool. It searches current sources, provides citations, and gives you a quick overview of any topic. For backgrounding a story, checking facts, and finding original sources, it's faster and more reliable than a Google search. The citations feature is essential for journalism... you can verify every claim.

• ChatGPT for Work — Use it for interview preparation, headline generation, story angle brainstorming, and draft-to-final editing. Give it your draft and ask "what questions does this story leave unanswered?" It's a decent editor, though it lacks the nose for what makes a story matter.

• Claude for Work — Best for long-document analysis. Government reports, court filings, corporate annual reports, leaked documents. Paste them in and ask specific questions. i use it for finding the newsworthy nugget buried on page 47 of a 200-page document.

• Otter AI for Meetings — Press conferences, interviews, panel discussions. Otter transcribes with reasonable accuracy and generates summaries. Not reliable enough for direct quotes without checking, but excellent for capturing the substance of a long conversation.

What to say in meetings

When editors discuss using AI for content: "AI can handle the commodity stuff. Match recaps, earnings roundups, event listings. Let it. That frees us to do the journalism that actually builds readership and reputation. The stuff people share and remember and subscribe for." Frame it as strategic, not defensive.

If a publication suggests AI-written articles: "CNET tried that and had to retract articles for factual errors. The reputational risk is real. AI-assisted journalism with human oversight works. AI-replaced journalism is a liability."

When discussing the future of the newsroom: "The journalists who survived previous waves of disruption, from print to web, from web to mobile, were the ones who adapted without abandoning the core skill: original reporting. AI is the next wave. Same principle applies."

If the worst happens

If you're made redundant from a journalism role, your skills are more marketable than journalism salaries suggest. Writing clearly, researching thoroughly, meeting deadlines, working under pressure, and telling stories that engage people... these are valuable in communications, PR, content strategy, brand journalism, and corporate affairs. They're also valuable in completely different fields like intelligence analysis, policy research, and UX research.

The most natural adjacent moves: corporate communications, content strategy, PR (it might feel like crossing to the dark side, but they pay better and the hours are shorter), brand journalism, media consulting, or teaching. The growing market for "branded content" in particular needs journalists because it requires the same storytelling skills applied in a commercial context. Some journalists struggle with this transition on principle. That's valid. But principles don't pay rent.

i'll say something that might be unwelcome. Journalism as a salaried career with benefits and job security was always a historical anomaly, and it's ending. What's replacing it is a model where journalists are more like independent professionals: building audiences, diversifying income, and maintaining their editorial standards while adapting their business model. AI accelerates this transition. It's not comfortable. But the journalists who accept it and plan for it are the ones who keep doing journalism. The ones who wait for the old model to come back are the ones who leave the profession entirely. The old model isn't coming back.