After Ukraine struck Russian data centres, a claim travelled quickly: the internet had gone quieter because Russian bot networks were down. President Volodymyr Zelensky suggested on Monday there was "less hate, less division and fewer provocations" online because of "fewer bots".
BBC Verify asked researchers who track these operations. They found no evidence the strikes have significantly reduced them.
What the chart actually showed
The claim spread on the back of charts built from Cloudflare data showing bot activity on Yandex's network declining. The inference was that Kremlin-linked disinformation accounts had gone offline.
Cloudflare clarified what its figures count: all automated internet requests to websites on its network — which includes bots crawling pages so they can appear in search results. The company said the drop was "not necessarily bot behaviour on social media platforms".
In other words, the measurement includes the ordinary machinery of the web. If a search engine's crawlers go down with a data centre, the chart falls, and nothing has happened to a single disinformation account.
How far it travelled
Tomáš Zdechovský, a Czech member of the European Parliament, shared a post on X reading: "If confirmed, this is telling: Russian data centres get hit, and online activity around AfD suddenly plummets." It has been viewed more than four million times.
He has since said he "did not present it as a clear fact". The post remains pinned at the top of his profile.
Researchers from dTeam, a US-based group tracking Moscow-linked influence operations, told BBC Verify they had not seen any traffic change from known Russian accounts.
The strikes themselves are real
None of this disputes the attacks. Ukraine has hit three of Yandex's five data centres since 8 October, in response to Russian attacks on Ukrainian internet hubs, and the second strike left Yandex saying it was struggling to maintain services.
What is unsupported is the inference drawn from them.
What it means in Bangladesh
The failure mode on display here will arrive in Bangladesh during election season, and its shape is more useful than its subject.
The claim was attractive because it was satisfying, carried a real chart from a real company, and confirmed what its audience already hoped. Nothing was fabricated. The error was entirely in what the chart was measuring — a definition problem, not a data problem, and definition problems do not look wrong at a glance.
The question to carry from this is the one BBC Verify effectively asked: what is being counted? A graph labelled "bots" may count search crawlers. A graph labelled "fake accounts" may count suspended accounts of every kind. A figure for "misinformation removed" may count duplicates of one post.
A Bangladeshi parallel already exists in this archive: 1.8 million sign-ups for 5,000 seats looks like overwhelming demand until you ask what a sign-up is, at which point it is a verification failure. Same number, two opposite stories, and the label decides which one you get.
Why the correction never catches up
Cloudflare clarified what its data measured, and BBC Verify asked the researchers. Both corrections exist and are easy to find. Neither will travel as far as the claim.
The post that spread the inference has been seen more than four million times and is still pinned. A clarification from a networking company about what counts as an automated request has no comparable audience, because it is a definitional note and definitional notes are not shareable. That asymmetry is structural rather than anyone's fault, and it is the reason the first telling of a story matters more than the correct one.
Source: BBC Verify.
