This article reflects information available as of 27 September 2026.
In brief
Of the 30 million paper ballots the Central Election Commission credited to United Russia in the 18–20 September State Duma election, at least 18.4 million fall within the statistically anomalous part of the result, according to a calculation by the investigative outlet IStories using the method developed by physicist Sergei Shpilkin. Researchers attribute that anomaly to ballot stuffing and rewritten protocols. It amounts to more than 60 per cent of the ruling party's paper vote. Independent estimates of United Russia's genuine share, produced by different methods, cluster between 34 and 40 per cent against an official 57.8 per cent; estimates of genuine turnout run to 40–42 per cent against an official 59.80 per cent. Meduza, having compared the distribution of polling stations in 2021 and 2026, calls this the dirtiest parliamentary election of the Putin era.
The official outcome is 349 of 450 seats for United Russia, more than any party has held in the Duma's history. But the result was not assembled by counting alone. The only party with an anti-war platform was struck from the federal ballot by the Supreme Court six weeks before polling. The required turnout was manufactured before the count began: through corporate mobilisation of public-sector workers who reported to their managers, and through online voting, which was extended from seven regions to thirty-three and sealed off from scrutiny. Observers were turned away by the hundred. And a party that was still below the threshold on the morning after the vote was lifted to exactly 5.00 per cent within an hour.
In 2026 manual adjustment was no longer confined to the ruling party's percentage. It extended to the composition of the Duma as a whole: who was allowed to run, who cleared the threshold and how the seats were divided among loyal parties. A chamber assembled this way will convene ten days after the vote, faster than any Duma in modern Russian history.
The official tally: a record majority
The Central Election Commission (CEC) certified the final results on 25 September. United Russia took 349 seats: 140 from the federal party list and 209 of the 225 single-member constituencies. The Communist Party (KPRF) won 37 seats (5 in constituencies), the Liberal Democratic Party (LDPR) 23 (2 in constituencies), New People 19 (all from the list) and A Just Russia 17 (4 in constituencies). One further constituency seat went to Rodina, won by the party's leader Alexei Zhuravlev, whose lawsuit had removed Yabloko from the ballot; four went to independents. United Russia holds more than two-thirds of the chamber, a constitutional majority.
Compared with 2021, United Russia gained 25 seats, New People six and the LDPR two. The KPRF lost 20 seats and A Just Russia ten. The CEC did not publish final list percentages when it announced the seat allocation; the last widely reported figures, after 96.95 per cent of protocols had been processed, were United Russia 57.83 per cent, KPRF 13.76, LDPR 9.01, New People 7.96 and A Just Russia 5.00.
Official turnout was 59.80 per cent, or 67.4 million voters. Russian state media called it a record. It was not: turnout was higher in 1995, 1999, 2007 and 2011. It is a record only against 2016 and 2021 (47.88 and 51.72 per cent). The official result also outran the forecasts of the state's own pollsters. On the eve of the vote VCIOM projected turnout of 51–53 per cent and a United Russia share of 51–53 per cent; FOM projected 54 per cent turnout and 47–49 per cent for the ruling party. The entire gap between the loyal pollsters' forecast and the CEC's final figure was closed in three days of voting.
Forty-six veterans of the war against Ukraine were elected, 41 of them on the United Russia ticket. For the first time, voting was held in the occupied parts of Ukraine's Donetsk, Luhansk, Zaporizhzhia and Kherson regions; the CEC opened 482 polling stations for troops in the combat zone. Among those elected from the occupied territories is Alexander Borodai, former head of government of the self-proclaimed Donetsk People's Republic.
18.4 million: what the protocols show
Shpilkin's method compares the distribution of votes for the leading party with the distribution for all other parties across polling-station turnout. At honest polling stations the two curves have the same shape; the surplus for the leader where turnout is anomalously high is the estimate of fraud. Applying the method to CEC data, IStories found at least 18.4 million anomalous votes for United Russia out of 30 million cast on paper, and put the party's genuine result at no more than 35 per cent. Analyst Andrei Kovrizhnykh arrived at a similar figure: 16–18 million stuffed ballots and genuine support of 36–40 per cent. Boris Ovchinnikov estimated the stuffed share at 31 per cent of all votes and genuine turnout at no more than 42 per cent.
