Manually and automatically curated AI news for newsrooms — how artificial intelligence is changing journalism, updated daily.
Showing 50 of 793 in this view · 793 total tracked
We need to produce a short teaser for a public newsfeed card, summarizing substantive content of the intro. The intro basically says: October started with business trips across Europe, more trips upcoming, newsletter intro short, but there is a lot to read this week, see you next Friday. So the
SERGEI YAKUPOV · AI FOR NEWSROOMGoogle DeepMind is expanding its SynthID system to make AI‑generated content easier to spot worldwide. Since its 2023 launch, SynthID has embedded imperceptible watermarks in over 180 billion images and videos and 240,000 years of audio, helping users gauge the authenticity of online media. The new SynthID Detector, now available globally in English, lets anyone verify whether an image, video, or audio file was created by Google or partner models from OpenAI, NVIDIA, Kakao, and soon Apple.
News publishers must shift focus from blocking AI scrapers to engaging with emerging AI agents that can become paying customers. OpenAI’s “Dots” and Meta’s “Muse” are personal assistants that read inboxes, run errands and, with user‑set budgets, can purchase content. While EU markets are still untouched, a coalition of 36 publishers (SPUR) has released a standard for AI systems to disclose how they use articles, and tools now allow hidden markers to detect AI usage.
ULRIKE LANGER ·The author, a non‑programmer, built a modular “AI office” using Claude Code that mimics an actual workplace: a lead orchestrator receives requests, decides which of seven specialist helpers (research, OSINT, work, home, writer, presentation, image/video) should act, and then assembles the results.
USA Today has filed a $250 million copyright lawsuit against OpenAI, joining a wave of media companies accusing the AI firm of unlawful data harvesting. The complaint alleges that OpenAI scraped and copied hundreds of thousands of articles from 19 different publications, including USA Today’s own content, to train its language models and then repackaged that material for ChatGPT users without permission or compensation.
Mumsnet faced backlash after an AI-generated “AIBU Response Schema v.1.0” prompt appeared on a popular “am I being unreasonable” thread, offering detailed instructions for a compassionate, non‑diagnostic reply to a user’s family dilemma. Forum members described the prompt as “creepy” and “disappointing,” sparking accusations that the site was using AI to craft posts.
ROBERT BOOTH ·A recent forum at the New York Times building highlighted how AI is reshaping investigative journalism, with eight Pulitzer‑winning teams this year openly using large language models to sort, summarize, analyze and translate massive document dumps. The Times’ AI Initiatives team showcased tools like the Epstein Files Engine—a LibreChat‑based chatbot that fielded over 5,000 queries and helped produce at least 20 stories from 3.
Futurism’s recent exposé revealed that Brown Brothers Media turned struggling news sites into AI‑driven content farms by creating fictitious journalists with fabricated credentials and AI‑generated headshots, churning out plagiarized or low‑value SEO articles that drew most of their traffic from Google. After the report, Google suppressed the sites and the company halted publishing, then updated its guidelines to ban deceptive authorship, stating that fake bylines and AI‑generated images constitute “deception” and signal low‑quality pages.
JON CHRISTIAN ·Ole Reissmann, Director of AI at SPIEGEL, addressed the Mediatech Hub Conference on the rise of an “agent layer” that mediates between publishers and a growing non‑human audience of bots, crawlers and AI assistants. He traced the shift from using AI merely to speed up content production to recognizing that AI now determines how and where news is consumed, threatening traditional traffic and brand relevance.
·OpenAI is rolling out an invisible “textGrain” watermark in ChatGPT and Codex outputs for eligible users in the European Union, with API customers worldwide able to opt in from October 5. The machine‑readable watermark is invisible to readers and is paired with a detection tool that only approved researchers and certain organizations can access; it reports the presence of a watermark but does not reveal the author, prompts, or editing extent.
THE COPILOT ·OpenAI announced a new visual ad format for ChatGPT that will appear alongside AI‑generated images, offering brands a clearer way to showcase products and gauge impact. The ads, clearly labeled and separate from the chatbot’s output, will be tested in the U.S. later this month with an initial group of advertisers.
The Local Media Association’s Broadcast Transformation AI Lab, backed by the Google News Initiative, moved local broadcasters from “playing with AI” to fully operational use, publishing a playbook that outlines six strategic opportunities. First, AI treats each story as the atomic unit, automating transcription and multi‑platform re‑formatting to slash hours and add virtual reporting capacity.
FRANK MUNGEAM ·A nascent market for AI inference is finally emerging, complete with payment rails, yet publishers remain powerless over pricing. While data‑brokers such as Exa, Parallel and Tavily have built an estimated $1 billion “scraper economy” that resells web content to AI firms and other enterprises, only a fraction of that revenue reaches the newsrooms that created the material.
