What 129 Wikipedia Articles Taught Me About Verification

By Christopher Geneth Kurg · PanoptikEU on English Wikipedia · Revised 9 September 2026

My saved contribution record covers 129 article creations, with much of the work focused on making Estonian scholarship and culture accessible in English. The central challenge was checking what the sources supported, what the translation preserved, and what needed to be left out.

This page documents the work, the questions other editors raised, and the limits of the conclusions I can draw from it. Recorded activity is separate from estimated working time; observed problems are separate from hypothetical risk calculations. The aim is to make the process examinable, not to claim it was flawless.

English article · 4,889Estonian article only · 17,655
One dot per human recorded with Estonian citizenship and an Estonian Wikipedia sitelink in the 8 September 2026 snapshot. My article list has not been matched to this population, so it is not plotted as an additional or highlighted group.
Coverage snapshot
21.7%
Article creations
129115 days
Publishing days
107
Recorded pageviews
392,382

Pageviews are stored request counts across all 129 titles, not unique readers. Reporting period: not retained in the export. Data collected 7–8 September 2026; these are not live totals.

What the numbers include, and what is missing

Article population. 129 rows in the saved creation list, dated 2025-12-24 to 2026-04-17. This is not an exclusively Estonian or exclusively biographical set: it includes Konrad Duden Prize, Estonian youth cybersecurity competitions, Chris Olah and Jared Kaplan. No verified count of the Estonian-biography subset is asserted here.

Coverage population. The original export describes humans with a Wikidata Estonian citizenship statement and an Estonian Wikipedia sitelink, grouped by whether an English sitelink exists. This is a database definition, not a complete list of Estonians or a determination of English Wikipedia notability.

Snapshot and matching. Coverage figures are from the recorded 8 September 2026 snapshot. The repository stores aggregates, not the exact original query response or an article-to-QID join. The query below expresses the stated inclusion rule for a future reproducible export; rerunning it now does not reproduce a historical snapshot. Count distinct QIDs, then match the creation list to English sitelinks within that same export before highlighting any contribution subset. Occupation groups overlap and must not be summed.

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/>.
  }
}

Pageviews. The displayed 392,382 is the sum of stored per-title values for all 129 rows, without selectively excluding the two AI-researcher articles. The earlier collection note identifies Wikimedia user-classified traffic and a 7–8 September 2026 retrieval. The request start and end dates are not retained, so the reporting period is unknown. This is neither a unique-reader count nor a claimed lifetime total, and it should not be used for growth comparisons. Wikimedia metric definition.

Retention is separate from quality. The earlier snapshot reported no deletions among its tracked mainspace titles, while recording the abandoned Arno Vihalemm draft as deleted on 21 July 2026. That draft outcome is outside the displayed creation list. This is not a fresh deletion-log audit, and an article remaining online does not establish that all its claims have been validated.

Time and verification. Dates and hours show when pages were published. An hour on a typical working day is my recollection, not a measured labour log. Preparation, attribution and later corrections also take time. Citation counts show references present, not independently certified claims.

A publishing record, not a timesheet

The saved list contains 129 creations across 107 active publication dates, spanning 115 calendar days. That is 1.12 creations per calendar day. I remember spending roughly an hour on a typical working day, but publication timestamps cannot verify that estimate or account for preparation and later corrections.