Novaya Gazeta Europe took a different route. On a two-dimensional plot of turnout against result it searched for the core of polling stations showing no sign of interference. That core sits at 34 per cent for United Russia at roughly 38 per cent turnout. Only 18,336 polling stations, 21.3 per cent of the total excluding Moscow and online voting, fall inside this statistically clean core. Polling stations reporting round-number turnout and results alone account for at least 2.8 million drawn votes. Roughly 9 per cent of protocols were entered into the GAS Vybory system faster than the ballots could physically have been counted; those stations delivered about 4 million votes to the ruling party. In North Ossetia 97 per cent of stations reported that fast, in Kabardino-Balkaria 93 per cent. The record belongs to precinct No. 752 in Belgorod region: 100 per cent turnout, 8,870 ballots, 98.6 per cent for United Russia, protocol in the system one hour and fifteen minutes after the polls closed. In Derbent, Dagestan, United Russia "won" at 45 polling stations with results between 93.84 and 95.03 per cent.
The spread between 16 and 18.4 million, and between 34 and 40 per cent, shows that these are estimates of magnitude rather than precise figures. Both Shpilkin's method and the search for an honest core rest on the assumption that, absent interference, the distribution of polling stations is close to normal, and different models yield different numbers. None of these estimates covers the roughly 7 million votes cast online. Analysts did not examine that dataset because it cannot be examined. Online voting, as Meduza puts it, "was already a black box, and has now become even more closed". Since those who chose to vote online are on average more loyal to the authorities, the party's genuine result including them is above 34 per cent; by how much, nobody outside the system knows.
There is direct evidence of rewritten protocols, not just statistics. At Moscow precinct No. 2700, according to analyst Ivan Shukshin, the paper protocol put New People first with 29.5 per cent, the KPRF second with 28.2 and United Russia third with 21.5. A day later, GAS Vybory showed United Russia at 86.8 per cent, the KPRF at 2.5 and New People at 2. Nor was the falsification all in one direction. In Cherepovets, where United Russia polled around 20 per cent and the KPRF nearly 37, the results at 111 of 119 polling stations coincide to within a few percentage points; analysts link this to the Vologda governor's feud with steelmaker Severstal. An administrative machine that can draw results for the authorities can draw them against the authorities too.
Where 59.8 per cent came from: turnout built before the count
By the evening of the second day official turnout had reached 44.38 per cent, more in two days than the 2021 election recorded across all three. After the first day alone the CEC reported 29.19 per cent, against 18.39 per cent at the same point five years earlier.
Meduza's sources close to the presidential administration described the machinery. Regions were set a turnout target of "50 plus", and, in the words of one deputy governor, "the plus is better made fatter". State institutions and corporations appointed "brigadiers" from middle management, each responsible for the turnout of their group. In Yakutsk, teachers reported to their superiors every two hours on their own attendance and that of their pupils' parents. Staff at Moscow's General Planning Institute were asked in advance to collect relatives' names and were required to submit screenshots of their vote. In Belgorod region, according to the rights group OVD-Info, pressure ran through the treasury, the regional justice ministry office, utility companies, factories and universities. The pattern was uniform: an assigned day and hour, and proof in the form of a screenshot, geolocation or polling-station number. OVD-Info and the Elections-2026 monitoring project logged 590 reports of pressure, including 119 of direct coercion to vote and 130 of pressure through an employer or institution. In St Petersburg, observers reported that the attendance of civil servants was tracked with facial-recognition cameras.
