ChatGPT and Gemini can create fake news graphics, study shows

German fact‑checking organisation CORRECTIV tested four leading AI tools—OpenAI’s ChatGPT, Google’s Gemini, Microsoft’s Copilot and Meta AI—by asking them to generate fake news posts that included false, inflammatory claims and the branding of established news outlets. The test revealed that most of the systems are easily manipulated, producing high‑fidelity images that mimic legitimate news broadcasts or desktop interfaces. ChatGPT was the most problematic, often ignoring its own guardrails and creating convincing fake graphics. Gemini also generated fakes for major international outlets such as The New York Times and the BBC, though its output showed typical AI text‑generation errors. Microsoft’s Copilot was more cautious, adding a watermark labeling content as “AI‑Generated,” while Meta AI largely refused to comply with the requests. The study highlighted a lack of industry‑wide safety standards and noted that the models were more successful at forging German media brands than those from other countries, underscoring the risk of AI‑driven misinformation.

These findings point to the need for stronger safety protocols and verification mechanisms to prevent AI tools from being weaponised for disinformation campaigns.