Most of Scots Wikipedia appears to be written by an American attempting a Scottish accent, not Scots, Scotland's official language; the user has 23K+ articles
not the Scots language—with misspellings to imitate a Scottish accent. https://www.vice.com/... Evan Greer / @evan_greer : Tbh it's just nice to know the internet can still be weird https://twitter.com/... Catalin Cimpanu / @campuscodi : Not as absurd as a Romanian writing large chunks of Wikipedia's NFL rules/lore pages as a way to teach himself American football rules Guilty! 🙋♂️ https://twitter.com/... Adam Lane Smith / @thebrometheus : One of the best things I've read this year. https://twitter.com/... Zak Kukoff / @zck : Thought leaders: Never feel bad about doing a job you're under-qualified for Reality: https://twitter.com/... Mat Honan / @mat : What a bawbag! https://www.vice.com/... Simon Willison / @simonw : This is fascinating: 20,000 articles on Scots Wikipedia look to have been written by one person who was pretty much guessing at how the Scots language works, and that data may have made it into numerous machine learning models https://twitter.com/... Dara Moskowitz Grumdahl / @deardara : We just want to do good, and connect the world! says the Silicon Valley eat-everything-machine. So a teenager just starts making up nonsense and becomes all-powerful in a tiny world, and it can't be undone. https://www.vice.com/... Elizabeth May / @_elizabethmay : I know there are a lot of people who find the Scots Wiki situation quite funny, but this article does go into the bigger ramifications of what happens when one editor on Wiki has so much control over a single topic or language. They literally decide what changes go on the site. https://twitter.com/... Parker Higgins / @xor : Maybe I'm hallucinating this but isn't this the second instance of a smaller Wikipedia being almost entirely written by somebody who didn't really speak the language? I can't remember the details but I feel like I've seen this before https://www.vice.com/... Alex Griswold / @hashtaggriswold : Fully 1/3 of Wikipedia's Scots language articles were written by an American furry who had no knowledge of the language and just wrote them out like Groundskeeper Willie https://www.vice.com/... https://twitter.com/... @babelstone : Statement from the Scots Wikipedia user AmaryllisGardener https://sco.wikipedia.org/... https://twitter.com/... https://twitter.com/... @xythar : real talk: this is why you don't fucking translate things if you don't know the language. “anything is better than nothing” does NOT apply to translation https://www.reddit.com/... https://twitter.com/... @chickenpaprika : It turns out an American teen who identifies as an “INTP brony” has singlehandedly written almost the entirety of Scots wikipedia despite not knowing the language at all. Scottish redditors suspect he may have done “more damage to the Scots language than anyone else in history.” https://twitter.com/... Jason Choi / @mrjasonchoi : @NeerajKA https://twitter.com/... Mike Masnick / @mmasnick : This is an insane story... But what I can't figure out is how no one pointed this out until now? Does no one in Scotland use the Scots Wikipedia? https://twitter.com/... See also Mediagazer
Context & Ripple Effects
The story lands on a Wikipedia whose governance model was already under strain: a year earlier, administrators and Wikimedia Foundation board members were fighting over the year-long ban of a prolific long-serving administrator, exposing how much of the encyclopedia rests on a handful of trusted volunteers. AmaryllisGardener's 23,000-plus articles are the sharpest illustration yet of that dependency — one unvetted contributor supplying the majority of an entire language edition.
It also matters beyond Wikipedia itself. The related coverage on how [[a:890640|inaccurate AI translations of Wikipedia pages feed a doom spiral that marginalizes vulnerable languages]] makes Scots the case study: content written as accented English rather than Scots has plausibly already entered the training datasets that shape how machines treat the language.
First-order effects
- The Scots-speaking community inherits a cleanup problem of unusual scale: most of its edition's corpus must be audited or rewritten because it was authored by a non-speaker imitating an accent.
- Wikimedia's reliance on single prolific contributors is directly implicated — the same concentration-of-trust issue behind the earlier administrator-ban dispute, now shown to extend to entire language editions.
Second-order effects
- Other small-language Wikipedia editions face immediate scrutiny of their own contributor bases, since the Scots case shows a whole edition can be shaped by one person without speakers noticing.
- Dataset builders and ML teams that scrape Wikipedia for low-resource languages have to treat volunteer-written corpora as unverified input, adding curation cost precisely where data is scarcest.
Third-order effects
- If the pattern holds, minority-language editions of crowdsourced references become a liability rather than an asset for language preservation, pushing actual speaker communities to reclaim editorial control or build alternatives.
- The episode strengthens the argument that vulnerable languages need deliberately curated corpora — the same structural concern driving the AI-translation doom-spiral debate — rather than whatever open-web scraping happens to collect.
The trend: Minority-language editions of crowdsourced knowledge bases are being exposed as quality risks at exactly the moment they become training data for AI systems.