The second instrument was online voting. In 2021 remote electronic voting was available in seven regions; in 2026 it was available in 32 regions plus Moscow. In the capital 78 per cent of voters, 3.34 million people, voted electronically; the other regions added 3.64 million. By midday on the first day more than half of those registered for online voting had already cast their ballots. That concentration in the first hours of a working day is consistent with reports of corporate mobilisation and with the deadlines employees were given for submitting proof. On the final day the electronic voter roll failed across Moscow for several hours; because the capital keeps no paper rolls, many polling stations refused voters a paper ballot and directed them to a terminal instead. Dmitry Reut, deputy chair of the Moscow election commission, described a request for a paper ballot as "an attempt at double voting".
What online voting delivers beyond turnout is visible in a comparison by Novaya Gazeta Europe: in regions with electronic voting United Russia took 51 per cent at paper polling stations, in regions without it 63 per cent. Where an unverifiable channel exists, the paper can be kept cleaner. The blockchain data that allowed Novaya Gazeta to identify traces of fraud in Moscow's online vote in 2021 were not published in 2026, and voters can no longer verify their own ballot: the system displays only a numerical identifier with no link to a party.
Why 2026 differs from 2021
Meduza compared the distribution of polling stations across the two parliamentary campaigns using data from 82,124 precincts, about 85 per cent of the total. In 2021 the core of honest polling stations lay at 35–40 per cent turnout and roughly 30 per cent for United Russia; in 2026 it shifted about five points upward on result. The classic "comet tail" of high-turnout stations has become a multi-cluster "body": several separate clouds at different levels instead of a single trail of anomalies. The number of polling stations reporting round-number turnout and results, calculated using the methodology of Dmitry Kobak, Sergei Shpilkin and Maxim Pshenichnikov, nearly doubled, from 1,300 to almost 2,250. St Petersburg, by Meduza's estimate, doubled both its turnout and the ruling party's result in five years. In Chechnya nothing changed: everything there was drawn in 2021 too.
The tools changed as well. For the first time since 2004, no optical ballot scanners were used at polling stations. In the 2011 election, by the calculation of mathematician Alexander Shen, United Russia polled 34 per cent at 45 per cent turnout where scanners were in use, and 55 per cent at 61 per cent turnout where they were not. Since 2020 voting has run for three days, leaving ballots unattended by observers for two nights. Novaya Gazeta Europe also identified a new pattern in 2026, a "cloud above the cloud": within a single region and at identical turnout, one group of polling stations yields 35 per cent for the ruling party and another 60 per cent. That is the signature of votes being shifted between parties within a completed protocol, not of ballot stuffing. In Crimea and the Moscow region each territorial commission draws its own group of stations at its own level, a "patchwork quilt" with no single tail.
The comparison of the two datasets points to a change in the structure of the anomalies, not merely their scale. Analysts at Novaya Gazeta Europe and Meduza read it the same way: manipulation has become multi-layered, distributed across territorial commissions and aimed no longer solely at the ruling party's percentage, while online voting has removed a large share of ballots from any form of scrutiny.
How A Just Russia was carried over the threshold
On the morning of 21 September, with 94.95 per cent of protocols processed, A Just Russia stood at 4.96 per cent. At 96.95 per cent it stood at exactly 5.00. Stanislav Andreychuk, co-chair of the Golos election monitoring movement, noted that the party gained almost 70,000 votes in an hour, which is "almost mathematically impossible". By Roman Udot's calculation, the party polled around 12 per cent in the final tranche of protocols against 4.9 per cent previously, and those protocols were half again as large as the average. United Russia's share in the same tranche came in at 65 per cent against 57 before.
Novaya Gazeta Europe traced the source of those votes: late-entered groups of polling stations in eight regions where the party's result leapt from 3–5 per cent to 15–40 per cent. In all, 534 stations in 43 territorial commissions and roughly 94,000 drawn votes, most of them in the Moscow region (about 35,000) and St Petersburg (about 25,000). At stations without anomalies the party polled about 7 per cent, but ballot stuffing for United Russia pushed its share below the threshold, and manual correction became necessary. In Pavlovsky Posad, United Russia's votes were split almost evenly: 4,074 went to A Just Russia and 4,073 to the LDPR, so that no single party's sudden jump would stand out. Ovchinnikov describes the mechanism plainly: the votes were first stolen from the party in United Russia's favour, then partly written back.
