AI in Journalism

The first news organisations that adopted AI all did so for a business reason and this could be why actual #newsroom adoption has been low or inconsistent. Most editors look at #AI with suspicion... as something imposed on them and not something they can trust or use to their advantage. Of course, AI can create content, but a newsroom is the last place it should do so. News organisations should protect their status as creators of credible primary knowledge and not outsource that job to machines. The smartest newsrooms use AI as a research assistant, data gleaner, and distribution agent. AI tools can also be used to translate with accuracy, summarise with a level of audience-based customisation, generate representational images with full disclosure, create video elements and audio where footage or clips are not available, as well as for marketing mailers, social posts, visualisations… but always with a human in the loop. If used wisely, AI can be a great force multiplier for news organisations, giving them an edge and speed that keeps them ahead in a competitive landscape. But using AI to create #content is like buying your death on a quick commerce site. 

I tested how well some AI tools actually work for journalism. Here's what I found, published today in the Columbia Journalism Review. 🤖🗞️ 👉 https://lnkd.in/eU4DVJsa As a reporter, I’ve often asked myself: Can I actually trust AI tools to support real journalism work? For me, and for folks like Hugging Face’s Florent Daudens, The Washington Post's Jeremy B. Merrill, and Sahan Journal's Cynthia Tu—“vibe checks” aren’t enough. I teamed up with a group of amazing researchers to run structured tests on some of the most popular AI tools. We used real-world editorial tasks, summarizing government meetings and reviewing scientific research, to see how these tools actually perform. The results? Surprising, frustrating, and occasionally impressive. 📝 Summarizing Local Government Meetings This is bread-and-butter work for many local journalists. Here's what we found: For short summaries (~200 words), tools like ChatGPT-4o, Claude Opus 4, and Perplexity Pro did surprisingly well, often capturing more facts (and hallucinating less) than the human-written summary we used for benchmarking. For longer summaries (~500 words), the quality dropped fast. On average, the tools retained only about 50% of the facts, hallucinated more, and missed key details. ChatGPT-4o had the most consistent and accurate output, with the lowest hallucination rate and best user experience. So: AI can help with quick recaps—if humans are verifying the work. But for more in-depth reporting, it still needs a human doing the work. 🔬 AI & Scientific Research: Not There Yet We also tested newer AI tools designed to help journalists and researchers make sense of academic work, especially tools that promise to surface related studies or verify the importance of a finding. Most tools surfaced less than 6% of the citations included in expert human literature reviews. Across the board, results were incomplete, or just plain wrong. 100% do not recommend (yet). Huge thanks to the brilliant team behind this work: Sophia Juco, Sandy Berrocal, Nneka Chile, Julia Kieserman, Jiayue Fan, Emilia Ruzicka, Mona Sloane, and Michael Morisy 🙌 (and anyone I may have missed!). I’m especially grateful for funding and support from the The Patrick J. McGovern Foundation, Vilas Dhar, and Nick Cain, who are deeply committed to journalism’s future. Next steps: If you’re experimenting with AI in your reporting—or you’ve read the piece and have thoughts—I’d love to hear from you. Drop a comment 👇 or shoot me a message. I'm also looking to connect with others interested in developing AI benchmarking standards for journalism, to help folks test tools more easily and responsibly. Burt Herman, Paul Cheung, Aimee, Nikita Roy, Silvia DalBen Furtado, Nicholas Diakopoulos, Jeremy Gilbert, and many others, I see you! #AIinJournalism #MediaTech #Journalism #AI SABEW Investigative Reporters and Editors Global Investigative Journalism Network Online News Association MuckRock Foundation, Tech Policy Press