065129129cumulativeper day · maximum 42025-122026-012026-022026-032026-04
Cumulative creations and daily counts. Hover a date for titles, or open the complete list below.
081600:00 · 0 articles0001:00 · 0 articles02:00 · 3 articles03:00 · 3 articles04:00 · 4 articles05:00 · 1 article06:00 · 4 articles0607:00 · 5 articles08:00 · 0 articles09:00 · 9 articles10:00 · 16 articles11:00 · 8 articles12:00 · 6 articles1213:00 · 10 articles14:00 · 4 articles15:00 · 14 articles16:00 · 10 articles17:00 · 5 articles18:00 · 7 articles1819:00 · 5 articles20:00 · 5 articles21:00 · 5 articles22:00 · 2 articles23:00 · 3 articles
Publication hours, labelled Tallinn time in the original export. No session length is inferred.
Inspect all 129 recorded article creations
Article titles and publication timestamps from the saved export
ArticleRecorded publication
Karl Pajusalu2025-12-24 05:20
Sven-Erik Soosaar2025-12-25 03:51
Sirje Olesk2025-12-28 02:51
Marek Tamm2025-12-29 08:31
Els Oksaar2025-12-31 08:52
Konrad Duden Prize2025-12-31 11:11
Michael Heltzer2026-01-01 10:05
Hasso Krull2026-01-01 11:05
Fedor Bucholtz2026-01-02 18:54
Henrik Hololei2026-01-03 19:06
Sven Sakkov2026-01-04 17:24
Ain-Elmar Kaasik2026-01-05 21:04
Aksel Kipper2026-01-05 21:04
Armin Tuulse2026-01-06 01:12
Ester Oras2026-01-08 02:31
Olivia Saar2026-01-09 05:27
Kadri Aavik2026-01-09 05:50
Jaan Vahtra2026-01-10 07:55
Eva Piirimäe2026-01-11 08:42
Toomas Rein2026-01-12 09:14
Hardi Tiidus2026-01-13 08:32
Ago Luberg2026-01-14 09:30
Aleksei Talisainen2026-01-14 11:26
Hendrik Sepp2026-01-15 07:21
Estonian youth cybersecurity competitions2026-01-15 07:41
Tarmo Uustalu2026-01-15 13:18
Rain Ottis2026-01-16 05:10
Jakob Kübarsepp2026-01-17 13:35
Katrin Tiidenberg2026-01-18 12:34
Aarand Roos2026-01-19 09:58
Arne Merilai2026-01-20 08:30
Mai Levin2026-01-22 09:24
Enn Põldroos2026-01-23 08:36
Ülo Lepik2026-01-24 09:19
Tuudur Vettik2026-01-25 08:54
Raimund Hagelberg2026-01-26 08:15
Raimo Pullat2026-01-27 00:04
Peeter Saari2026-01-28 09:19
Aija Sakova2026-01-29 08:26
Jared Kaplan2026-01-29 18:08
Dimitri Kaljo2026-01-30 11:13
Ilmar Tomusk2026-02-01 11:19
Ants Viires2026-02-02 15:44
Tõnu Parming2026-02-03 11:52
Evi Tihemets2026-02-04 12:40
Chris Olah2026-02-05 13:35
Jüri Selirand2026-02-06 13:44
Ago Künnap2026-02-07 13:34
Ülo Kaasik2026-02-08 16:20
Ferdinand Eisen2026-02-09 17:55
Nikolai Kormašov2026-02-10 16:44
Ülo Niinemets2026-02-11 14:02
Rein Müllerson2026-02-12 07:54
Arved-Ervin Sapar2026-02-13 08:21
Kuno Areng2026-02-14 16:12
Jaak Tomberg2026-02-15 14:36
Liina Lukas2026-02-16 17:24
Agu Laisk2026-02-17 18:57
Maaja Vadi2026-02-18 13:35
Kalle Kasemaa2026-02-19 19:34
Ilmar Trull2026-02-20 08:08
Vive Tolli2026-02-21 21:57
Ingo Normet2026-02-22 19:14
Aleksander Loit2026-02-23 04:50
Hendrik Krumm2026-02-24 08:17
Jaak Soans2026-02-25 07:33
Art Leete2026-02-26 07:39
Lea Tormis2026-02-27 07:37
Urmas Varblane2026-02-28 10:32
Malle Talvet-Mustonen2026-03-01 14:12
Peeter Vihalemm2026-03-02 12:38
Sirje Runge2026-03-03 13:43
Juhan Kahk2026-03-04 10:33
Mikk Sarv2026-03-05 11:47
Martin Ehala2026-03-06 09:32
Viljar Ansko2026-03-07 10:11
Eve Pormeister2026-03-08 16:49
Peeter Kann2026-03-09 17:37
Gennadi Vainikko2026-03-10 19:25
Anu Saluäär2026-03-11 15:51
Andres Uibo2026-03-12 14:44
Raul Meel2026-03-13 04:06
Jüri Kuuskemaa2026-03-14 14:31
Märt Väljataga2026-03-15 14:08
Lauri Mälksoo2026-03-16 14:19
Hellar Grabbi2026-03-17 04:46
Ottomar Maddison2026-03-18 16:15
Mati Mandel2026-03-19 09:57
Jüri Sepp2026-03-20 15:06
Lauri Leesi2026-03-21 13:07
Mati Kaal2026-03-22 15:15
Ants Üleoja2026-03-23 18:07
Jaan-Eik Tulve2026-03-24 05:06
Valdur Mikita2026-03-25 07:19
Toomas Hiio2026-03-26 04:32
Jaanus Vaiksoo2026-03-27 07:25
Bruno Pao2026-03-28 08:03
Mati Karelson2026-03-29 09:21
Arvi Altmäe2026-03-30 10:02
Indrek Treufeldt2026-03-31 07:09
Daniele Monticelli2026-04-01 09:50
Mihkel Veiderma2026-04-02 11:22
Arne Mikk2026-04-03 13:24
Jaak Järv2026-04-04 12:17
Ljubov Kisseljova2026-04-04 12:53
Urmas Nemvalts2026-04-05 10:45
Ott Valdeko Kurs2026-04-05 10:58
Voldemar Ilja2026-04-06 19:49
Irina Belobrovtseva2026-04-06 19:58
Jaan-Mati Punning2026-04-07 13:11
Aili Aarelaid-Tart2026-04-07 13:19
Ene-Liis Semper2026-04-08 17:58
Ülo Langel2026-04-08 18:31
Tiit Ojasoo2026-04-09 00:43
Ilona Martson2026-04-09 00:52
Mati Õun2026-04-10 12:09
Mihkel Kaljurand2026-04-10 12:22
Valter Lang2026-04-11 14:50
Andrus Saar2026-04-11 15:34
Richard Kleis2026-04-11 23:04
Alo Raun2026-04-11 23:08
Aune Valk2026-04-13 01:32
Jaan Elken2026-04-13 01:34
Heldur Niit2026-04-15 15:51
Raimo Raag2026-04-15 16:19
Herald Eelma2026-04-16 12:12
Lembit Jaanits2026-04-16 12:30
Mati Palm2026-04-17 07:16
Avo Keerend2026-04-17 07:40