The story of A Just Russia raises a question wider than one party's result: in 2026 the administrative system determined not only United Russia's percentage but the roster of parties that cleared the threshold. A Just Russia's return cut United Russia's seat count from a projected 355 to 349, but preserved a five-party Duma in which every faction supports the war.
Yabloko's removal: how the court stretched the law
Yabloko, registered with a federal list and holding the second line on the approved ballot, was the only party calling for a ceasefire. On 7 August, Rodina filed a suit to cancel the registration of its list. On 10 August the Supreme Court granted it, and on 17 August the appeals panel upheld the ruling; outside the courthouse that day, according to Meduza, more than 85 people were detained. The Supreme Court declined to hear a supervisory appeal, and the party has turned to the Constitutional Court. According to Meduza's sources close to the presidential administration, the party was registered and then removed once its ratings began to climb, in order to "contain the risks"; according to journalist Mikhail Zygar's sources, the decision was taken personally by Vladimir Putin. Another recognisable anti-war politician, Boris Nadezhdin, had already been forced out of the campaign, as FACTUM.PRESS reported earlier.
The plaintiff brought a sprawling set of claims: copyright infringement (lines from Soviet songs, AI-generated images, a 1945 photograph of Hiroshima, a logo resembling an El Lissitzky poster), unlawful campaigning, and support for extremism, by which Rodina's representative meant calls for a truce. The CEC and the prosecutor's office backed the removal primarily on the copyright ground.
The plaintiff's strongest ground was financial. According to Rosfinmonitoring data read out in court, 74 people were found among the party's donors and counterparties who between April 2023 and August 2026 had received around 80 million roubles from 310 foreign persons; the court held that the party had failed to prove it had returned these donations. The court also relied on a memorandum from the Association of Bloggers and Agencies classifying publications in support of Yabloko as party expenditure exceeding 35 million roubles. Taken whole, the construction runs as follows: the party received money of indirectly foreign origin and spent more on its campaign than passed through its electoral fund.
This construction departs from the letter of the law. The list of prohibited donors in Article 58 of the Law on Basic Guarantees of Electoral Rights and Article 73 of the Law on Elections of State Duma Deputies is built on the donor's status: foreign states, organisations and citizens, stateless persons, foreign agents, Russian legal entities with foreign ownership above 30 per cent, anonymous donors. A Russian citizen who has at some point received transfers from abroad is not on that list, and nothing in the published materials suggests that any of the 74 held a prohibited status. The court shifted the test from the donor's status to the origin of the donor's income, then placed on the party the burden of proving it had returned money the law does not require it to return. The bloggers' association memorandum treats publications by media outlets and private individuals, which the party neither commissioned nor paid for, as party spending, although electoral law defines campaign expenditure as electoral fund resources and campaigning paid for by the party. On that reading, as Yabloko's leader Nikolai Rybakov observed, any participant can be removed from any election. Formally, the Supreme Court found a violation. In substance, it interpreted the law expansively and created a practice under which an opposition party's access to the ballot depends on how a court chooses to read the origin of its donors' money.
The party's constituency candidates were thinned out too: by The Insider's count, 98 of 135 made it to polling day. None won a seat, although in several constituencies they officially polled 10–15 per cent. In St Petersburg, Yabloko candidates received zero votes at several polling stations where voters said they had voted for them. Yabloko did win seats in the Novgorod regional legislature.
Observers: barred, detained, beaten
On the first day alone, by the count of the regional outlet 7x7, at least 1,334 observers from the KPRF and Yabloko were refused entry to polling stations. The largest numbers were in Voronezh region (481 KPRF observers), Leningrad region (279), St Petersburg (250 Yabloko observers and 17 from the KPRF) and Tyumen region (more than 200 KPRF representatives whose lists, the party says, never appeared in GAS Vybory). Territorial commissions closed during working hours and rejected documents on formal pretexts, down to lists being delivered on a disc rather than a flash drive. In Chuvashia, according to OVD-Info, observers were offered money to stay away.
