The distinction at the heart of the report is between stopping an account from editing and reviewing what that account already changed. ADL argues that sanctions address future activity but do not, by themselves, repair earlier omissions or framing. It wants individual changes examined rather than every contribution automatically reversed.
What the 91 Percent Finding Measures
ADL examined the editing history of the banned account Iskandar323 across six English Wikipedia articles: Israel, the Gaza war, the October 7 attacks, Hamas, Zionism and Jews. Researchers identified 485 edits, of which 398 could be compared with the live text in August 2026.
Of those assessable edits, 362 remained, or roughly 91%, according to the report. The other 87 edits were excluded because they involved page moves, deleted revisions or changes that could not reliably be traced.
That percentage measures the survival of changes by one account within a defined sample. It does not measure the proportion of Wikipedia that is inaccurate. ADL itself stresses that persistence is not a judgment on whether an individual edit was correct, making a detailed review central to its recommendation.
Moving Material Can Change What Readers See
One example concerns material about the terrorist organization Hamas, its antisemitism and its use of human shields. ADL says the account moved substantial material from the main Hamas entry to a separate article about criticism of the group on October 10, 2023.
The information was moved, rather than deleted, in that example. ADL's objection is that a reader arriving at the main entry would be less likely to encounter it without opening another page. The report therefore focuses on prominence and organization as well as whether individual statements remain somewhere on the site.
For the proposed audit, ADL asks reviewers to assess removed source material and altered framing, restoring sourced information and balanced wording where bias is documented.
Visible Coordination Fell, According to ADL
The organization compared six month periods before and after its March 18, 2025 report on approximately 30 editors. It reports an 88% fall in edits made to the same page by members of that group within an hour of one another, and a 95% drop in messages exchanged between the accounts.
Those are measures of observed activity, not proof that all coordination ended or that bias fell by the same percentages. ADL also identified 19 active accounts in a July 2026 review, comprising eight from its earlier group and 11 newly documented accounts. Newly documented does not mean newly registered.
A Request for Review Within a Volunteer System
ADL's proposals include a structured content audit, stronger safeguards for contentious subjects and clearer disclosure when search services or AI tools rely on Wikipedia. The group presents these as ways to address the consequences of past editing alongside future enforcement.
The Wikimedia Foundation described the division of responsibilities in a June 26, 2024 statement: volunteers make editorial decisions, while the foundation supports the platform and does not write or determine its content.
That earlier explanation concerns Wikipedia's governance. Against that structure, ADL's new demand is specific: examine the surviving changes themselves, so that an account sanction is followed by a reasoned assessment of what readers continue to see.
