The tool is a large-scale, AI-powered database that turns millions of Nazi Party membership cards from scattered, hard-to-browse PDFs into a fast, public search engine where anyone can look up individual names and basic biographical details. Using a combination of OCR systems and Google’s Gemini to read mixed fonts and old German handwriting, Die Zeit’s team extracted and standardized fields like name, birth date, occupation, and place of birth, deduplicated overlapping records, and wrapped it all in an intuitive search interface with strong historical context. It’s designed less as a simple “Nazi search engine” and more as a starting point for personal and historical investigation, making previously buried archival material accessible while clearly communicating gaps, OCR errors, and the limits of what a membership card can actually tell you about someone’s motives or actions.
AI-powered editorial assistant being co-developed by Observatorul de Nord, Internews, and NarativAI to support a small local newsroom in Soroca, Moldova. This assistant is designed to streamline routine newsroom work: organizing information, supporting editorial workflows, helping with service journalism and daily updates, and enabling responsible, transparent use of AI without replacing human reporters or editors. It fits into a broader innovation strategy that includes social video, podcasts, mobile journalism, and experiments like virtual presenters, with clear internal AI guidelines to keep core editorial judgment, ethics, and accountability firmly in human hands.
Signària is an experimental AI-powered system that lets a deaf journalism student host live radio shows by turning spoken audio into Catalan Sign Language in real time and then back into personalised synthetic speech. Built by Pompeu Fabra University and Sorensen.ai, it combines a sign-recognition pipeline (camera + linguistic rules + the student’s own writing as context) with an LLM that proposes three candidate sentences for her to choose via a foot pedal, which are then broadcast using a unique “radio sound identity” trained from her cousin’s prosody and her own vocal timbre. Instead of treating sign language as an accessibility add-on, Signària makes it a native medium, uses linguistics and a 3D engine to generate synthetic sign-language training data, and aims to create adaptable, locally controlled models that can preserve and extend minority sign languages beyond Catalan.
Document Rake is a conversational AI tool that lets Helsingin Sanomat journalists query and explore large collections of documents in natural language, similar to Notebook LM. It uses a custom RAG architecture to generate summaries and answers, lowering the time and skill needed for investigative and data‑driven reporting.
Jagran New Media’s SmartCMS is an in-house content management system enhanced with Google Cloud and Gemini AI to create a “one‑click” AI-powered newsroom workflow, where journalists can automatically generate headlines, summaries, drafts, metadata, and multi-format content variants inside the CMS, while every article passes through an AI-driven EEAT quality scoring layer and a real-time infrastructure (Google Pub/Sub + Apache Solr) that handles over 100,000 messages per second to eliminate indexing latency, boost Google Discover visibility, and safely scale high-speed, high-volume digital publishing without sacrificing editorial standards.
The Timeline Machine is an internal tool developed by Danmarks Radio (DR) that scans large collections of PDF documents, extracts dates, orders the excerpts chronologically, and generates summaries using OpenAI’s GPT‑4/3.5 models. Journalists use it to create structured metadata for archival research and to streamline open‑source investigations. The system operates with human oversight, allowing reporters to verify and refine the extracted information.
ABC News’s ABC Assist tool is an in‑house AI assistant that helps regional journalists quickly turn verified local radio bulletin scripts into short digital news briefings for the website and app, without generating any new stories itself. It ingests copy that has already been reported and broadcast, then suggests tightened text, minor style edits, reduced word counts, and a few headline options tailored for digital audiences. Journalists work in a side‑by‑side interface to verify every change, fact‑check, and make additional edits before the piece goes to an editor and digital sub for final approval. Built and trained on decades of ABC’s own Australian content and governed by ABC’s AI editorial guidance and AI Principles, the tool’s sole purpose is to scale distribution of original local journalism while freeing reporters’ time for more on‑the‑ground reporting rather than manual repurposing work.
AI-powered listening feature for EL PAÍS converts most written news articles into synthetic-voice audio, available only to subscribers on the website and mobile app. Each eligible story includes a headphone icon that opens a built-in player, allowing users to listen while doing other activities, adjust playback speed, skip ahead in 15‑second jumps, and share the piece via WhatsApp, social networks, or email. The tool is being progressively rolled out across all sections except very short or highly visual pieces, and complements other audio initiatives such as listening to the print edition and the VERA assistant, with the overarching goal of offering a more flexible, hands-free way to consume EL PAÍS’s journalism.
Suspilne Ukraine is piloting C2PA content authenticity verification, an open standard that cryptographically records an image’s origin and edit history and exposes it to users via a small “cr” mark showing Content Credentials. Integrated into Suspilne’s editorial and technical workflows, the system lets audiences quickly distinguish original newsroom photography and, later, audio and video from synthetic or manipulated media, without needing to understand the underlying cryptography. Developed with engineering and mentoring support from the Content Authenticity Initiative community, SSL.com certificates, and funding from BBC Media Action, the tool creates an end‑to‑end chain of trust from the camera to publication, helping counter deepfakes, enhance transparency, and align with requirements like the EU AI Act’s machine‑readable labeling of AI‑generated content.
Vera EL PAÍS is a premium, chat-based AI assistant that lets subscribers ask natural-language questions about current events and topics and get answers grounded exclusively in EL PAÍS’s historical archive and related brands like Cinco Días. Built with OpenAI technology, it acts as a conversational layer over the newspaper’s journalism, helping readers quickly find and explore articles, authors, recommendations (from news pieces to books, recipes, series, films, and music), and contextual background. Vera is positioned as an innovation to improve how subscribers search, navigate, and consume EL PAÍS content, and was validated in a pilot with 3,300 users before being rolled out to all premium subscribers.