The Elections 2026: Observing Together project collected 1,064 reports from 65 regions: 379 concerning pressure on observers and commission members, 169 on the use of administrative resources, 121 on coercion to vote, 114 on threats from security forces and 55 on ballot stuffing and other direct falsification. OVD-Info documented 18 detentions during the voting days: three candidates, six observers and a territorial commission member; two people were beaten while being detained. At Moscow precinct No. 1919, according to OVD-Info, a police officer threw an observer to the floor and pinned him with his foot. The campaign manager for Yabloko candidate Shatunovskaya-Byurno said she was beaten at a polling station. Observers recorded rewritten protocols, safe-bags opened after overnight storage and precinct results vanishing from the system.
Failures, delays and disappearing data
This was the first election run in full on GAS Vybory 2.0, the state's new vote-tabulation system. Days before polling, IStories published an investigation based on a leak of almost 200 GB of documents from the system's developer, Tsifrotekh. Experts consulted by the outlet found that the system has no protection against protocol substitution, allows protocol data to be edited manually and uploads online-voting protocols without authenticity checks. Weeks before the election the developers were still fixing dozens of bugs, including in the generation of final protocols; the project's cost, by the journalists' calculation, exceeded 20 billion roubles. In August CEC chair Ella Pamfilova had insisted that GAS Vybory could not be hacked.
Publication of results was uneven. Protocols from Bashkortostan appeared only on the morning of 23 September, and Leningrad region's were still missing that afternoon. After cases of rewriting were publicised, journalists at RusNews found that precinct-level results in GAS Vybory would no longer open: in place of figures the system returns the message "No data to build report". In Kirov region, two registered candidates for the regional legislature were left off the ballot in one constituency; the vote was declared invalid and rescheduled for September 2027.
Abroad, according to Meduza, United Russia lost in 40 of 152 countries, and in 39 of those New People came first. Independent exit polls cited by IStories gave New People 43 per cent abroad and United Russia 5.5 per cent; the official figures were 22.42 and 45.49. Pamfilova claimed 58 per cent for the ruling party among the 280,000 who voted abroad; CEC data show that result was delivered by polling stations in Abkhazia and Transnistria.
What the CEC says
Pamfilova called the campaign the most difficult in modern Russian history. According to the CEC, the election infrastructure withstood more than a thousand waves of DDoS attacks originating in Britain, Ukraine and Brazil and more than 28 million potentially hostile actions; none, the CEC chair said, succeeded. Pamfilova also said a representative of one of the "systemic" parties had tried to plant malware in Moscow's online voting system, naming neither the party nor the person. The Moscow election commission attributed the failure of the electronic voter roll to hacker attacks. None of these claims has been independently verified; the CEC has published no technical reports.
By the CEC's account, confirmed violations did not exceed 0.01 per cent across some 91,000 precinct commissions. Pamfilova dismissed reports of mass exclusion of observers as fake: complaints on the subject, she said, numbered 23 on the first day and 3 on the last, and courts removed 15 of 184,106 party observers from polling stations. The pro-government Independent Public Monitoring group published a map of more than 351,000 messages from its own observers, seven of which were flagged as violations. On 23 September Putin linked the outcome to the war and addressed the troops: "this vote is for you".
After the vote
One of the campaign's most prominent public critics, political scientist Alexander Kynev, designated a "foreign agent" in February, was detained in Moscow on drug charges; according to Mediazona, the Perovsky district court remanded him in custody, and the charge carries 10 to 20 years. Kynev had called the campaign "the most unfair and over-regulated" in the history of Duma elections and put genuine turnout at 40 per cent and United Russia's genuine result at no more than 35. No one has officially asserted a link between those assessments and the criminal case; readers may draw up the chronology for themselves.