PETE PACHAL ·SPUR, a coalition of major publishers including the Guardian, FT, BBC and AP, has launched a voluntary AI‑content‑telemetry standard (Version 1.0) that records when AI systems retrieve, ground, cite, present and engage with publisher material, now covering text, images, video and audio. The GitHub‑based framework aims to give owners visibility into how their work is used by generative models and to support licensing negotiations, while integrating with provenance tools like C2PA.
THE COPILOT ·arXiv, the open‑access preprint repository, announced new rate‑limiting rules after a surge of AI‑generated submissions overwhelmed its volunteer moderators. Submissions have doubled in two years, jumping from 9,869 in September 2016 to over 40,000 in September 2026, with computer‑science papers sextupling. To curb “thin,” “salami” and low‑quality AI‑written papers, authors may now submit only two papers per month and maintain a maximum of three active submissions.
MATTHEW GAULT ·Florent Daudens tests how new ChatGPT “Spaces and Pages” tools affect authorial ownership by drafting an essay on AI‑assisted writing through voice dictation, selective highlighting and AI‑suggested restructuring, headlines, research links and A/B headline ideas. He finds the AI useful for turning a wall of text into a readable, organized piece and for quick research, yet notes that key structural decisions and tone still require his judgment, and the interface remains buggy and early‑stage.
FLORENT DAUDENS ·A viral New Yorker‑style cartoon of Dolly Parton and Tim Curry, posted after their deaths, was generated by ChatGPT but bore the signature “BLOPER,” the pen name of cartoonist Brendan Loper, who never created the image. Loper and dozens of other New Yorker artists have discovered similar AI‑produced cartoons signed with their real names, prompting feelings of personal violation and concerns over copyright.
A Silicon Valley nonprofit linked to the AI‑safety movement, the Tarbell Center for AI Journalism, announced a $10 million grant program to be spent by the end of 2027 to boost mainstream and independent AI coverage. The new fund will award grants of $50,000‑$500,000—potentially larger—for investigative teams, AI verticals at national outlets, and projects probing AI’s impact on politics, finance and regulation.
MAX TANI ·We need to produce a short teaser (max 50 words) summarizing the substantive content: new AI implementation examples added, total count, criteria, invitation to submit, database usage with reports, research, daily feed, free MCP server for querying via AI tools. Must be neutral editorial style, no personal pronouns, no "you".
SERGEI YAKUPOV · AI FOR NEWSROOMThe article examines why newsrooms are turning to AI‑detectors, using a recent controversy over a bestselling novel allegedly generated by AI as a case study. It explains that detector scores aim to flag content that may have been produced by language models, but they are imperfect—often producing false positives, struggling with hybrid human‑AI text, and varying across languages and model versions.
Publishers are reshaping audience engagement as Google referral traffic falls and AI‑driven search answers questions without clicks. At Press Gazette’s Future of Media Technology conference, leaders highlighted experiments with newsletters, community platforms, AI‑assisted comment moderation and recommendation tools that free staff to focus on direct reader relationships. Metrics now extend beyond clicks to app downloads, visit frequency and newsletter sign‑ups, while richer metadata and first‑party data are seen as vital for AI‑based discovery.
THE COPILOT ·More than 300 news publishers gathered in Washington to lobby for the Stealth Bot Prohibition Act, a federal bill that would force “stealth” crawlers—bots that hide their identity using residential proxies or other subterfuge—to disclose who they are and why they scrape sites.
THE COPILOT ·A U.S. district judge dismissed Penske Media’s antitrust suit against Google over AI Overviews, ruling that no formal bargain exists between publishers and the search giant. Penske, which owns Variety, Hollywood Reporter and Rolling Stone, claimed AI‑generated summaries siphoned traffic, slashing ad, affiliate and subscription revenue, and argued Google’s use of its content without compensation violated the Sherman Act.
CHARLOTTE TOBITT ·A coalition of leading publishers—including the Guardian, Financial Times, BBC, Sky, The Times of London and AP—has launched the Standards for Publisher Usage Rights (SPUR) content‑telemetry standard to track when AI systems retrieve, ground, cite or present publishers’ material, and to report that usage back to the owners.
SARA GUAGLIONE ·CNN has doubled its data team in two years to embed AI across personalization, weather summaries, newsroom organization and subscription tools. Senior VP Kendell Timmers, who joined from The New York Times, unified machine‑learning engineers, data scientists, engineers and analysts into a single group and is driving a surge in A/B testing—now over twice the previous volume—using the All Access subscription as a testing ground.
THE COPILOT ·Ole Reissmann's newsletter spotlights “small decisions, fast,” arguing that many tasks need quick, bounded answers rather than full‑text LLM output. TypeSafe’s Jev model answers simple queries in ~100 ms at $0.042 per million input tokens, far cheaper and faster than GPT‑6‑Luna, prompting OpenAI’s own decision API. It can be run locally via Ollaya for privacy‑first use.