Spent the day gobbling up the 2026 edition of the always-interesting Nieman Lab Predictions for Journalism. My takeaway? AI isn’t destabilising journalism as much as it’s exposing what was already wrong. And it’s forcing clarity about what actually matters. Everything that was already weak or in trouble in our profession is being laid bare, and that in turn is forcing us to double down on what is truly valuable about what we do. Why? 𝟭. 𝗔𝗜 𝘀𝘁𝗿𝗶𝗽𝘀 𝗮𝘄𝗮𝘆 𝘁𝗵𝗲 𝗶𝗹𝗹𝘂𝘀𝗶𝗼𝗻 𝗼𝗳 𝘃𝗮𝗹𝘂𝗲 AI hoovers up anything routine. Anything that can be automated, will be. What you’re left with is the work that can’t be done by machines, the work that carries real meaning and value, like: -> Investigations with real follow-through -> Relationships, local insight, context and nuance  -> Spotting shifts before they become stories In that sense, AI becomes a pressure test. If a machine can do your job, you weren't doing the job that mattered. 𝟮. 𝗔𝗜 𝗽𝗿𝗼𝘃𝗲𝘀 𝘄𝗲'𝘃𝗲 𝗯𝗲𝗲𝗻 𝘀𝗼𝗹𝘃𝗶𝗻𝗴 𝘁𝗵𝗲 𝘄𝗿𝗼𝗻𝗴 𝗮𝘂𝗱𝗶𝗲𝗻𝗰𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 People avoid news because they don't find it valuable or useful, not because they're misinformed. AI helps us deliver information faster and better, and audiences still aren't satisfied. -> Misinformation travels through emotional friction, not factual error -> Trust lives inside community networks: WhatsApp groups, organisers, neighbours -> People move toward what’s actionable, not overwhelming We keep telling ourselves the audience is the problem. More often, the gap is between 𝘵𝘩𝘦𝘪𝘳 𝘳𝘦𝘢𝘭𝘪𝘵𝘺 and 𝘰𝘶𝘳 𝘢𝘴𝘴𝘶𝘮𝘱𝘵𝘪𝘰𝘯𝘴. Journalism that's human and useful outperforms bland content produced at scale. 𝟯. 𝗧𝗵𝗲 𝗔𝗜 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝗶𝘀 𝗮𝗯𝗼𝘂𝘁 𝘀𝘆𝘀𝘁𝗲𝗺𝘀, 𝗻𝗼𝘁 𝘁𝗼𝗼𝗹𝘀 AI amplifies whatever system you already have in place: clarity or chaos. Most newsrooms are struggling to use AI well because their architecture is broken or inadequate, with chaotic taxonomies, inconsistent tagging, or disconnected workflows. And the most interesting AI use cases aren’t coming from well-resourced newsrooms. They’re coming from small teams who can move quickly because they’re not weighed down by legacy systems. 𝗦𝗼... 𝘄𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝗮𝗹𝗹 𝘁𝗵𝗶𝘀 𝗺𝗲𝗮𝗻 𝗳𝗼𝗿 𝗷𝗼𝘂𝗿𝗻𝗮𝗹𝗶𝘀𝗺'𝘀 𝗽𝗮𝘁𝗵 𝗳𝗼𝗿𝘄𝗮𝗿𝗱? It isn't about more content or more technology. We know that what people really want is journalism that’s: -> Embedded and relational, not transactional -> Intentional about purpose, not filling an arbitrary pipeline And AI could help us get there, but only if we stop trying to solve a tool problem and start trying to solve a systems problem. That means rebuilding workflows around what matters.  Getting honest about what fills space and what creates value. Accepting that the path forward requires shedding weight, not adding features. And the newsrooms that figure this out are going to be the ones willing to ask the hard questions they’ve been avoiding.

In a blind test, readers preferred the AI-written data story over the human journalist's 3 to 1. Before you read this as "journalism is over", that number points somewhere more interesting. I've been testing the 7-agent "virtual newsroom" from Oxford and Stanford. I pasted a WHO dataset. Minutes later: a finished, interactive, fully-sourced article. You don't even need to write a prompt. (see for yourself) In the study, 39 of 53 reviewers preferred the AI version when it was compared with published pieces from The Economist and The Pudding. In others words, la crème de la crème of journalism. Why? Mainly because of a design choice: auditability. 93% of the agent’s claims had a machine-checkable trail back to the data, code or cited source that produced them. For the human articles, that number was 25%. Not because the journalists were wrong. Because they almost never expose the wiring. In AI articles, wvery figure was clickable. Not “trust me.” Show me. But another number matters more: the agent recovered only about 50% of the human journalist’s editorial angle. Given repair café data, it can rank what breaks most often. It can’t tell you that manufacturers designed those products to be difficult to repair. That angle lives outside the dataset, with the person who went and asked. I’m seeing this in our World Cup experiment at Mizal. Our agent team has access to a ton of match data. With that alone, it can produce something decent. But when the agents can also choose what to read, listen to and watch, they can find the angles that make a story land. The artifact is becoming cheap. The editorial judgment that frames it is not. Agents are strong at what’s in the data. They’re most useful when they surface and connect the human reporting that lives outside it. The more I work with these systems, the clearer it becomes how much expertise matters. And, perhaps more fundamentally, how much we need better collaboration mechanisms and interfaces between journalists and agents.

