2026 election misinformation trends: From “will it happen?” to “was it stolen?”
Bangladesh held its much-anticipated 13th National Parliament election on 12 February 2026, an event widely expected to be heated, polarising and closely contested. In the months leading up to the vote, there were growing concerns that misinformation would play a larger role than in previous elections, shaping public perception and potentially influencing electoral behavior.
Between 11 December 2025 – when the schedule was announced – and 20 February, one week after polling, nine fact-checking organisations documented and debunked 1,185 unique pieces of political misinformation. Nearly half were directly related to the election and around 12% of those were generated or manipulated using artificial intelligence.
For this analysis, cases were grouped according to the election calendar: pre-election, nomination, campaign, 48-hour silence, Election Day and immediate post-election phases. The classification follows formal procedural milestones – from candidate submission to results – allowing trends to be measured against moments when voter attention intensifies and regulatory conditions shift.
Half of this factchecked misinformation was packaged to resemble journalism itself – with graphic cards mimicking newsroom layouts and videos styled as official announcements or news clips.
The findings show that while the formal campaign period accounted for the largest share of election-related misinformation (41% of the total) the most concentrated surge occurred during the legally mandated silence period. In the final 48 hours before polling, narratives circulated alleging election cancellation, voter suppression and fabricated candidate withdrawals. Nearly one in five false or misleading claims were detected in those last two days alone, underscoring how misinformation intensified at the most fragile moment of the electoral cycle.
This analysis also draws on 528 election-specific cases, tracing how narratives shifted as Bangladesh moved from schedule announcement to final results. In the early weeks, rumours focused on whether the election would take place at all, with claims of postponement, cancellation or conditional participation. As the campaign advanced, fabricated surveys and altered statements shaped expectations about who was leading. By polling day, the emphasis turned to alleged irregularities – exit polls, ballot seizures and candidacy withdrawals. In the aftermath, attention moved to margins and manipulation, recasting the debate from whether the vote would happen to whether it had been won legitimately.
Trends in Data
Political falsehoods have historically dominated Bangladesh’s misinformation space. During the 2026 election cycle, that pattern intensified, and sharpened as polling approached.
Across December and early January, fact-checked political claims fluctuated at relatively moderate levels, typically ranging between 10 and 12 cases per day, and consistently outnumbered those explicitly tied to voting procedures or candidacy, which ranged from just two to three. But from late January onwards, the trajectory shifted. Both categories rose in parallel, with election-specific misinformation accelerating more steeply in the final stretch.
The sharpest surge came during the 48-hour silence period. In just two days, 95 election-related false claims were documented – accounting for nearly 18% of the entire election-related dataset. On 11 February alone, more than 70 combined political and election-specific claims were fact-checked, marking the single highest daily spike of the cycle. Compared with the early-January baseline, misinformation volumes increased roughly six to seven-fold.
Of the 528 documented election-specific cases recorded between 11 December and 20 February, the formal campaign period accounted for the largest overall share (41%). But volume alone does not tell the full story. When adjusted for duration, the sharpest concentration occurred during the legally mandated 48-hour silence period.
In just two days, 95 cases were detected – nearly 48 per day – making it the most compressed spike of the election cycle. By comparison, the 19-day campaign period averaged about 11 to 12 cases per day, reflecting sustained pressure rather than a sudden surge. The nomination phase averaged just under five cases per day, while Election Day recorded 26 cases within a single day. In the week after polling, misinformation declined but persisted at around five to six cases per day. The pre-election phase, though longer, remained relatively subdued, with the lowest daily concentration across the cycle.
The data also points to a shift in emphasis over time. In the pre-election and nomination phases, fact-checkers recorded an average of around 13 combined political and election-related cases per day. Of those, fewer than five per day were explicitly tied to voting procedures or candidacy. Political misinformation dominated, but much of it was not yet centred on the mechanics of the election itself.
As polling drew closer, that balance began to tilt. During the campaign period, combined political and election-related claims rose to an average of 21 per day, while election-specific cases alone climbed to 11 or 12 daily – more than double the rate seen during nominations. By the 48-hour silence period, 88% of the political claims fact-checked were directly tied to the election.
