2025 in review: The rise of AI in misinformation, familiar faces in disinformation headlines, and unfamiliar actors behind the scene
By the numbers alone, 2025 can be described as a year of misinformation, with fact-checkers identifying 30 percent more false information in Bangladesh than in the previous year. It was also a year of evolution in the nature of how disinformation is spread. Two out of every three fact-checks involved political content last year, half of which were spread through videos and one in every ten using artificial intelligence (AI), signaling an increasing use of technology to circulate false information.
In 2025, nine fact-checking organizations focusing on Bangladesh published at least 5,706 fact-check reports. Of them, 4,131 were unique pieces of misinformation, compared with 3,167 in 2024. Dismislab’s analysis of these contents show 10 individuals appeared most frequently in disinformation headlines. By examining more than 22,000 posts used as evidence in these fact-check reports, Dismislab researchers also identified the accounts that often spread disinformation on social media.
Throughout 2025, waves of disinformation took to social media closely mirroring the major events that dominated public discourse—politics, elections, disasters, violence, religious tensions, and international conflicts. Nearly 58 percent of all misinformation fact-checked last year involved politics, followed by law and order and religion. Periods of crisis saw sharp spikes in false claims. In July, for example, a significant share of misinformation spread around the military jet crash in Dhaka. In November, disaster-related falsehoods surged during an earthquake. In the international sphere, the India–Pakistan and Iran–Israel conflicts became the most significant drivers of misinformation.
Video emerged as the most widely used medium for spreading disinformation, with content generated using AI tools playing a major role. Across nearly every issue, AI-created fake images and videos were circulated. At the end of the year, however, there was a marked increase in impersonating media outlet photocards to spread fabricated quotations attributed to political leaders, activists, government officials, or members of law enforcement agencies.
This analysis next explores who appeared most frequently in disinformation-related headlines in 2025; which platforms hosted the largest share of identified false content and who spread it most actively; and how videos, photocards, and AI made the spread of disinformation faster and easier.
Who Dominated Disinformation Headlines
Dismislab analyzed the headlines of more than 4,000 distinct fact-check reports published in 2025 and compiled a top-ten list of individuals who appeared most frequently in disinformation headlines. Ousted prime minister Sheikh Hasina topped the list, with her name appearing in 274 fact-check headlines. A large share of this misinformation involved her old videos, photographs, and statements that were spread as “recent”. Another major strand centered on rumors of Sheikh Hasina’s return to power. Edited videos, clips of rallies and slogans, and even promotional contents were used to make that claim. False information claiming international “support” for her as the “legitimate prime minister” also appeared regularly. Throughout the year, rumors of Sheikh Hasina’s arrest or death circulated as well, alongside fabricated statements or so-called “confessions” created using AI.
Next on the list is interim government Chief Adviser Dr. Muhammad Yunus, who appeared in 150 fact-check headlines. Most of these false claims involved portraying his government as unstable and illegitimate. Repeated false claims also circulated stating that “Yunus has resigned” or would resign, and that the army had issued him a “deadline” for resignation. To lend credibility to these rumors, edited photocards and fabricated quotations attributed to various media outlets were widely used. Alongside this, false information invoking the United States, China, Europe, international media, or foreign leaders was circulated, claiming sanctions, revocation of citizenship, or disruptions in diplomatic relations. For instance, one post claimed that U.S. President Donald Trump called Dr. Yunus a “dictator.” Beyond politics, AI-generated fake statements attributed to Yunus were used to promote fraudulent schemes and sell consumer products.
A large volume of false information circulated about Sharif Osman Bin Hadi, the spokesperson of Inqilab Moncho and a parliamentary election candidate who was shot dead in December 2025 in Dhaka. His name appeared in nearly 100 fact-check headlines. Disinformation targeting him began as early as September 2025, initially aimed at discrediting his character and political position. Among these early false claims were fabricated allegations of intimate scenes involving Osman Hadi and a woman, as well as accusations that he had subjected a child to “bad touch.” After Hadi’s death, however, the scale of misinformation involving him intensified.
Rumors were amplified by manufacturing artificial “evidence” surrounding the attack, the police investigation, and arrests of suspects to make the false narratives appear credible.
Dramatic claims around his shooting, hospitalization, physical condition and even death were also circulated. Parallel to this, blame-focused campaigns also emerged, seeking to link Hadi’s killing to intelligence agencies of various countries, the Bangladesh Nationalist Party (BNP) and Bangladesh Jamaat-e-Islami, or specific political figures.
