The Washington Post developed Heliograf internally to cover the 2016 Rio Olympics and US elections. The system published over 850 articles during these events, focusing on races and competitions where human reporters could not physically be present.
The Technical Implementation
Heliograf connects to structured data feeds from election boards and sports organizations. When vote totals update in a county commissioner race in rural Virginia, the system generates a brief article with current percentages, candidate names, and historical context from previous elections. Templates include conditional logic: if a race tightens within 2 percentage points, the article notes the competitive nature.
What Gets Automated vs What Gets Reported
During the 2016 election, Post reporters concentrated on 12 major races requiring interviews, polling analysis, and political context. Heliograf handled 500 down-ballot races in Virginia and D.C. These automated articles averaged 150 words and updated as results changed throughout the night. No human reporter could cover 500 simultaneous events.
The Beginner View
New journalists worry this technology eliminates local reporting jobs covering city councils and school boards.
The Expert Assessment
Experienced editors know these hyperlocal races rarely received any coverage before automation. The Post didn't replace reporters; it filled coverage gaps. When Heliograf identifies an unexpectedly close race through data anomalies, it alerts human reporters who can investigate why traditional voting patterns shifted. The system extended reach rather than replaced judgment.