Spotting Misleading Charts and AI-Generated Data Visuals

Charts compress information, which makes them useful and easy to misuse. Begin with the title, axes, units, date range, source and definitions. A bar chart with a truncated vertical axis can exaggerate differences; a line chart with a compressed time window can hide long-term context. Neither design is automatically wrong, but the scale should be visible and appropriate to the claim. Check whether the chart starts at zero when comparing bar lengths, whether a log scale is labeled, and whether category intervals are equal. Next inspect what was included. A percentage without a denominator, a survey without its question wording, or a trend without a baseline may be impossible to interpret. Compare like with like: same units, population, time period and definitions. Look for missing data, cherry-picked start dates, inconsistent axes between panels and uncertainty that was omitted. Find the original report or dataset instead of relying on a screenshot or a repost that may have detached the chart from its notes. Generative image tools can create convincing infographic layouts and text, but a realistic appearance does not show that the numbers came from a real dataset. A model may invent a source or render labels incorrectly. Treat any chart without a traceable source as an illustration until its values can be checked. Search the title or attribution, visit the linked report and reproduce a few key values from its tables. If no source exists, say that the data are unverified rather than asserting that the image must be AI-generated. A fair critique identifies the specific design or evidence problem and explains how it affects the conclusion. A chart may be technically accurate but framed selectively, or visually awkward while using valid data. Separate those questions: are the numbers supported, and does the display communicate them honestly? This prevents an accusation about intent from replacing a check of the evidence.