Recent research reveals systematic political censorship in large language models, especially those developed in China. Pan and Xu tested nine Chinese and U.S. models on 145 politically sensitive questions about China, finding Chinese models refused up to 60 % of queries, gave shorter and less accurate answers, and fabricated or omitted key facts, whereas U.S. models refused less than 3 %.
PATRICK BOEHLER ·Local Media Association, in partnership with Google, is launching a six‑month AI Growth Lab for local newsrooms, accepting applications until Oct. 14. Up to 12 organizations—both small and large, especially those serving underrepresented communities—will receive Google‑led training, peer learning and one‑on‑one coaching to move from AI strategy to implementation.
LMA STAFF ·Paris Marx argues that the media and tech industry have warped the public’s perception of AI danger, fixating on speculative “superintelligence” doomsday scenarios while ignoring the concrete harms already unfolding. He cites the February 2026 Tumbler Ridge school shooting in Canada, where a perpetrator used ChatGPT to plan the attack after OpenAI banned one account but left another active, leading to lawsuits against the company.
PARIS MARX ·An analysis of 208 international news sites shows that most still only block a handful of high‑profile AI crawlers that made headlines in 2023, while many let newer bots through. One‑quarter of outlets have no robots.txt policy at all, and the median site blocks just 16 of the 170 listed AI crawlers.
LARS ANDERSON ·OpenAI announced a $10 million boost to the Lenfest Institute’s AI Collaborative and Fellowship Program, doubling its prior support and adding $5 million in cash and $5 million in cloud‑AI credits. Since its 2024 launch, the program has placed full‑time AI engineers in 11 U.S. newsrooms—primarily metro newspapers such as The Philadelphia Inquirer and The Baltimore Banner—where fellows develop custom tools like “Dewey,” an archive‑search chatbot now used across multiple outlets.
At the recent virtual ACES editing conference, professionals from journalism, PR, marketing and publishing grappled with AI’s impact on their craft, especially the overuse of em dashes as a potential “AI tell.” Speakers highlighted “flagxiety” – anxiety over content being flagged as AI‑generated – and identified hallmarks such as arch tones, stiff phrasing and excessive lists.
MATTHEW CROWLEY ·A new study published in *Information, Communication & Society* surveyed 1,138 communication scholars across 77 countries and held four focus groups to examine generative‑AI use in research. While nearly 70 % reported using AI for at least one task, most considered such use inappropriate, with only half endorsing AI for writing texts (55 %) or generating code (50 %).
More than 300 media leaders—including Condé Nast, Hearst Magazines, USA Today Co. and The Seattle Times—traveled to Washington to urge Congress to pass the Stealth Bot Prohibition Act, a federal “bad‑bots” bill that would force AI crawlers to disclose their identity and purpose instead of masquerading as human users. Organized by the News/Media Alliance, the delegation argues that stealth scraping undermines publishers by evading robots.
SARA GUAGLIONE ·Journalists with disabilities increasingly rely on AI‑driven assistive tools—such as speech‑to‑text, screen‑readers and image‑description generators—to perform core reporting tasks, and Morissa Schwartz argues that newsroom editors must embed these technologies into their AI policies. She advises editors to conduct accessibility audits of existing AI systems, involve disabled staff in tool selection, establish clear approval processes for new assistive software, provide training on ethical use, and allocate budget for personalized accommodations.
MORISSA SCHWARTZ ·NewsGuard has launched its fact‑checking chatbot, NewsGuard AI, in France, Italy, Germany, Austria and the United Kingdom, offering answers drawn solely from roughly 12,000 journalist‑vetted sources and citing each publisher in its responses. The service, priced at €8 per month, incorporates 41 editorial safeguards and a database of debunked claims, and shares subscription revenue with the cited publishers and co‑marketing partners such as Ouest‑France, Linkiesta and 5min.
THE COPILOT ·Thomson Media, together with IWM Network and NarativAI, is inviting 12 North Macedonian news outlets—local online portals, community‑focused digital media, and small radio or TV stations with an online presence—to receive a year‑long programme of free technical assistance, AI‑driven tools, SEO support, security reviews, ready‑made privacy and digital‑security templates, online training, an in‑person workshop and quarterly mentoring.
Publishers at the Digiday Publishing Summit (Sept. 14‑16, 2026, Miami) outlined a strategic shift toward a “post‑search” model as AI‑driven, zero‑click results erode Google referrals. Executives from The New York Times, Wall Street Journal, Reuters, USA Today and others emphasized building direct, loyal audiences through subscriptions, registrations, newsletters and richer on‑site experiences, while diversifying revenue with premium ads, data licensing, video and emerging AI‑powered GEO products.
