Public engagement through outreach is a key mechanism for learning about science and to communicate societal impacts of government‐funded science. However, outreach effectiveness could be limited if approaches are not evidenced based. Partnerships with cognitive scientists who study fundamental learning processes suggest helping people learn how earthquakes happen would improve understanding of hazards and motivation for preparedness. We used a seismology–cognitive science partnership to evaluate understanding by nongeoscientists of widely viewed U.S. Geological Survey (USGS) ShakeMap and Prompt Assessment of Global Earthquakes for Response (PAGER) products originally developed for technical use. After discussions with USGS staff, we constructed 13 free response questions probing understanding of these products, including the graph literacy scale (GLS) for comparison. Through Zoom interviews of 101 participants, we found poor performance (28% correct) on the PAGER and ShakeMap questions despite good performance (76% correct) on the GLS. When coding free responses, we identified an average of 12.4 misconceptions per participant, approximately 1 per question. Five misconceptions were observed in over half the participants, including how the reports are constructed and used in real time, treating reports using modeling or simulations as containing only direct observations, and difficulty interpreting probabilities. The performance and misconceptions led us to develop a revised PAGER with scatter plots to visualize fatality and damage probabilities among other simplifications. We used a multiple‐choice survey to assess understanding using the original and new visualizations with another 100 participants. We found significant improvement: 49% correct when seeing the revised visualization first, compared with 36% when seeing the original visualization first. Finally, we surveyed 147 participants on whether they could distinguish expert‐defined shaking and damage descriptors for three macroseismic terms (moderate, strong, and severe). Participants struggled, choosing greater consequences than intended for lower intensity terms. Overall, our findings suggest that the key aspects of these earthquake information products are poorly understood by nongeoscientists in aseismic regions, but adjustments can improve effectiveness.