Study of 187 African Authors Finds No Racial Bias in Pangram's AI Detection

A study of 187 African authors has found no evidence of racial bias in Pangram, a tool that scans written text to judge whether artificial intelligence (AI) produced it. The report was published on 17 November 2026 in Publishers Marketplace's Lunch newsletter.
The study was carried out by Tuhin Chakrabarty, a professor of computer science at the State University of New York at Stony Brook. The question he set out to test was simple to state. Does the tool treat the writing of African authors differently from everyone else?
The first group was 97 Zimbabwean authors. The study scanned their books published between 2000 and 2025. That range matters because it covers two periods: before and after the arrival of large language models, the kind of AI systems that can generate fluent prose on request.
The result for that group was stark. Pangram labelled zero passages from the 97 Zimbabwean authors' works as AI-written.
Chakrabarty then repeated the experiment with books by 90 more authors from other African origins. Together, the two groups make up the 187 writers in the study. The published conclusion is that the tool shows no evidence of racial bias in how it analyses their work.
Why does a detection tool's fairness matter to a book reader? Detectors are often used to decide whether a piece of writing is original. If a tool flagged one group's sentences as machine-made more often than another's, the cost would fall on real writers, not on the software. A clean result does not settle every question about such tools. It does answer this one, for this set of authors.
The study's scope is narrow. It looked at one tool and one question, and it did not test every kind of writing or every country on the continent. Still, it is a rare example of a detector being checked against a large group of authors who are often left out of such tests.
For now, the finding is that Pangram labelled none of the Zimbabwean passages as AI, and that the wider study found no evidence of racial bias.