Velora is a coordination layer for AI-assisted newsrooms: a software platform that plugs into publishers’ existing stacks to orchestrate multiple AI agents—research, drafting, transcription, image selection, quote/source verification, originality scoring, and story ideation—so a tiny editorial team can operate at the speed of a much larger one while keeping humans in charge of judgment calls. It automates commodity tasks, traces every claim back to original sources to catch hallucinations already cascading through journalism, and uses metrics like an “originality score” to nudge editors away from low-value rewrites toward distinctive reporting, all sold as a licensed workflow (about £300/site/month) with Velora Cycling itself serving mainly as the live demo of what the stack can do.
NewsConnect is a publishing infrastructure that lets newsrooms push their content directly into AI assistants like ChatGPT as branded, interactive apps rather than just raw training data or link snippets. It turns articles and archives into structured, queryable “journalism-as-a-service” inside the AI interface, preserving attribution, context, and updates while enabling users to ask follow‑up questions, compare, translate, and use news as a decision tool. Strategically, it treats AI not as a back-end productivity helper but as a front-end distribution channel and utility layer where audiences increasingly discover and use journalism, aiming to give publishers presence and some leverage in an environment otherwise controlled by AI platforms.
AI Extra is a browser-based, real-time media intelligence dashboard that tracks how artificial intelligence is covered across global news. Using live counts from the GDELT Project’s worldwide media index, it measures press prominence of AI-related countries, sectors, companies, universities, regulators, and political actors over the last 30 days, plus a few static metrics like citations and AGI predictions.
The Kyiv Independent uses Google's NotebookLM to analyze questions from paying members about the war in Ukraine. By feeding stripped anonymized member questions into the model, the newsroom identifies knowledge gaps and emerging topics among their audience, which then informs editorial coverage and content strategy.
Observador’s Subscription Concierge is an AI-powered assistant that holds one-to-one, negotiable subscription conversations with readers over SMS or WhatsApp at scale. It plugs into the publisher’s data stack (subs CRM, billing via Stripe, usage analytics) so it “knows” each subscriber’s plan, payment status, reading and listening habits, and past engagement, then uses an LLM to initiate personalised outreach to upgrade active subscribers or win back lapsed ones. Operating under strict editorial tone rules and commercial guardrails, it can discuss options, make tailored offers, and complete transactions, with A/B-testable prompts and a dashboard tracking conversion, response rates, and cost per transaction—aiming to outperform telemarketing while being operable directly by the subscriptions team.
The Reporter is building a tool to better understand and nurture long-term donor relationships in a low-data, long-form journalism environment. Instead of classic “big data” recommendation engines, the tool aggregates sparse signals from reading behaviour, story topics, donations, newsletters, and social media to map how “invisible readers” quietly engage before choosing to give, and how deeper engagement links to recurring support. It uses AI to classify articles into more granular, meaningful topics, then combines those quantitative patterns with qualitative input from editors and donor interviews. The goal is not aggressive optimisation, but to surface subtle patterns that help a small nonprofit newsroom design for trust, retention, and sustainable reader revenue without sacrificing its mission.
An AI-generated local news outlet launched by Justin Evans to fill the gap left by declining local news in Massachusetts. The outlet uses Claude for article generation and Google Gemini for transcription of government meetings, and has published 1,900 AI-generated stories with hundreds of paying subscribers generating $25,000 in annual revenue on Substack.
An AI-assisted local news startup launched in a community with little remaining local coverage, using AI to support reporting, content production, and operational efficiency. The project demonstrates that audiences care more about the quality and usefulness of information than the use of AI itself, provided there is transparency about how the technology is applied. Built from the ground up as an AI-assisted newsroom, the initiative aims to make local journalism economically viable in underserved communities and create a sustainable model for expanding local news coverage.
Forbes’ “The Daily Brief” is an AI-powered audio tool that automatically transforms the day’s top three Forbes stories into a concise, five‑minute podcast briefing. Using Bertie, Forbes’ internal AI system, writers feed in story summaries; Bertie then selects three pieces, generates tight audio summaries, and voices them into a single daily episode. Product and editorial teams review and fine‑tune these selections, preserving human editorial judgment while offloading most of the production work to AI. The result is a lightweight, homepage-embedded daily news brief that repackages Forbes’ core journalism into a convenient, listenable format for time‑pressed audiences.
BCN’s tool is a commercial service that helps brands get prominently and accurately represented inside AI assistants like ChatGPT, Gemini and other LLM-based agents, not just on classic search results pages. It works by analyzing how large language models describe and surface brands, then optimizing publishers’ content and metadata so that when AI systems answer user queries, they pull in richer, brand-safe, publisher-controlled descriptions. Essentially, it turns “being visible in LLM answers” into a premium media product that advertisers can buy, similar to SEO but focused on AI overviews and conversational agents rather than traditional SERPs.
USA Today Co.’s tool is an AI-assisted “shell file” system that lets editors pre-build article templates around every likely World Cup storyline before the tournament starts, automatically pulling in related archive material, stats and background so that when real news breaks they only need to drop in the fresh facts and hit publish. The goal is to have high-quality, context-rich match reports, news updates and explainers live within minutes — ideally before Google’s AI Overviews and other assistants have enough data to synthesize their own answers — preserving USA Today’s search visibility and traffic in an AI-first discovery environment.