Can AI truly understand cause and effect? The new CHECKWHY dataset puts AI fact-checking to the ultimate test by tackling the complexities of "why" questions. Existing fact-verification models excel at simple claims like "who," "what," "when," and "where," but causal claims—those that probe the "why"—demand deeper reasoning. Imagine a claim like, "Military crises forced Marcus Aurelius to debase Roman silver currency." Verifying this requires more than just matching keywords. It involves understanding the chain of events: military crisis leading to financial strain, which then prompts currency devaluation. CHECKWHY challenges AI with over 19,000 complex scenarios like this, complete with evidence and structured arguments. The dataset uses a unique "argument structure" representing the logical steps connecting evidence to a claim. This structure mirrors human reasoning, allowing researchers to assess not just *if* an AI gets the answer right, but *how* it arrives at its conclusion. Early experiments show that even the most advanced AI models grapple with these causal puzzles. While they might identify individual pieces of evidence, weaving them into a coherent argument remains a major hurdle. CHECKWHY’s innovation lies in its detailed, human-like reasoning framework. This allows for more nuanced evaluation, revealing the gap between current AI capabilities and true causal understanding. The dataset is a crucial step toward more robust and transparent fact-checking, pushing AI beyond simple keyword matching and into the realm of logical deduction. It highlights the need for future research focusing on causal reasoning in AI, paving the way for systems that can truly understand and explain the "why" behind the facts.