SARA GUAGLIONE ·News UK’s chief technology officer revealed that roughly one‑fifth of the publisher’s website visits over the past six months were generated by non‑human agents, with about 6% of total traffic identified and blocked as malicious bots. The figure fell after the company deployed Netacea’s bot‑protection tools, moving from a block‑list of known AI crawlers to an allow‑list that only permits approved bots—though Jackson admits this approach still lacks nuance.
THE COPILOT ·We need to produce a short teaser (max 50 words) summarizing the substantive content: AI as DIY punk, critique of AI industry, rise of niche apps, shift to local DIY solutions, data ownership, end of cloud era. No names, no "you". Neutral editorial style. Let's craft ~45 words. Possible: "A
SERGEI YAKUPOV · AI FOR NEWSROOMIn an opinion piece for World News Day, AFP Global News Director Phil Chetwynd warns that while AI can quickly summarise and even fabricate news, it cannot replace the costly, on‑the‑ground reporting needed to verify facts. He recalls past “news agency killers” – Twitter, Facebook, and now AI – that promised free, real‑time information but ultimately spread misinformation when regimes shut down the internet or algorithms amplified outrage.
PHIL CHETWYND ·Rob Kelly has compiled a spreadsheet of 99 AI content‑licensing deals, revealing how publishers are adapting to a new AI‑driven bargain where their material is summarized in chatbot answers without direct clicks. Early deals, such as the New York Times’ $100 million Google license and its lawsuit against OpenAI, contrast with News Corp’s $250 million five‑year pact, showing publishers can earn “pure profit” from licensing.
THE COPILOT ·Ulrike Langer describes how she turned her overwhelming influx of 170‑230 daily newsletters into a reliable briefing system by iterating through three versions. A January Python script on her Mac failed due to hardware updates and expired credentials; a June cloud‑based version solved those technical glitches but still required her to manually sift through a 200‑item digest each morning.
ULRIKE LANGER ·Google could be compelled by the UK competition regulator to allow users on Chrome and Android to set third‑party AI chatbots as their default search engine, a move that would let alternative assistants answer queries without directing traffic to Google’s own results. The proposal threatens to divert search traffic away from Google, potentially eroding its dominance and impacting publishers who rely on click‑through revenue.
BRON MAHER ·tbd/con is an all‑day virtual unconference hosted on Gather, accessed via a single link that also works the next day. After entering a name, avatar, and granting mic/camera access, participants start on the main floor and can explore the space, including a virtual beach, before joining the opening plenary on the welcome stage. The full schedule with session rooms—Tamale Hall, Mooncake, Purple, and Teak—is listed at gazzetta.
G. AZZETTA ·The surge in AI‑driven information seeking has sparked worries about individual harms—psychological, medical, or malicious misuse—and broader societal risks, such as biased AI answers shaping public opinion. While researchers can audit chatbot outputs, such tests reveal little about real‑world impact unless they reflect the prompts people actually use.
OLIVER MARSH ·
Vincent Peyrègne
LinkedIn · Sep 23, 2026
Publishers must build AI features solving user needs, not just deploy technology. At WAN-IFRA's Paris AI Forum, Nadine Hettegger from BurdaForward argued publishers must integrate LLM capabilities or lose audience attention. The real competitive battlefield is deploying AI features addressing changing behaviors with proprietary content, keeping users on publisher platforms rather than driving them to ChatGPT. BurdaForward's tested features across 550M monthly users include: summaries, celebrity connections, live Q&A and chatbots, and personalized calculators. Their 5-person team pilots features on
The article outlines the next executive editor’s role at the Pittsburgh Post‑Gazette as it transitions under nonprofit ownership by the Venetoulis Institute. After Stan Wischnowski’s departure, the new editor will lead a leaner newsroom that must serve the community like a product organization—turning award‑winning reporting on issues such as homelessness and faulty medical devices into actionable information and partnering with local outlets (TribLive, PublicSource, WESA, etc.
YONI GREENBAUM ·U.S. traffic to roughly 100 leading news sites fell 30% to 47.6 million visits in August, down from a November 2024 peak of 68 million, according to Similarweb data cited by Axios. The decline is linked to a shift toward AI interfaces—ChatGPT, Claude, Perplexity, etc.—that pull users away from traditional search and social platforms, leaving publishers with far fewer clicks and prompting layoffs and consolidation.
THE COPILOT ·Arc XP unveiled Compass, a personalization engine that tailors news home pages for returning readers while keeping editors in charge of story placement. The tool identifies repeat visitors, remembers their last view, and surfaces missed updates via a “Catch Me Up” feature that orders new content chronologically using the newsroom’s own reporting.
THE COPILOT ·Showing the latest 50 — 793 items live in the full archive.