A Cautious Embrace of Automation
The people tasked with separating fact from fiction are increasingly turning to artificial intelligence, but they are far from ready to hand over the reins. According to the latest 'State of the Fact-Checkers' report from the Poynter Institute's International
Fact-Checking Network (IFCN), a growing number of organisations are integrating AI into their workflows. More than half (53.3%) now use AI tools, a significant increase from previous years. However, the consensus is clear: a vast majority, 73%, believe AI should play a supporting role, not a leading one, over the next three years. This highlights a central tension in the fight against disinformation: the need for AI's speed and scale versus the irreplaceable value of human judgment.
The Need for Speed and Scale
The sheer volume of false and misleading content flooding the internet makes a purely manual approach to fact-checking impossible. AI offers a crucial advantage in efficiency. For many organisations, the most common use of AI is for research, information gathering, and translation. Tools can scan vast amounts of data, transcribe audio and video, and identify claims that need verification far faster than any human team. One senior adviser at a Norwegian fact-checking group noted that a thorough, transparent fact-check can take anywhere from half an hour to three days, but newsrooms often need answers immediately. AI-powered tools can help bridge that gap, allowing human experts to focus their energy on more complex investigative work.
The Limits of the Algorithm
Despite its power, AI struggles with the very things that are often essential to good fact-checking: context, nuance, and intent. A study from the Reuters Institute for the Study of Journalism concluded that much of what human fact-checkers do requires a sensitivity that remains far beyond the reach of automated systems. Algorithms can be tripped up by satire, irony, or culturally specific references. There are also significant concerns about bias, as many AI models are trained on data that underrepresents smaller languages and non-Western contexts, potentially leading to inaccurate or unjust outcomes. Furthermore, AI models can 'hallucinate' or fabricate information, making them unreliable as a sole source of truth.
Keeping Humans in the Loop
The preferred solution, echoed by experts and practitioners, is a 'human-in-the-loop' model. This approach uses AI for what it does best—large-scale monitoring, claim detection, and preliminary analysis—while ensuring that human fact-checkers oversee the process and make the final judgment. This hybrid model balances the machine's efficiency with human expertise. Research has shown that while the public trusts AI for large-scale scanning tasks, they trust humans more for nuanced work that requires corroborating evidence from multiple sources. As one outlet put it, the fact-checking process remains fundamentally human because it requires a level of judgment that machines cannot yet replicate.
The Road Ahead: A Partnership for Truth
As fact-checking organisations navigate a difficult financial landscape, with many facing budget and staff cuts, the strategic use of AI becomes even more critical. The goal is not to replace journalists, but to augment their abilities. At the same time, the rise of AI-generated deepfakes and synthetic media presents a major new challenge, cited by nearly half of organisations as a significant obstacle. The future of fact-checking will likely depend on building a robust partnership between humans and machines, developing AI tools that serve human editors and empowering them to combat increasingly sophisticated disinformation. Pairing AI's power with human judgment is seen as the most promising path forward in the crucial fight for truth.