Agentic AI journalism has arrived. According to the UK Press Gazette this morning, Mediahuis, one of Europe's largest news publishers with 25 titles across five countries, just revealed it's experimenting with a chain of AI journalism agents to produce routine "first-line" news. Not just one AI tool. A full agentic AI pipeline: commissioning, writing, legal checks, fact-checking, multimedia sourcing, and discourse monitoring - all handled by specialised AI agents before a human journalist reviews and publishes. The goal? Free their (currently) 2,000 human journalists to focus on "signature journalism" — investigations, interviews, community-connected depth reporting. What does this mean for PR and communications professionals? How long before: 1. Your press release may be triaged by AI first. Mediahuis is building curated source databases — wire agencies, parliaments, think tanks, political leaders on social. If your organisation isn't in those source pools in a structured, machine-readable way, you may not even make the first cut. Being findable by validated AI system sources may become as important as knowing the right journalist. 2. The two-tier newsroom needs a two-tier pitch strategy. Routine announcements will increasingly flow through AI-mediated workflows. But "signature journalism" — the pieces that build reputations and break stories — still requires human relationships. Know which tier your story belongs to, and invest your time accordingly. 3. AI monitoring is now part of the editorial cycle. Mediahuis's monitoring agent tracks public discourse around published stories. When polarisation spikes, it flags the topic for deeper editorial investigation. That means how audiences react to initial coverage can now algorithmically trigger follow-up journalism. The crisis response window just got shorter and more complicated (if that's possible). The multi-agent workflow Mediahuis describes - commissioning, producing, checking, monitoring - maps directly to how many PR teams operate. Is there an opportunity to apply similar thinking to comms content production: use AI for the routine, preserve human expertise for the strategic? Though fewer routine journalism roles will mean an even thinner pipeline of experienced reporters long-term. And if multiple publishers adopt similar AI systems drawing from the same source databases, do we risk even more homogenised news coverage? What happens when dealing with agentic AI journalism systems becomes the norm? What changes are you already seeing in how newsrooms handle incoming stories? As ever, welcome your comments below. Read the original Press Gazette article here: https://lnkd.in/eZ5_SgpS

A story I wrote for a client showed up in a Google summary 30 minutes after it hit the wire. First, the article was picked up by a local newsroom in a midsize American city. Then, within half an hour, it was indexed and quoted as an AI answer. AI is reading your local news, and it's easy to understand why. Before GPT or Claude can answer a question, these AI models make a thousand small decisions about who to trust. There are too many sources to cover, so they fall back on the same heuristic newspapers have used for two hundred years: trust the institutions that have to put their name on something, that have editors, that are accountable if they get it wrong. Local newspapers fit that description even when they're operating with a fraction of the staff they had a decade ago. Which means the humblest little daily paper in Wisconsin is a more authoritative source in the eyes of a frontier AI lab than a Fortune 500 brand's polished blog. This is an aspect of syndication that nobody factored in two years ago. It wasn't even a real consideration when I started doing this work. Now it's the thing I lead calls with. Brands whose content is sitting in a hundred local newsrooms today are going to be in a very different position than brands whose content is sitting on their own website, hoping to get crawled. Local news has gotten a value boost as far as real estate on the internet right now.

📌 I have a new report for the Thomson Reuters Foundation out today, on how journalists in the Global South and emerging economies are using AI, and the challenges they face in using these technologies. The research is based on a Q4 2024 survey and responses from more than 200 journalists in over 70 countries. 📊 Some key findings: 1️⃣ More than 80% of our sample uses AI, with many journalists using it for transcription, translation, and content editing. 2️⃣ Yet, only 13% of respondents said their newsroom has an AI policy. 3️⃣ Skill gaps are a challenge – over 50% of journalists using AI are self-taught, emphasizing the need (and opportunity) for better training. 4️⃣ Ethical concerns, western bias in LLMs, and lack of awareness of how to use AI, are all factors inhibiting further take-up and adoption. 5️⃣ AI tools remain expensive – affordability is an additional barrier for many newsrooms in the Global South. 6️⃣ Respondents believe that regulation is needed to address a myriad of factors, from ethical concerns to fears around misinformation, and more. Awareness of existing policies and discussions is low among journalists. 🤔 So, where do we go from here? The report outlines key recommendations for journalists, policymakers, funders, and media development organizations, designed to foster the responsible and ethical development of AI and its integration into journalistic work in emerging economies and the Global South. 🎯 📖 Read the full report here: https://lnkd.in/gYZKRdg3 #AI #Journalism #Digital #DigitalTransformation #Media #MediaDevelopment #Research

💡 What happens when local newsrooms get direct access to AI talent + tools? We have been testing exactly that at The Lenfest Institute for Journalism AI Collaborative & Fellowship Program. In partnership with OpenAI + Microsoft, 10 major news organizations have hired two-year AI fellows to build real newsroom solutions. Early projects already live: 📰 The Philadelphia Inquirer: AI archive research assistant 📊 The Seattle Times: AI ad sales prospecting agent 🎙️ Chicago Public Media: Multilingual transcription + translation 🍽️ The Minnesota Star Tribune: launching an AI-driven restaurant finder and sharing the source code for their underlying "Agate" framework 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: Local journalism is under huge pressure. This program is a model for how AI can strengthen, not replace, the mission of journalism by transparently fueling sustainability, collaboration, and innovation. 👉 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝘆𝗼𝘂:  How should AI be applied in industries where trust and mission matter most? 🔗 𝗙𝘂𝗹𝗹 𝗱𝗲𝘁𝗮𝗶𝗹𝘀 𝗵𝗲𝗿𝗲: You'll find case studies, projects, guides, code, and more. https://lnkd.in/g2vnFXqC The Lenfest AI Collaborative & Fellowship cohort includes: + Baltimore Banner + Boston Globe Media + Chicago Public Media + DallasNews Corporation + Newsday Media Group + NEWSWELL at Arizona State University + ProPublica + The Minnesota Star Tribune + The Philadelphia Inquirer + The Seattle Times Gratitude to these organizations for their openness and collaborative spirit as well as their fellows: Aaron Brezel, Brian Prichard, Dana Chiueh 👩🏻💻, Kevin Hoffman, Mark Chonofsky, Omar Reid, Paulino Daniel Ramírez Torres, Rajesh B., and Shanni You.

We at Epicenter NYC have been part of a consortium sharing best AI practices among small publishers, thanks to a grant from the The Patrick J. McGovern Foundation. I wanted to share a piece we compiled with an assist from D. Mishra's GovWire, which monitors public meetings and creates an article, summary, transcript, podcast, among other content items. And I share some learnings. Last week, there was a community board meeting in Queens that none of us could get to (a frequent challenge for small, strapped news outlets in big newsy cities). So Carolina V. Valencia and I sent some of the basics to GovWire staff just to see what was possible. The result is this piece that is a combo of humans and tech, and I suspect that is going to be more the norm for publishers like us versus either-or. Some takeaways: - Aspects of AI make the fact-checking process much faster and smoother; in this case, we checked quotes and titles against a transcript. - We found the summary option to be most valuable for users. Traditional articles (inverted pyramid style) are not really how we deliver information because we strive for action and context. - You definitely need subject experts reviewing content to check spellings, offer fuller descriptions of personalities, explain where things go next... - I am bullish on the multiple formats that are suddenly possible like short video and audio, even though we did not use these offerings (yet). - Original reporting after a meeting and the AI summaries might help reporters be more relevant and useful to readers. I found myself interviewing someone and really pushing on the WHAT NOW versus WHAT HAPPENED. - It helps to have an AI skeptic or ethics czar or copy editor in the mix, even if it's a pain in the ass in the moment or slows things down. Explaining how we know something to be true or original feels a crucial difference between independent, thoughtful, useful journalism and the alternatives. - I estimate we saved about 6-8 hours thanks to the tool; the meeting was 3+ hours long and my process of stitching from the AI-generated content to my hybrid piece took about 2 hours total. - I've said this often and will say again: You have to know the rules to break the rules. If I hadn't been doing this for a very long time now, I dunno if I'd have the confidence to approach journalism in this masala way. We editors and managers likely need to get in there so we can make our mistakes and share best practices and ethical models from a more privileged perch versus junior reporters and others on the front lines who can't read our minds. I'll keep sharing as I tinker. https://lnkd.in/e64MqiVk

It started as a data project on school closures. It ended with uncovering a political proposal built on fake studies generated by ChatGPT. itromso.no, a 25-person newsroom in Norway, published 90 stories in six weeks, exposed how politicians used AI without oversight, and even gained new subscribers along the way. When I spoke with Lars Adrian Giske this week, he walked me through how his team combined good journalism with AI-native tools (like their RAG system, Djinn, and an AI-checker for “low information density” texts) to uncover the truth. But this story is also the story of a small media climbing the AI competence ladder, a journey Giske describes as: 1. Getting to learn what AI is 2. Integrating AI into daily workflows 3. Expanding into data-journalism, with AI tools extending the work  4. Developing in-house tools through an innovation pipeline 5. Commercialising the tools "Having the AI framework and prompting classes is great, but it won't get people to understand and implement AI, he said. You do that through hands-on projects, and it's been our strategy since 2019." The media now has more subscribers than it did 25 years ago, pre-Internet, pre-social media, pre-AI. To me, this case shows both the risks of untrained AI use in public policy and the opportunity for small newsrooms to lead with smart, practical AI adoption. (Also! Lars Adrian Giske will share more at the upcoming WAN-IFRA, the World Association of News Publishers AI Forum in Paris 🇲🇫 )