Fourth on the list is BNP Chairman Tarique Rahman. An analysis of 72 headlines shows that misinformation surrounding him alternated between praise and misleading claims. Among these were false assertions that footballer Lionel Messi met Tarique or that he was set to attend the inauguration ceremony of Donald Trump. Fabricated claims also circulated, including alleged statements by him in support of extortion or declarations that he had announced plans to make Dr. Yunus president. The dominant trend, however, centered on his return to the country. Using edited videos, old clips, and AI-generated images, posts variously claimed that he would return soon or that he had been spotted in different locations, including at London’s Heathrow Airport. In addition, edited photocards of different media outlets were used to lend credibility to false statements attributed to him about elections.
National Citizen Party (NCP) leader Hasnat Abdullah appeared in 58 headlines. A large share of the disinformation surrounding him consisted of sensational claims, such as his arrest, being attacked, seriously injured, or even killed, often using old or unrelated videos and images presented as evidence. Edited photocards bearing the names of media outlets, fake press releases, and fabricated quotations were also circulated in attempts to link him to political conspiracies, intra-party deals, or incidents of violence.
Others who featured in the top 10 disinformation headlines are U.S. President Donald Trump, Bangladesh Army Chief General Waker-Uz-Zaman, BNP Secretary General Mirza Fakhrul Islam Alamgir, Indian Prime Minister Narendra Modi, and BNP’s late chairperson and former prime minister Khaleda Zia.
Key Amplifiers of Disinformation
False or misleading information is spread by different actors for different reasons, some do it for political purposes, others for commercial gains. Also, not every post gains equal prominence in fact-check reports. Typically, only a handful of posts are cited as evidence in each verification report. Dismislab documented and analyzed 22,129 such posts from the fact-check reports published by nine organizations. Researchers then identified more than 10,000 accounts across Facebook, TikTok, X, and YouTube, which were subjected to verification. Facebook accounted for the largest share in this campaign.
Fact-checkers verified 114 pieces of false information circulated from a Facebook profile named “Gurudaspur Upazila Juboleague.” Posts from this account regularly promoted the Awami League party while portraying the BNP, Jamaat, and the interim government in a negative light. Although the account name itself signals a party affiliation (Jubo League is Awami League’s youth wing), the URL lists the name “Sadia Sultana,” a common female name in Bangladesh. Second on the list is an account titled “Amora Harbona,” whose profile features photographs of Sheikh Hasina and her son Sajeeb Wazed Joy. In the profile description, the account holder claims to work in “political sciences.” Its daily posts also consistently supported the Awami League. Over the last year, 81 posts from this account were flagged in fact-check reports. The three others on the top five Facebook account list are Shipon Islam, Pintu Yousuf, and Tamanna Akhter Yesman (which is no longer accessible). These accounts were also aligned with the same political party.
A different pattern emerges on YouTube. While the top accounts on Facebook spread misinformation primarily for political purposes, the top five accounts on YouTube appeared to be commercially motivated. All of them were monetized. These channels circulated misinformation across political lines and presented themselves as news outlets. Bijoy Shorts YT, topped the list, followed in order by Public Insight, Alo Entertainment News, TNB NEWS TV, and News 365.
On X (formerly Twitter), the leading accounts were found to be primarily spreading religious misinformation. Most of these accounts appeared to be Indian or aligned with the Hindu nationalist movement in the country. At the top of the list was a handle named Voice of Bangladeshi Hindus, whose 25 posts were fact-checked. The handle spread religiously-framed false claims, portraying attacks stemming from family disputes as Muslim assaults on Hindus, or presenting clashes between Chittagong University students and local residents as persecution of Buddhists by the army and Jamaat. Midway through the year, the account was suspended by X. Fourth on the list was an account named Pakistan Untold, which is operated from “Delhi, Gujranwala,” according to its profile. Its description states that it presents “Pakistan as seen through the eyes of Hindus.”
On TikTok, the account named Touhidur Zaman Touhid appeared most frequently in fact-check reports. In the profile description, the user identifies himself as an “expatriate Mujib loyalist,” and circulated AI-generated misinformation in support of the Awami League (1, 2, 3).
AI in Disinformation
Of all the distinct instances of misinformation fact-checked in 2025, 10 percent, or 417 cases, were created using AI. In other words, one in every 10 misinformation involved AI-generated content. From politics to disasters, international conflicts, and environmental issues, AI-driven misinformation appeared across nearly all topics. As shown in the figure above, the spread of such false information increased steadily as the election period progressed.
The use of AI in political misinformation was particularly striking. For example, on December 25, multiple AI-generated images circulated claiming that Tarique Rahman returned to the country, showing him at the airport with family members or stepping off a plane and touching Bangladeshi soil. Similarly, while the BNP’s late chairperson Khaleda Zia was hospitalized and undergoing treatment, several images and videos circulated, falsely claiming that she had recovered and was able to walk.