A pool of potential subjects

The snapshot records 17,655 people with an Estonian article but no English article. The occupation chart describes that missing population only. It does not overlay my unmatched creation list or assume every candidate meets English Wikipedia's notability requirements.

journalist1,026artist661writer599translator473university teacher398historian313painter309music educator275poet257economist257chemist253architect252composer233linguist227physicist221sculptor208literary scholar149printmaker143mathematician142
Wikidata occupation groups overlap: one person may appear in several. These values must not be added together as distinct people.

Why make this work available in English?

A person's work should not become difficult to discover simply because the best sources are in Estonian. English-language biographies offer an entry point for readers who cannot use those sources directly. My aim is to make that work easier to find without losing the evidence, context or distinctions in the original material.

That practical aim does not require a position on changing Estonia's principal spoken language. The broader language-policy argument is separate from this account of the work.

The work behind a publishable article

The workflow starts with 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 particular statement beside it. Reading a source is not enough when the wording overstates what it says.

Recorded citations
1,81314.1 per article
Source domains
540
Link-in edits
59
Follow-up fixes
29+ 15 replies
Riigi Teataja (state gazette)57DOI (journal articles)49ERR Kultuur40University of Tartu37DIGAR (national digital archive)32Office of the President (decora…32Estonian Academy of Sciences31Sirp30ETIS research information system29
Recorded domain usage across the creation list. Library authority records were excluded in the original tally. Citation presence is not a claim-level quality audit.

A worked reference check: Sven-Erik Soosaar

A reference in the Soosaar article was suspected of being fabricated when its original link was unavailable. On 1 August 2026, Jishara supplied an archived University of Tartu page from 2019 in the noticeboard discussion. Credit for finding it belongs to that editor. Discussion · Archived reference.

Question

Did the cited page exist, or was the reference invented?

Evidence found

Another editor supplied a dated archive capture of the original university URL.

Limit of the result

Recovering a source addresses its existence. Each statement citing it still needs to be compared with the source text.

The lesson is to distinguish an inaccessible source from a nonexistent one, and to distinguish both from a source that exists but does not support the claim. This example does not clear unrelated wording, attribution or sourcing questions.

A capacity calculator, not an accuracy study

This calculator explores assumptions, not measured accuracy. It shows how a fixed labour budget behaves when article time, overhead and hypothetical checking rates change. The risk curves do not estimate the actual error rate of these articles.

