Yandex Bot Traffic Fell to 0.4% of Its Two-Week Peak. The Metric Doesn't Say Disinformation Dropped.

Does the Yandex number measure what the headlines say it does? As of October 12, the figure going around is that bot traffic from Yandex's network fell to 0.4 percent of the maximum request volume Cloudflare recorded over the prior two weeks, after Ukrainian drones hit three Yandex data centers between October 8 and 11. The version that climbed Reddit's technology forum over the weekend called it Russian disinformation bot traffic nearly eliminated overnight.

Three strikes in four days

The first hit came Thursday, October 8, at Sasovo, a Yandex hub that Al Jazeera says houses two of the three supercomputers used to develop the company's AI model. Friday brought Kaluga, southwest of Moscow. On Sunday it was Vladimir, east of the capital, where Yandex told users on Telegram that operations at the site were "completely suspended" and that the remaining resource configuration was "unstable." The company advised customers to activate disaster-recovery plans.

Kyiv has said it is targeting places that provide revenue for Moscow's war machine, according to the BBC. President Volodymyr Zelenskyy, speaking after the Sasovo strike, said "We always respond in mirror-like fashion," pointing to Russian hits on Ukrainian data centers. Neither statement in the reporting mentions disinformation.

What the 0.4 percent counts

Cloudflare's Radar glossary defines bot traffic as any non-human traffic to a website or an app. Requests with a bot score from 1 to 29 are labeled likely automated, and 30 and above likely human. Cloudflare's bot documentation lists search engine crawlers among its examples and sorts bots by what they do on a site. Neither document sorts them by who sent them or why.

Yandex runs a search engine, which the BBC calls "Russia's Google," and a cloud platform whose status page listed at least 84 services offline, per Tom's Hardware. Crawlers, cloud automation, and a hypothetical propaganda script would all land in the same bucket if the requests came from Yandex addresses. When the data centers go dark, every one of them stops.

The number is also a ratio. It compares one day against the busiest day in two weeks, so it says nothing about how many requests that was. RuntimeWire, which traced the chart to a Bluesky post, said the public chart gives no usable baseline or time interval, so it cannot show the size of the drop or who operated the traffic.

A second measure points the same direction but does not match. Tom's Hardware, citing NetBlocks, put Yandex Cloud's overall network traffic at 30 percent of normal and Yandex LLC's at 63 percent. Those figures come from a different monitor and may describe a different moment, so Trestlewire has not reconciled them with the 0.4 percent.

Where the disinformation label came from

The Tom's Hardware article by Jowi Morales speaks only of bot traffic. Its web address still reads russian-disinformation-bot-traffic-nearly-eliminated-overnight, but neither the current headline nor the text uses that word. The Maltese Herald goes furthest, reporting a significant reduction in Russian bot propaganda on social media, including pro-Trump and pro-AfD content. The text gives no dataset or account counts, and its only link for the claim is a single post from an X account called TheOhioRob. It also says Russian state-aligned media and propaganda sites suffered outages, which is a more direct effect of servers going down.

The strongest case against skepticism is simple. Some influence operations may well have run on Yandex infrastructure, and an outage would hit them like everything else. Nothing published that Trestlewire could find shows it either way. The outlets that report the bot figure appear to rest on the same Cloudflare Radar chart, so three headlines may be one data point.

No one has published what share of Yandex's automated requests were search crawls versus anything else, and that base rate is the missing piece. What would settle it is a category breakdown of those requests from Cloudflare, or a researcher tracking named influence accounts across the same dates. Trestlewire found neither as of October 12.