In Kazan, Perm, Izhevsk, Syktyvkar, Yekaterinburg, Krasnodar, Tyumen, Ulan-Ude, St Petersburg and Moscow, activists from across the political spectrum, from the KPRF to Yabloko, have filed notice of rallies against fraud in the first ten days of October. In Kazan, Perm and Izhevsk the authorities have already refused; in Perm and Izhevsk organisers are preparing lawsuits to annul the results.
On 21 September the European Union declared that it would never recognise the vote in Ukraine's occupied territories or its results, and announced sanctions against organisers and candidates. The statement noted separately that Russia had not invited OSCE observers.
The first sitting of the ninth Duma is scheduled for 30 September, ten days after the vote. That is the shortest interval between an election and the start of the lower house's work in modern Russian history: in 1994, 1996 and 2000 it was 30 days, in 2003, 2007 and 2021 22–23 days, in 2011 and 2016 17 days. According to Novaya Gazeta Europe, 15 days were originally planned before the schedule was cut by a further five. A Duma assembled in three days of voting is needed at work immediately.
Conclusions
What is established. United Russia officially won 349 seats and a constitutional majority on a turnout of 59.80 per cent, which is not a historical record and exceeded the state pollsters' forecasts. The only anti-war party was removed from the federal ballot by a Supreme Court ruling on a suit brought by a rival party whose leader himself won a seat. At least 1,334 observers were refused entry to polling stations on the first day; detentions and beatings of observers are documented. A Just Russia cleared the threshold in the final percentage points of the count, gaining almost 70,000 votes in an hour. Discrepancies between paper protocols and GAS Vybory data, a multi-hour failure of paper voting in Moscow, delayed publication for entire regions and the subsequent closure of precinct-level data are all documented. The number of polling stations reporting round-number turnout and results nearly doubled compared with 2021. Some 7 million electronic votes are closed to any independent scrutiny.
What remains an estimate. The precise scale of fraud: 16–18.4 million anomalous paper ballots, a genuine United Russia result of 34–40 per cent and genuine turnout of around 40–42 per cent are statistical estimates that cannot be verified without the original protocols and the online-voting data. It has not been documented that A Just Russia's passage over the threshold was a centralised decision, or that the failure of Moscow's electronic voter roll was deliberate. How much of the CEC's account of external attacks would survive scrutiny is unknown.
The official result is known to the hundredth of a percentage point. What is not known is how far those hundredths reflect the votes of voters. Independent analyses using different methods converge not on a precise figure but on the scale of the discrepancy: tens of percentage points on the ruling party's result, millions of statistically anomalous ballots, turnout some fifteen points below the announced figure. Nearly seven million electronic votes, meanwhile, remain beyond any form of verification. The principal outcome of the 2026 election is not only United Russia's 349 seats but the impossibility of independently reproducing the result on which those seats were allocated.
A Duma assembled on that basis will convene on 30 September without a single faction opposed to the war. The decisions that were postponed until after the election, from tax rises to a possible new wave of mobilisation, which FACTUM.PRESS has examined in earlier coverage, it will now take against a backdrop of strikes on infrastructure and fuel shortages, and with no further need to look over its shoulder at the voter.
Sources: Central Election Commission of Russia (final results of 25 September, statements by Ella Pamfilova); Supreme Court rulings on Rodina's suit; IStories (Shpilkin-method calculation, investigation into GAS Vybory 2.0); Novaya Gazeta Europe (anomaly analysis, interview with Roman Udot, data on convening dates); Meduza (comparison of 2021 and 2026 distributions, sources in the presidential administration); The Bell; OVD-Info; The Insider; 7x7; DOXA; Mediazona; ASTRA; Radio Liberty; Electoral Atlas; the Elections 2026: Observing Together project; the Yabloko party; Vedomosti; Kommersant; Interfax; RBC; VCIOM and FOM forecasts; statement by EU High Representative Kaja Kallas. Statistical estimates of fraud were produced by different methods and are given as ranges; CEC claims regarding cyberattacks and violations, accounts from anonymous sources and some observer reports have not been independently verified. This article will be updated as complete data are published.