In addition, AI-generated videos were circulated claiming to show Sheikh Hasina delivering a speech in the Indian parliament. Elsewhere, images were shared purporting to show students holding placards reading “We want Hasina” or painting similar slogans as graffiti on walls. In some cases, entirely fabricated video campaigns were run using techniques such as so-called “puppet video messages,” aimed at promoting narratives about Sheikh Hasina’s “return to the country” or conveying political messages attributed to her. Throughout the year, AI-generated images and videos were also used to portray leaders and activists from various political parties in a negative light (1, 2, 3).
Figures from the entertainment industry also fell victim to AI-driven misinformation. For example, a video circulated claiming to show model and activist Farzana Sithi dancing was actually created by manipulating a dance video of an Indian performer using deepfake technology. In another instance, images from an Instagram account operated from India were edited and falsely promoted as belonging to actress Sadia Ayman.
The same pattern was evident in disasters as well. After the earthquake in Bangladesh in November (1, 2), as well as in July when a training aircraft crashed at Milestone School and College in Uttara in the capital, a number of AI-generated images were presented as footage from the actual incidents.
The Surge of Political Misinformation
Throughout the year, misinformation related to politics circulated consistently and was fact-checked in greater volume compared to any other topic. In every quarter, politics ranked first among all other issue-based false information. However, the volume was highest between September and December as elections approached. While an average of around 200 political misinformation items were fact-checked each month, the number more than doubled to 408 in December.
One of the main contributors to the rise in political disinformation from September onward was the student union elections at different public universities across the country. In the following months, false information began circulating about the timing and method of the national parliamentary election and party nominations (1, 2, 3). In mid-November, a lockdown program linked to the verdict in a crimes against humanity case involving the ousted prime minister Sheikh Hasina (1, 2) played a significant role in amplifying political disinformation. In December, the killing of Osman Hadi and Khaleda Zia’s illness and subsequent death were two top drivers of political disinformation.
Following the announcement of the parliamentary election and referendum dates, at least 55 instances of election-related misinformation were identified within just one month, most of them centering on various political alliances and fabricated survey results. At the same time, AI-generated videos were circulated claiming to show vandalism and arson by BNP leaders and activists who were denied nominations, alongside false claims of expulsions or resignations by party leaders. Claims were also spread misidentifying individuals as NCP nominees, further adding to the volume of misleading election-related content.
Beyond politics, large volumes of misinformation circulated around law and order (7 percent of all verified false information), religion (6 percent), international affairs (6 percent), and disasters (4 percent). Overall, there was a persistent tendency to spread fear and anxiety about the country’s law and order. Analysis shows that at least 90 fact-check reports had headlines related to rape, more than 100 focused on murder or killings, and 45 involved cases of abuse or torture. In the religious category, false information about attacks on or persecution of Hindus or other minority communities in Bangladesh circulated throughout the year. The spread of misinformation related to international affairs was most pronounced in May and June, largely centered on the India–Pakistan and Iran–Israel conflicts.
How Disinformation Circulated Online
In 2025, video emerged as the dominant medium for the spread of misinformation. Nearly 52 percent of all false information circulated in video form. Videos from older or unrelated incidents were frequently shared with false claims that they were recent. In addition, AI-generated videos were circulated to support false claims.
Graphic cards accounted for 21 percent of all misinformation fact-checked. Fabricated quotations attributed to leaders and activists of various political parties, government advisers, or law enforcement officials were spread using photocards. Of them, 74 percent were disseminated by impersonating established media outlets. In December, misinformation circulated more frequently through graphic cards than through video. Meanwhile, about 18 percent of verified misinformation spread in the form of images.
Methodology
To analyze misinformation trends and narratives in 2025, reports published on the websites of nine organizations that produce fact-checks on issues related to Bangladesh were taken into account. These sites were Rumor Scanner, BoomBD, NewsChecker, Fact Crescendo, Fact Watch, AFP Bangladesh, Ajker Patrika, Dismislab, and The Dissent. Based on the subject matter, the reports were classified into 17 categories.
Although a total of 5,706 fact-check reports were published on these sites in 2025, multiple outlets often published reports on the same false claim. When multiple websites published fact-check reports on the same piece of misinformation, only one report was counted as unique. Using this method, 4,131 fact-check reports were identified.
In some cases, a single incident involved multiple false claims that were verified within one long-form report. In such instances, despite multiple verifications, the report was counted as a single item. As a result, the total number of fact-checked false information items may be slightly lower than the actual number verified.
