As AI-generated content continues to appear across social media, news outlets, and individual feeds, experts are urging the public to take extra care in verifying images and videos. On the 10th annual International Fact-Checking Day, authorities highlighted ways to identify AI-generated disinformation, noting its presence in topics ranging from the US-Iran conflict to European elections.
A recent study published in PNAS Nexus surveyed 27,000 people across 27 EU countries to assess how accurately they could identify AI-generated headlines. Researchers found that nearly half of AI-created headlines were considered “mostly” or “completely real,” slightly higher than 44% of human-written headlines. Participants were more likely to share and trust AI-generated stories when they appeared linked to real events. The study indicated that distinguishing human-created content from AI-generated material remains a significant challenge.
Experts recommend a combination of visual scrutiny, research, and fact-checking to reduce the risk of spreading false content. The Global Investigative Journalism Network (GIJN) advises looking for visual cues in images and videos, such as objects that disappear, humans with unrealistic features, or overly polished appearances. For example, a person who appears magazine-ready in a conflict zone might be a sign of AI generation, as could unusually smooth or uniform skin textures.
Research and verification tools are also critical. Reverse image searches using Google or platforms like TinEye can help trace the first appearance of a photo or video online. Technical tools may reveal watermarks or metadata indicating whether content is AI-generated. Google’s Gemini AI, for example, embeds invisible digital watermarks in images using SynthID, which can be detected to confirm authenticity.
Fact-checking organizations provide additional guidance. In Europe, groups such as the European Fact-Checking Standards Network (EFCSN), European Digital Media Observatory (EDMO), and EUvsDisinfo publish research, trends, and verified debunks of AI-generated misinformation. The Database of Known Fakes collects fact-checks from journalists and researchers to help users confirm the legitimacy of content.
AI detection software can also assist, though accuracy varies. Tools like Winston AI, TruthScan, and Originality AI can flag AI-generated images or text, but the absence of a watermark does not guarantee authenticity. Users are urged to remain cautious and not immediately share content that evokes strong emotional responses. Comments or reactions from other users may provide clues about whether an image is genuine or altered.
Experts emphasize that while it is not always possible to determine if content is AI-generated, adopting a careful and critical approach can reduce the spread of misinformation. Slowing down, checking multiple sources, and relying on verified fact-checkers remain the most effective steps in navigating today’s complex information environment.