The Game of Tactics and Formats
On average, fact-checkers detected just over two AI-generated misinformation cases per day across the election cycle. Of these, roughly one per day was directly tied to the election itself. While AI-linked content did not dominate overall volume, its presence was steady and persistent throughout the period.
AI-generated misinformation accounted for roughly 12% of election-related cases, and nearly seven in ten of those were video-based. About 24% were images, indicating manipulated photos or synthetic visuals, but far less dominant than video.
Political and election combined, of the 150 AI-generated misinformation documented, nearly one-third (31%) occurred during the formal campaign period, making it the phase with the highest share. But it was not limited to the campaign alone. More than 44% of AI-linked cases were recorded before the campaign officially began – 23% in the pre-election phase and 21% during nominations.
Although the 48-hour silence period recorded the highest overall misinformation spike, only 11 documented during that window. If we look at formats, the silence-period spike appears to have been more text-heavy, including short claims, procedural rumours and quickly circulated allegations about cancellations, withdrawals or voting disruptions – an easier and faster way to produce and distribute than synthetic visuals or manipulated media.
In terms of tactics, most misinformation sought to look like journalism or look like evidence. Graphic cards mimicked newsroom layouts. Videos mimicked on-the-ground reporting or official announcements. Even AI-generated content was frequently styled to resemble news broadcasts or security footage. More than half of the documented cases (54.5%) appeared as graphic cards designed to resemble legitimate news media outputs, allowing false claims to circulate with the appearance of institutional credibility.
Video accounted for another 30.3% of cases, making it the second most dominant format. These included AI-generated clips, repurposed old footage and miscaptioned security drills presented as unfolding election events. Static images made up 9.3% of cases, while pure text posts accounted for just 2.1%, indicating that standalone written claims played a limited role in comparison to visually packaged content.
Narrative analysis
Pre election phase: The pre-election phase was marked less by procedural rumours and more by broad political messaging. Misinformation during this period focused largely on party reputations, leadership credibility and potential postponing or cancelling the elections.
For example, after the announcement of the election schedule on December 11, an AI-made image showed a torch procession by the Shariatpur District Awami League and its affiliated organizations demanding the cancellation of the schedule.
Narratives linking the assassination of Inqilab Moncho’s spokesperson, Sharif Osman Bin Hadi, to the election also circulated. By altering news media photocards, claims were spread that the Jamaat Ameer did not want a national election before the arrest and trial of Osman Hadi’s attackers.
Several false surveys often claiming the win of Either Bangladesh Jamaat-e-Islami, or that 65% of the people will not vote without the Awami League emerged in this phase. Candidacy appeared to be a popular theme as old videos were circulating claiming to be releasing a list of Awami League candidates for next election or individuals receiving nomination from political parties.
Fake attribution also emerged during this phase, with posts on social media alleging that Bangladesh Jamaat-e-Islami Secretary General Mia Golam Parwar had declared that if Jamaat came to power, it would introduce Pakistan’s legal system and make travel to and from Pakistan free. The claim was later found to be false.
Nomination phase: As the nomination window opened, misinformation shifted from reputational positioning to the mechanics of participation. Claims began to centre on who would contest, who had been disqualified, and whether elections would proceed at all. Posts circulated alleging that BNP would boycott the polls without Awami League participation, that nominations of senior leaders – such as, Jamaat-e-Islami’s Ameer Dr. Shafiqur Rahman, and BNP Chairman Tarique Rahman (now prime minister) – have been cancelled.
Seat sharing among parties was at the center of political rumors. Various claims suggested that Jamaat demanded a specific number of seats from BNP, that certain candidates had switched parties, or that certain alliances were breaking apart. The narrative focus also widened to include referendum rhetoric and religious framing tied to ballot symbols. Posts claimed that voting for a particular symbol would lead to spiritual reward, or that the constitution would change if a “yes” vote prevailed.
AI-generated videos and images appeared more frequently, including synthetic clips about election participation and party endorsements. One AI-made fake video of Chief Adviser Dr. Muhammad Yunus circulated widely, claiming that he deferred the election. Fake statements of senior officials, including Election Commissioners, Army Chief and foreign diplomats appeared to reinforce doubts about the legitimacy of the process. These claims centred on election postponement, cancellation, foreign backing of specific parties, or the withdrawal of international observers.
Campaign: During the formal campaign period, misinformation intensified and became more directly tied to the conduct and outcome of the election. Posts attributed exaggerated or fabricated promises to candidates, including claims that Tarique Rahman would close the metro rail and introduce digital local buses if elected, or that free data packages would be distributed during the campaign. Other posts framed voting choices in religious terms, alleging that voting for a particular symbol would guarantee paradise or that failure to support a party would threaten Islam’s constitutional status.
Manipulated campaign imagery also circulated widely. AI-generated rally photos suggested inflated turnout for Jamaat or BNP events, while old footage from 2023 and earlier protests was repurposed as current clashes between rival supporters. Multiple AI-generated images and videos (1, 2, 3, 4) were also circulated showing larger crowds in election rallies.
Synthetic videos portrayed army officers or police officials making statements about election participation or alleged plans for violence. At the same time, posts claimed that foreign actors — including the US ambassador or Indian officials — had taken positions on the election, and fake survey campaigns projected sweeping victories, including assertions that an alliance would win 240 to 250 seats or dominate expatriate ballots. A cluster of posts centred on postal and expatriate ballots, falsely claiming overwhelming overseas margins for Jamaat, premature counting and even cancellation of votes — narratives that sought to shape expectations about the outcome before ballots were cast.
Another widely discussed claim during this period involved NCP leader and Dhaka-11 parliamentary candidate Md. Nahid Islam, with reports circulating in several media outlets that a writ petition had been filed in the High Court seeking cancellation of his candidacy on the allegation that he had acquired citizenship of the Caribbean nation Dominica on April 20, 2025; however, verification report noted that the passport copy attached to the petition was sourced from social media and had previously been proven false through fact-checking.
Religious and identity-based framing intensified, with posts warning of violence, moral decline or constitutional change depending on the vote. For example, a forged media card cited the Jamaat chief, alleging that he said Hindus must vote for the Islamist party or else they will be sent to India. Another fake graphic claimed that a leader of BNP’s student wing threatened that Hindus would be killed inside their home if they did not vote for the party.
Campaign silence phase: The final two days before vote marked the most compressed surge of misinformation in the election cycle. Unlike earlier phases, narratives concentrated on immediate disruption and fearmongering. Headlines repeatedly framed alleged incidents: clashes between party supporters, arrests of candidates with money or weapons, recovery of ballot papers, attacks on polling stations, or intervention by joint forces. In many cases, old footage – joint force exercises, past bus arson, previous protests or earlier polling-station incidents – was repurposed and presented as unfolding election violence. The format suggested immediacy: something was happening now, and it was linked directly to the vote.
A second theme focused on enforcement and procedural integrity that ballot papers had been seized from party workers’ homes, that advance votes had already been cast, or that expatriate ballots had determined results before polling began. Posts alleged detention of candidates, recovery of weapons, or discovery of irregularities at polling stations. These narratives did not argue ideology; they implied administrative breakdown or manipulation in real time.
Fragmented and recycled statements were also prominent. Years-old boycott comments, edited speech clips and miscaptioned videos were circulated as fresh declarations. Rumours suggested candidates had withdrawn, alliances had fractured, or leaders had issued last-minute reversals.
Election Day: Election day misinformation centered on claims of polling station takeovers, ballot box snatching, fabricated exit poll results in various constituencies and false announcements of candidacy withdrawals. The objective appeared to be to cast doubt on whether the election had been free and fair.
Multiple videos of election security drills were circulated with claims that polling stations were being seized or ballot boxes were being snatched. For example, a video shared on Facebook falsely claimed that a man fleeing with a ballot box from a polling station in Rajshahi had been apprehended by the army. It was later found to be old footage of a joint forces drill conducted to ensure election security. Similarly, a video of ballot stamping from the 2024 election was recirculated with claims that it depicted current irregularities.
AI-generated videos were also used to fabricate withdrawal announcements by at least two candidates: Ataur Rahman Khan Angur, an independent candidate in Narayanganj-2, and Samiran Dewan, an independent candidate in Khagrachhari. The manipulated videos claimed that they were withdrawing from the race and urging voters to support rival candidates.
Another widely shared claim alleged that the brother of a candidate had been killed in election violence. Posts asserted that the brother of NCP leader Abdul Hannan Masud, the 11-party alliance candidate in Noakhali-6 (Hatiya), had died in an attack by BNP supporters. The claim was later found to be false.
Several fabricated photocards bearing the names of different news outlets claimed to show exit-poll results for various parliamentary constituencies. In each case, the results depicted the Jamaat-e-Islami candidate as leading and the BNP candidate as trailing. In one instance, it was claimed that in the Thakurgaon-1 constituency exit poll, 33 percent of votes had gone to the BNP and 61 percent to Jamaat. In another, it was claimed that in the Dinajpur-1 constituency exit poll, 34 percent of votes had gone to the BNP and 60 percent to Jamaat.
Post-election Phase: The post-election phase was marked by attempts to reinterpret the outcome. Misinformation during this period focused largely on vote legitimacy, alleged manipulation and retaliatory violence. The narrative environment shifted from contestation over candidacy to contestation over credibility.
For example, several posts shared on social media claimed that, without rigging or conspiracy, Jamaat alliance candidates would have won in many more constituencies. For example, one post stated: “The number of seats where Jamaat lost by fewer than 5,000 votes is 53. In fact, Jamaat was defeated through rigging in these 53 seats. Jamaat actually won 135 seats. But the deep state is trying to reduce the number to 70–80.”
Later, an analysis by Dismislab found that victory or defeat was determined by a margin of fewer than 5,000 votes in a total of 22 constituencies, not 53. Of these heavily contested 22 seats, 11 went to Jamaat and its allies, and nine were taken by BNP and its allies. Posts also alleged nationwide recounts in 150 constituencies, or claimed that official message sheets showed winners being declared despite receiving fewer votes. Both narratives were false.
Violence-related claims also intensified. Footage from earlier clashes between BNP and Jamaat supporters during the campaign was recirculated as evidence of post-election reprisals. In some cases, the party forming the government was accused of attacking citizens for voting freely. In others, Islamists were alleged to have attacked Hindu communities following Jamaat’s defeat. These narratives were found to be false.
Fabricated statements attributed to the United Nations or the United States suggested that the election had not been free or fair; even cabinet formation became a target of misinformation (1, 2), with false reports assigning ministerial portfolios to specific political figures.
Methodology
This analysis is based on fact-check reports published by nine Bangladesh-focused verification organisations: Rumor Scanner, BoomBD, Newschecker, Fact Crescendo, FactWatch, AFP Bangladesh, Ajker Patrika, Dismislab, and The Dissent.
Between 11 December 2025, and 20 February 2026, these organisations debunked at least 1185 political misinformation cases. Within this pool, 528 cases were explicitly election-related and form the basis of this report.
A claim was classified as election-related if it directly referred to the 2026 National Parliament Election and included content about nominations, candidate eligibility, seat-sharing, campaign activities, voting procedures, postal ballots, exit polls, result announcements, and post-election disputes.
For temporal analysis, cases were mapped against the election timeline: pre-election (11-29 December), nomination (30 December-21 January), campaign (22 January-10 February), the 48-hour silence period (10 February 7:30am – 12 February 7:30am), Election Day (12 February), and the immediate post-election period (13-20 February).
This report was produced with the financial support of the European Union within the framework of the AHEAD Bangladesh project. Its contents are the sole responsibility of Digitally Right and do not necessarily reflect the views of the European Union.
