What 129 Wikipedia Articles Taught Me About Verification
Christopher Kurg · · revised 9 September 2026
Between 24 December 2025 and 17 April 2026 I created 129 articles on English Wikipedia, where I edit as PanoptikEU. Much of the work was about making Estonian scholarship and culture readable in English.
The hard part was the checking: what the sources actually support, what a translation keeps, and what has to be left out. This is how that went, including the questions other editors raised along the way.
The gap
Wikidata lists 22,544 people with Estonian citizenship and an article on Estonian Wikipedia. Only 4,889 of them, 21.7%, also have one in English.
A person's work should not become hard to find simply because the best sources about it are in Estonian. An English article gives readers who cannot use those sources a way in, as long as it keeps the evidence, context and distinctions of the original.
The 17,655 without an English article include over a thousand journalists and hundreds of artists, writers and translators. Not all of them would meet English Wikipedia's notability rules, so this is a research queue, not a to-do list.
Four months of articles
The 129 articles went up on 107 of the 115 days between the first and the last, a little over one a day. I remember it as about an hour on a typical working day, though publication times cannot prove that. When I collected the data in early September they had 392,382 recorded pageviews.
All 129 articles
Checking the sources
Every article starts from the Estonian original and its references, then follows the evidence: research records, university publications, official honours registers, newspaper archives and independent reviews. A citation has to support the exact sentence beside it. Having read a source is not enough if the wording says more than the source does.
Across the 129 articles that came to 1,813 citations, about 14 per article, from 540 different websites. These were cited most:
After publishing came 59 edits linking the new articles from existing pages, 29 follow-up fixes and 15 talk-page replies.
One reference, checked
In the Sven-Erik Soosaar article, one reference was suspected of being made up because its link no longer worked. On 1 August 2026 another editor, Jishara, found an archived copy of the University of Tartu page from 2019. The credit for finding it is theirs.
It taught me to keep three things apart: a source I cannot open, a source that does not exist, and a source that exists but does not say what the article claims. The archive shows the page existed. Each sentence citing it still has to be checked against what it says.
What other editors said
On Wikipedia, publishing is only the start. Other editors check sources, fix wording and question whether a subject belongs. That scrutiny is useful, but it does not take responsibility for what I submit off me.
I used AI assistance to translate from Estonian Wikipedia. English Wikipedia's guideline allows that on conditions, including knowing both languages, checking the sources yourself and crediting the translation. In July 2026 a discussion opened on the AI noticeboard about how my articles were made. After my reply, the reviewer asked for translation attribution and for checks on anything added beyond the Estonian originals. As of 9 September the noticeboard still lists that cleanup as ongoing.
- First article: Karl Pajusalu.
- An editor questions whether Marek Tamm is notable enough.
- A reviewer asks what makes Ago Luberg notable as an academic.
- Arno Vihalemm is moved to draft space for better sourcing. I reply and resubmit.
- The notability tag on Aija Sakova is removed after further review.
- A conflict-of-interest notice about a draft on an AI researcher. I reply that I have no connection to him.
- My first barnstar, an Estonian Award of National Merit.
- I join WikiProject Estonia.
- I thank a reviewer of Estonian articles.
- An exchange of thanks with another editor.
- A draft on Sam McCandlish is declined on notability.
- Last article: Avo Keerend.
- The abandoned Arno Vihalemm draft is deleted.
- A discussion opens on the AI noticeboard about how the articles were made and their quality.
- Jishara finds an archived copy of the questioned Soosaar reference.
- Kowal2701 reports that some source checks passed and decides not to go ahead with deletion nominations for now.
- I reply, explaining the translation assistance, my own review, and that corrections are my responsibility.
- The reviewer asks for translation attribution and checks on anything added beyond the Estonian originals.
Me · Other editors. Selected events; the full record is on my talk page and the noticeboard.
How long would the rest take?
Say each article takes an hour and a half, including preparation, attribution and follow-up, 6 days a week. Move the sliders to see how far that goes against the 17,655 people on the list.
85yearsAbout 0.7 articles a day, 208 a year.
Longer days are not free, though. As an illustration, not a measurement: suppose 15% of drafts contain a claim the source does not support, careful checking catches 88% of them, and after two hours the catch rate slips by 4% for every extra hour.
At 1 hour a day, about 18.0 articles in every 1,000 would keep an unsupported claim. The inputs are illustrative; the shape is the point. Past a certain length, a longer day buys more articles and a higher share of mistakes.
Change the assumptions
Keeping track
Alongside the articles I have been building a small monitor for page status, edits, notices and pageviews, so the next thing that needs checking is easier to spot and to document.
About the data
- The 129 articles are every creation in my saved list, 2025-12-24 to 2026-04-17. Not all are Estonian biographies: the list also includes the Konrad Duden Prize, Estonian youth cybersecurity competitions, Chris Olah and Jared Kaplan.
- Coverage counts people who have a Wikidata Estonian-citizenship statement and an Estonian Wikipedia article, split by whether an English article exists, in a snapshot from 8 September 2026. That is a database definition, not a list of every Estonian, and it says nothing about notability. Occupation groups overlap. My own articles have not been matched against this population, so they are not highlighted in it. The query below states the rule; running it today will not reproduce the snapshot.
- Pageviews are the sum of stored per-article counts of human (Wikimedia “user”) traffic, collected 7–8 September 2026. The start of the counting period was not saved, so it is not a lifetime total, and views are not unique readers. Wikimedia's definition.
- Deletions. An earlier snapshot found none among the tracked articles; the abandoned Arno Vihalemm draft, outside the list, was deleted on 21 July 2026. An article staying online does not prove every claim in it is right.
- Time. Dates are publication times, not a work log. The hour a day is my recollection. Citation counts show references present, not a claim-by-claim audit.
- Revisions. Revised on 9 September 2026 to separate all creations from Estonian biographies, drop an unmatched coverage overlay, label the pageviews, and keep observed issues apart from the hypothetical model. The quoted replies are unedited.
SELECT DISTINCT ?person ?et ?en WHERE {
?person wdt:P31 wd:Q5; wdt:P27 wd:Q191.
?et schema:about ?person;
schema:isPartOf <https://et.wikipedia.org/>.
OPTIONAL {
?en schema:about ?person;
schema:isPartOf <https://en.wikipedia.org/>.
}
}