When numbers mislead: Data interpretation in fact-checked misinformation

Abstract

Misinformation is defined as false or misleading information, irrespective of whether the person sharing it intends to deceive. Yet many misleading public claims do not invent data. Instead, numbers become misleading when denominators, time frames, definitions, uncertainty, source quality, or comparable baselines are omitted. This study asks which forms of data misinterpretation professional fact-checkers most often identify and how those forms overlap. The dataset comprises 3344 English-language fact-check records retrieved through the Google Fact Check Tools API using neutral quantitative search terms and coded with a deductive, LLM-assisted framework that used claim text, review titles, textual ratings, and article text where available. The findings show that quantitative claims were common and that misinterpretation most often involved missing context, weakly supported numbers, and distorted comparisons. Multi-label coding further shows that missing context frequently accompanied more specific errors, including misleading comparisons, causal overreach, sampling problems, and selective statistics. These results indicate that data misinformation often operates through incomplete interpretation rather than numerical fabrication. Fact-checking organizations, statistical agencies, journalists, and educators should therefore make denominators, baselines, definitions, comparable units, source provenance, time periods, and uncertainty explicit when correcting or communicating quantitative claims.