Confronting gender bias in Nature's journalism

In 2021, Nature — the world's most recognized scientific journal — turned its analytical tools on itself. Researchers Natalie Davidson and Casey Greene used Genderize.io to audit 15 years of the journal's own journalism, asking a simple question: who gets quoted?

The audit

Davidson and Greene analyzed more than 16,000 articles from Nature's News, Features, and Careers sections published between 2005 and 2020. They used Genderize.io to classify the gender of people quoted in each piece, building a year-by-year picture of whose voices appeared in science's most prominent news outlet.

The trend

In 2005, 87% of quoted sources in Nature's journalism were men. By 2020, that figure had dropped to 69%. The trajectory was clear and consistent — roughly a percentage point of improvement per year. But at that rate, parity was still decades away.

The imbalance was not simply a reflection of who does science. Even accounting for the gender composition of active researchers in each field, men were overquoted relative to their representation. Journalists were disproportionately reaching for male sources, whether through habit, network effects, or the greater visibility of men in senior positions.

Why Nature published it

The decision to publish a self-audit in the journal's own pages was notable. Scientific journals rarely subject themselves to the same scrutiny they apply to the research they publish. By making the findings public, Nature acknowledged the problem as institutional rather than individual — not a matter of one journalist's Rolodex, but a pattern embedded in how science journalism operates.

The editorial accompanying the study committed Nature to tracking the metric going forward and actively working to diversify its source base.

The broader pattern

Davidson and Greene later extended this work into a full peer-reviewed paper published in eLife in 2024, which analyzed 22,001 non-research articles in Nature and found systematic underrepresentation not just of women but of people with East Asian names — a regional disparity layered on top of the gender one.

The two studies together illustrate a recurring theme in Genderize.io's academic use: the tool is most valuable not when it confirms what people already believe, but when it quantifies patterns that institutions prefer not to measure.