Hypothetical risk assumptions

The defaults of 15%, 88% and 4% are illustrative choices, not measured rates. Community comments are not a random audit and cannot calibrate the number of undiscovered problems.

Review slots / day
0.7one person
Slots / year
208
Years for this pool
84.9
0.03.06.01h2h3h4h5h6h7h8hreview slots per day
Linear only because this scenario holds time per article constant, including the overhead entered.
0.017351h2h3h4h5h6h7h8haffected articles per 1,000
Fading catch rate · Constant catch rate. The selected scenario has the highlighted point.

Selected assumption: fading attention. At 1 h/day, 18.0 articles per 1,000 would retain at least one unsupported claim under this model.

The curves coincide for the first two hours or when fade is zero. Constant attention is not a proven benefit of a checkbox or a cost-free control. Add reviewer handovers and rework to the time budget; independent second reviews would require additional labour.

The pool is 17,655 Wikidata candidates, not 17,655 articles ready for publication. Notability checks, unavailable sources, coordination and continuing maintenance can change the workload.

Publication, scrutiny and responsibility

Publication is followed by community scrutiny. Other editors may check sources, correct wording, question notability or identify problems I missed. That scrutiny is valuable, but it does not replace my responsibility for the material I submit or guarantee review before a reader sees a page.

Translation assistance

Drafting assistance must stay within the applicable translation rules. A fluent sentence can still misrepresent its source.

Editorial responsibility

Check the original and the cited evidence, remove unsupported additions, and provide the required translation attribution.

Community scrutiny

Patrol, corrections and discussion happen after publication too. Respond to specific issues and link completed fixes.

English Wikipedia's LLM guideline, checked 9 September 2026, permits translation assistance under specified conditions, including competence in both languages, source verification, compliant wording and attribution. Personal review is not a general exemption for AI-generated prose beyond translation. New-page patrol is not a certification that every claim has been independently audited.

The review record

Selected events from the saved record are summarised below. Expand the stored exchanges to read their wording. These are not a complete edit history or a substitute for the talk page and noticeboard.

  • 2025-12-24First article in the saved creation list: Karl Pajusalu.
  • 2025-12-29Notability discussion about Marek Tamm.
  • 2026-01-14A reviewer asks about the academic-notability basis for Ago Luberg.
  • 2026-01-21Arno Vihalemm moved to draft space for sourcing; a response and resubmission followed.
  • 2026-02-01The saved record notes removal of the Aija Sakova notability tag after additional reviews.
  • 2026-02-06Conflict-of-interest notice relating to a draft on an AI researcher.
  • 2026-02-16An Estonian Award of National Merit barnstar.
  • 2026-03-09Joined WikiProject Estonia.
  • 2026-03-27A thank-you to a reviewer of Estonian articles.
  • 2026-04-09An exchange of thanks with another editor.
  • 2026-04-11The saved record notes a notability-based decline of the Sam McCandlish draft.
  • 2026-04-17Last article in the saved creation list: Avo Keerend.
  • 2026-07-21The saved record notes deletion of the abandoned Arno Vihalemm draft, separate from the tracked article set.
  • 2026-07-25AI-noticeboard discussion opened about the creation process and article quality.
  • 2026-08-01Jishara supplied an archived version of the questioned Soosaar reference.
  • 2026-08-21Kowal2701 reported that some source checks passed and declined to proceed with LLM deletion nominations at that point; this was not a corpus-wide clearance.
  • 2026-09-06My response explained translation assistance, personal review and responsibility for corrections.
  • 2026-09-06The reviewer requested translation attribution and checks on additions beyond the Estonian originals. Completion is not established here.

The monitor

Alongside the articles, I have been developing a monitor for page status, edits, notices and recorded pageviews. Its useful role is to make the next check easier to identify and document. A list of missing subjects is a research queue, not an invitation to publish without assessing notability and sources.

Written by Christopher Geneth Kurg, who edits English Wikipedia as PanoptikEU. Activity, citation and coverage figures retain the original 7–8 September 2026 export. The article was revised on 9 September 2026 to distinguish all article creations from an unverified Estonian-biography subset, remove unmatched coverage overlays, label pageviews and their missing reporting interval, and separate observed issues from hypothetical model inputs. Public-source links are provided beside the relevant discussion. The stored reply texts have not been rewritten.