Brisbane: The rise of artificial intelligence (AI) is reshaping journalism, raising concerns over reliability, transparency, and its implications for democratic societies, according to experts from the University of Melbourne and Queensland University of Technology.

Generative AI (GenAI) systems rely on vast datasets scraped from the internet, including news articles, academic outputs, and user-generated content. These datasets are processed and labelled by human workers, often in low-cost countries, before being used to train AI models capable of producing text, summaries, and images on demand.

“AI models like ChatGPT do not understand facts or concepts. They generate content by predicting patterns from training data, not by verifying accuracy,” the researchers noted. This predictive nature, while enabling rapid content generation, can lead to “hallucinations” where outputs appear plausible but are inaccurate or misleading.

Experts caution that AI-generated content, when unlabelled or unchecked, may blur the line between verified reporting and simulated information, undermining trust in journalism. They called for strengthened editorial standards, rigorous verification practices, and greater transparency in how AI models are trained and maintained.

“Journalism's future depends on whether institutions can govern AI use effectively, ensuring data, labour, and energy inputs are accountable,” the study said. It stressed that human oversight is crucial to maintain the credibility of news and preserve the foundation of informed public debate.

The researchers argued that without robust protocols for verification and responsible AI deployment, democratic societies risk erosion of agreed facts essential for rational decision-making.