Telegram channel
Every post the moment it goes live.
Free · no email · app or web
Real-world AI projects from newsrooms worldwide, curated news about AI in newsrooms, case studies, reports, researches and guides to help journalists and news organizations understand and implement AI.
The Daily Mail is shifting its focus from page views to engagement metrics such as time spent, repeat visits, and "quality engagement" as it adapts to a zero-click world where AI assistants and search engines answer user queries without sending them to publisher sites. The publisher is overhauling its product, introducing an AI-powered dynamic paywall that weighs user and content propensity to maximize impact, and revamping its ad model to prioritize high-value placements. The Daily Mail aims to increase loyalty, habit, and subscription revenue, with a goal of 1 million subscribers to Mail+ by 2028. The company is also investing in AI-powered tools to free up newsroom resources, building a team of AI product engineers, and focusing on games, vertical hubs, and community engagement to keep users returning. The goal is to create a loyal, direct audience and establish a global news and entertainment business.
Source: Digiday
Co-developing StyleCheck, an AI application built to help journalists verify compliance with editorial guidelines before publication, highlighted how critical the data layer is in high-stakes AI systems. The tool was built using Retrieval-Augmented Generation (RAG) and required a carefully normalized and deduplicated dataset to remove redundancy and create a single source of truth. To reduce the risk of retrieval failures, the system relied on a layered retrieval architecture. While the project was successful, looking at it now, the data layer would be treated much more explicitly as a legal and editorial asset, with stronger logging architecture and a formal review process built in from the start. The main lesson is that grounding is not a one-time feature but an ongoing commitment, and that robust data governance and review processes are essential — especially in the context of frameworks such as the EU AI Act and other responsible AI requirements.
Author: Nathalie Samaha Martin
Source: LinkedIn
AI has already proven its worth for broadcast workflows like automating closed captioning and performing metadata enrichment and search for archived content. As the industry heads to the NAB Show in Las Vegas (April 18-22), vendors and broadcasters say the technology is now ready to help streamline day-to-day news production, particularly through the use of agents that will communicate changes in stories across systems from different vendors and automatically perform tasks like updating graphics or removing a clip from a rundown.
Source: TV News Check
Aron D'Souza (helped Peter Thiel bankrupt news media Gawker), has launched Objection, a startup that uses AI to adjudicate the truth of journalism. For $2,000, anyone can challenge a story, triggering a public investigation into its claims. Objection's AI system evaluates evidence, including primary records and whistleblower claims, and assigns a trust score to reporters. Critics argue that this could chill whistleblowing and make it harder to publish investigative reporting that relies on confidential sources. Media lawyers and experts warn that Objection's system could be used to silence whistleblowers and erode public trust in the press. D'Souza claims his goal is to restore trust in journalism, but experts question whether Silicon Valley entrepreneurs are equipped to evaluate what serves the public interest. The proposal raises concerns about bias, hallucinations, and transparency in AI systems, and whether Objection's pay-to-play model will primarily benefit powerful actors.
Source: Techcrunch
The recent Source Code event brought together journalists and techies to discuss the impact of AI on the news industry. Key takeaways include the emergence of a new AI licensing landscape, with structured news licensing deals like Really Simple Licensing (RSL) allowing publishers to set their own machine-readable licensing terms. Publishers need to manage access to their content to prevent unauthorized scraping, and know their value as AI companies seek "grounding data" for fine-tuning models. The news sector must remain united to negotiate with AI companies, with initiatives like SPUR bringing major outlets together to work out shared standards. Additionally, experts discussed the "zero-click" future, where Google AI overviews may herd users away from publishers, but noted that the impact varies by news category, with some types of content more likely to appear in AI overviews than others.
Source: Journalism.co.uk
Two years on, Le Monde’s partnership with OpenAI looks less like a pure “news distribution” deal and more like a strategic data and market-expansion agreement. The likely core value in 2024 was access to Le Monde’s high-quality French-language archive, which helped OpenAI strengthen its non-English capabilities at a time when French and broader European training data were especially valuable. For Le Monde, this probably meant a meaningful short-term revenue stream — potentially several million dollars a year — plus guaranteed attribution, links, and brand visibility inside ChatGPT. But the bigger question is sustainability: once the bulk of training value has already been extracted, renewal terms may shift away from archive licensing toward a usage-based revenue-share model tied to actual traffic and citations. If that happens, future income could be materially lower than the initial deal, making this first phase highly lucrative but potentially temporary.
Author: Thomas Baekdal
Source: Baekdal.com
Stay in the loop
Real-time posts on Telegram, or a curated digest every Friday.
Every post the moment it goes live.
Free · no email · app or web
One curated Friday roundup, no noise.
0 issues sent · 0 subscribers
Submit AI initiatives your newsroom is working on, resources, policies, or media content to help journalists and newsrooms worldwide stay ahead in the AI era.