In the world’s top newsrooms – whether print, digital, or broadcast – where there’s a PC, there’s now artificial intelligence. Before the rise of generative AI – including ChatGPT, Google Gemini (formerly Bard), and Microsoft Copilot – only a few organizations could afford the investment and skilled staff needed to implement AI.
Today, however, these tools have become more accessible, enabling widespread experimentation and use. The Washington Post’s Heliograf, for instance, generates short articles on sports scores and financial earnings reports from structured data. Reuters’ News Tracer employs machine learning algorithms to quickly identify breaking news stories and assess their credibility, and Lynx Insight analyzes large data sets to provide journalists with relevant insights and context for investigative reporting. Meanwhile, Corriere della Sera leverages AI to create audio versions of articles. And the list goes on.
Since 2019, the London School of Economics (LSE), along with POLIS (LSE’s media think tank) and the Google News Initiative, has conducted a global survey called Generating Change to explore AI’s role in journalism. The latest survey compiled data and insights from surveys, interviews, and public statements by journalists, experts, and media managers across 105 news outlets in 46 countries, including Il Sole 24 Ore. The participants represented newspaper professionals (28%), publishing groups (20%), broadcasters (16%), and news agencies (13%). Across all types of media, 90% of respondents reported using generative AI in news production, 80% in content distribution, and 75% for data collection and information gathering.
Content production
Initially, automated AI-driven content creation was mostly limited to short articles on stock market updates (Wall Street Journal), sports events (New York Times), and similar topics. Today, with consistent human oversight, generative AI assists in fact-checking, proofreading, natural language processing (NLP), trend analysis, headline generation, and the writing of summaries, abstracts, and newsletters. It’s also widely used for tasks like translation, data streamlining from multiple sources, spell-checking and content quality enhancement. According to the Generating Change survey, “for instance, NLP applications are assisting with factual claim-checking. They identify claims and match them with previously fact-checked ones.” This technology applies to images as well.
Content distribution
AI-powered distribution tools help reach broader audiences and drive engagement. Recommendation systems personalize content delivery based on user preferences, and other tools ensure consistent language conventions, such as enforcing British or American spelling. Speech-to-text technologies can convert written content into audio, while tools like Echobox and SocialFlow optimize social media content scheduling. Some organizations also use chatbots on WhatsApp for news distribution, while SEO tools guide authors in selecting effective keywords and trends.
Newsgathering and data collection
AI in newsrooms is particularly valuable for newsgathering on two fronts: scanning and converting various media (text, audio, images) and identifying recurring patterns in vast amounts of documentation. Speech-to-text applications like Whisper, Colibri.ai, Otter.ai, and SpeechText.ai facilitate media conversion. For data collection, Dataminr and Rapidminer are popular choices. For tracking trends, newsrooms still rely on Google Trends for web data and CrowdTangle for social media.
Why AI is being used (and what’s not working)
More than 50% of survey respondents cited efficiency and productivity as the main drivers behind AI adoption. AI helps automate repetitive tasks and streamline workflows, allowing journalists to focus on “more creative, relevant, and innovative work.” Around one-third also hope that these technologies will help expand their reach and personalize the reader experience. However, challenges remain, particularly with language limitations (AI performs best in widely spoken languages), regional accents or dialects, and the difficulties of integrating new tools with existing workflows and platforms.