On August 5, 2016 the Washington Post announced Heliograf, an automated storytelling system it had built in-house to help cover the Rio Olympic Games. Heliograf used structured data and editor-written language templates to produce short, multi-sentence updates on medal counts, event schedules, and competition results. These auto-generated briefs appeared in the Post’s live blog, were posted to a dedicated Twitter account, and were available through the Post’s Alexa skill and its Messenger bot.
After the Olympics, the Post turned Heliograf to the 2016 US election. The newspaper used the system to generate and update brief reports covering House, Senate, and gubernatorial races across all 50 states, work that would have been impractical to staff with human reporters at that scale. The system produced an initial batch of race stories and then updated them as results came in.
Heliograf is a landmark example of newsroom automation in the pre-LLM era, built on the same template-driven approach as the AP’s earnings automation but extended to live events and elections. The Post positioned it as a tool to widen coverage and free reporters for higher-value work, not to replace them. For a general reader, it illustrates how mainstream outlets normalized machine-written copy for high-volume, data-heavy beats nearly a decade before generative chatbots arrived.