A bot called Lsjbot is responsible for 24M out of 29.5M edits and 99.2% of articles for Wikipedia Cebuano edition; all but 5 of the top 35 editors are also bots
Kyle Wilson is an administrator on the English Wikipedia and a global user renamer . He does not receive payment … Tweets: @motherboard and @motherboard Tweets: @motherboard : One bot is responsible for nearly all of the edits and articles on Wikipedia's Cebuano language edition. https://www.vice.com/... @motherboard : “Wikipedia consensus is that an unedited machine translation, left as a Wikipedia article, is worse than nothing.” https://www.vice.com/...
Context & Ripple Effects
The Cebuano Wikipedia is, in practice, a machine-written encyclopedia: the concentration study showing 1% of editors produce 77% of Wikipedia's content found the extreme version of that pattern here, where a single bot, Lsjbot, accounts for 24M of 29.5M edits and 99.2% of articles, and all but five of the top 35 editors are bots. The VICE report lands in a lineage of mass-content scandals on small-language editions — months later, the same outlet documented Scots Wikipedia being written by an American faking a Scottish accent rather than a speaker of the language.
What makes the Cebuano case more than a curiosity is the site's own consensus, quoted in the report, that unedited machine translation left as an article is worse than nothing — and the fact that AI models now train on these pages, a feedback loop MIT Technology Review later traced as a doom spiral for vulnerable languages. The governance answer arrived years after Lsjbot: a 2025 policy giving administrators fast-delete authority over AI-generated articles.
First-order effects
- Cebuano Wikipedia's article count is effectively a bot output statistic, not a measure of human coverage — the edition's real editor base is five humans among its top 35, and its content is largely unreviewed machine translation.
- Human administrators on the project inherit a cleanup problem at machine scale: per Wikipedia's own consensus, most of the 99.2% bot-written corpus fails the bar for acceptable articles.
Second-order effects
- Small-language editions become cautionary data for AI training pipelines: models ingesting Cebuano and Scots-style content absorb machine-translated or non-speaker text, degrading output for the very language communities the editions nominally serve.
- The pattern forces Wikipedia's governance to scale enforcement to bot-speed vandalism-by-volume — the same bot-editing-war dynamics documented in the study of bots undoing each other's edits, now applied to article creation rather than reverts.
Third-order effects
- Wikipedia's headline growth to 65 million articles increasingly overstates real knowledge coverage, pushing the project toward quality-based metrics and deletion authority — the direction the 2025 admin fast-delete policy codifies.
- If the pattern holds, small-language editions face a structural choice between bot-built breadth with no human verification or shrinking back to genuinely human-curated scope, with AI translation quality determining which path is viable.
The trend: Wikipedia's expansion is increasingly driven by machine-generated content in small language editions, forcing the project to shift from growth metrics to enforcement and deletion as its core governance challenge.