Nearly 500 journalists at The Guardian walked off the job in December 2024 to protest the proposed sale of the Observer. During the strike, management allegedly used generative AI tools - including ChatGPT and Claude - to suggest headlines and write photo descriptions for visually impaired readers. The National Union of Journalists called it an attempt to undermine collective bargaining. The incident, which the outlet's management disputes, shows how quickly AI is entering daily news work and the friction it can create around accuracy, editorial values, and labor practices.
A new study by two media scholars examines how three major outlets - Reuters, the BBC, and The Guardian - are integrating AI into their workflows. The researchers analyzed public documents through 2025, including policy statements, editorial guidelines, and trial reports. Their findings reveal that no single approach dominates. Instead, each newsroom's use of AI reflects its funding model, mission, and operational pressures.
Speed as market strategy
Reuters, a global wire service, treats speed as a competitive necessity. A 30-second delay in a financial alert can cost traders real money. To meet that demand, the newsroom uses an internal tool called Lynx Insight to analyze financial data in real time, and Fact Genie to process corporate press releases. These tools let Reuters publish some breaking financial alerts in six to eight seconds, turning a race once measured in minutes into one measured in seconds.
Reuters acknowledges that AI tools can introduce factual errors. All AI-generated claims must be verified by a journalist before publication. Financial markets also provide an immediate feedback loop, with traders quickly flagging incorrect information. That same approach, the researchers note, could create greater challenges in political or investigative journalism, where errors might not surface for days or months.
AI with guardrails
The BBC, a publicly funded broadcaster, relies on editorial impartiality for its credibility. Unchecked automation could threaten that core value. The BBC has taken a cautious path, using AI primarily for summarizing long articles and reformatting local stories to match the broadcaster's house style. Its tool At a Glance creates bullet-point summaries, while BBC Style Assist helps editors rewrite regional stories in the BBC's voice, cutting a 30-minute task substantially.
No AI-assisted story is published until a journalist reviews and approves it. The BBC also published a study in October 2025 showing that AI assistants misrepresented news content - through faulty sourcing, fabricated details, or outdated information - 45% of the time. Rather than downplaying the finding, the broadcaster created AI labels - a hexagon icon paired with the phrase "How we used AI" - that now appear at the top of relevant content.
Principles before profit
The Guardian's ownership structure - a trust since 1936 - shields it from commercial and political pressure. The outlet relies heavily on voluntary reader contributions. Its original AI policy barred generative AI tools from creating text or images for direct publication, except in rare circumstances. In March 2026, the policy was updated to permit limited use for tasks like suggesting alt text, analyzing parliamentary documents, or transcribing audio, provided the tool aligns with the paper's editorial standards, is subject to human oversight, and receives explicit permission from a senior editor.
"Our approach to generative AI framework is designed to support our journalists' expertise, never replace it," a Guardian representative told the researchers. "Our journalists are always accountable for the journalism they create, and use of these tools requires absolute rigour and responsibility." When management and workers clashed during the 2024 strike, many staffers saw any use of AI to cut costs as a betrayal of those principles.
Why this matters for writers
The study introduces the concept of a "verification tax" - the time reporters and editors spend fact-checking AI outputs can erode the technology's value. If a writer spends more time correcting chatbot errors than crafting original prose, the tool becomes a burden. For writers learning to use generative AI, this underscores the need for rigorous editorial judgment. Writers aiming to build skills with these tools can find dedicated ChatGPT courses and AI for Writers courses that focus on integrating AI responsibly without sacrificing accuracy.
The researchers predict that AI will accelerate differences already underway. Commercial newsrooms will lean harder into automation. Public broadcasters will add more human oversight. Independent outlets will emphasize stories reported and written entirely by people. The most successful newsrooms, they argue, will not be the fastest or the most technologically advanced. They will be the ones that use AI wisely, treat workers fairly, and protect the trust that makes journalism a valuable resource.
This article is republished from The Conversation under a Creative Commons license. It has been updated to include The Guardian's dispute of the claims surrounding AI use during the 2024 strike.
Your membership also unlocks:
