AI-generated videos burst into political campaign amid surge in election misinformation

The use of AI-generated content in the spread of misinformation has steadily been growing in Bangladesh for quite some time, but the second quarter of 2025 has broken all past records in terms of the nature and the medium used to make fake or misleading content. In a break from the past, some of the content now consists of synthetic videos, rather than images, of political campaigns created using Google’s Veo-3 technology. The phenomena has raised concerns over the potential impact of AI-generated content on voters as Bangladesh prepares for national election. Bangladesh’s chief election commissioner himself sees the misuse of AI in the upcoming election as a serious challenge.

This quarter has also seen a significant surge in misinformation centering on global events, from just 1% of all misinformation in Bangladesh in the first quarter to 19% in the second quarter. This spike in misinformation on international events coincides with the terror attack in Kashmir’s Pahalgam in April reigniting tensions between India and Pakistan and the Iran-Israel conflict in June. Both events reverberated in Bangladesh throughout the period.

Overall, spread of misinformation surged by 17% in this quarter compared to the first quarter of 2025. Politics remains the most frequent topic of misinformation in Bangladesh, as was also evident in the second quarter of this year, with 44% of all the unique pieces of misinformation related to politics. The national parliamentary election was one of the key subjects of political misinformation. Some of the top political figures, who were subjects of misinformation regarding Bangladesh’s election and politics, include Chief Advisor Dr. Muhammad Yunus, US President  Donald Trump and Bangladesh Nationalist Party (BNP) leader Salahuddin Ahmed. Various false claims were also spread concerning the alleged involvement of politicians in corruption and criminal activities. Misinformation involving religion saw a slight decline in this quarter.

Dismislab has identified this pattern through an analysis of reports published on eight fact-checking websites focused on Bangladesh in the second quarter of 2025 (April–June). During this period, a total of 1,361 fact-check reports were published across these websites. In cases where multiple reports were published on the same claim, only one was counted as a unique item for analysis. Through this method, 1,013 unique pieces of misinformation were identified—a 17% increase compared to the previous quarter (868).

Evolving nature and medium of AI-generated misinformation

From early 2024 through the first quarter of 2025, AI-generated misinformation was more commonly disseminated through images. However, in the second quarter of 2025, for the first time, the spread of AI-generated misinformation through video surpassed that of images. This shift coincides with the release of Google’s new AI video generation tool, Veo-3, in May this year. Since then, the use of this tool for producing videos has steadily increased. These AI-generated videos are not only being used to spread misinformation (1, 2, 3) but are also being deployed in political and electoral campaigns. In June 2025, an investigation by Dismislab showed how videos of fake interviews of voters created with Veo are emerging as a tool for election campaigning.

A shift is also visible in the themes of AI-generated content, both misinformation and other content likely meant for fun. In 2024, one in every five (20%) AI-generated misinformation outside the political domain focused on environment and nature, with images of flowers created using AI (1, 2, 3) being circulated as real photographs. However, in the first two quarters of 2025, the share of such content came down to just 4%. Meanwhile, misinformation related to international affairs and entertainment registered a significant rise of 31% and 20% respectively.

Most of the AI-generated misinformation about entertainment involved deepfake videos of models, film actors, or popular content creators, where the face of one person was superimposed onto someone else’s body in unrelated footage.

Election and spike in political misinformation

Since the fall of the Awami League government on August 5, 2024, and the assumption of power by the interim government on August 8, discussions over the next parliamentary election have dominated Bangladesh’s political landscape. Chief Advisor Dr. Muhammad Yunus has stated that the election could be held in February or April 2026. Meanwhile, the BNP, one of the country’s major political parties, has been demanding elections by December this year. Amid intense discussions about the timeline, misinformation centering on the election has also spread far and wide this quarter. 

In the first quarter, only seven instances of election-related misinformation were recorded. In the second quarter, that number rose to at least 55, registering an eightfold increase. As BNP has been the most vocal political party about holding elections swiftly, its leaders have become frequent targets of various false claims. 

For example, a fabricated photo card attributed to BNP Standing Committee Member Salahuddin Ahmed falsely quoted him as saying, “BNP will not participate in the election if it is not held by December.” In another instance, he was misquoted as saying that if elections are not held by December, BNP would form an alliance with its rival party, the Awami League, and launch a movement. False statements were also spread in the name of BNP’s acting chairman Tarique Rahman, claiming that BNP would initiate a strong protest if the election is not held within the year. One fabricated claim, circulated through a doctored media photo card, falsely suggested that U.S. President Donald Trump had urged BNP to prepare for the election.

Misinformation also emerged around Dr. Yunus, including false claims that he had already announced the election date. Additionally, a fabricated quote attributed to Barrister Asaduzzaman Fuad, general secretary of the Amar Bangladesh Party, claimed that instead of holding parliamentary elections, Dr. Yunus should remain in power for five years.

False claims also circulated concerning the participation of the ousted Awami League in the next national election. One such piece of misinformation stated that the United Nations declared that no election would be held without the Awami League. Another claimed that Indian Prime Minister Narendra Modi called on the Bangladeshi government to ensure security of Awami League leaders during the election.

Misinformation also continued to circulate globally over the ban on the Awami League’s activities on May 10, falsely claiming international concern over the matter. For instance, a fake photo card styled after BBC Bangla falsely stated that India, Russia, and China expressed reservations over the decision. Another piece of misinformation claimed that Donald Trump issued a 24-hour ultimatum demanding the ban be lifted.

Beyond election-related narratives, false claims have also targeted leaders and activists of various political parties in connection with corruption and criminal activities. Members of the newly formed National Citizen Party (NCP) were falsely accused of hoarding luxury vehicles worth millions of taka, rice allocated under a social safety net program called Vulnerable Group Development (VGD), and large sums of cash. An old video from India was circulated to falsely claim that BNP leaders were involved in extortion.

On the other hand, the misinformation targeting female politicians often aimed at character assassination. Edited explicit or semi-explicit images and videos collected from the internet were falsely attributed to female leaders of the NCP. For example, a doctored video of an Indian model was circulated, falsely claiming to show footage of quota reform movement protestor Nafsin Mehanaz. Similar types of misinformation were seen in the name of NCP leader Arpita Shyama Deb.

AI-generated images and videos have also been used in spreading political misinformation. For example, a deepfake video falsely claimed that Chief Advisor Dr. Muhammad Yunus had resigned. Another deepfake video circulated the false claim that Donald Trump praised Dr. Yunus. One such post was even shared by the interim government’s Law Advisor Asif Nazrul.

Big jump in misinformation on international affairs

On April 22, a terror attack in Pahalgam, Kashmir, left 26 people dead. In response, India launched a military operation named “Operation Sindoor” on May 6, blaming Pakistan for the assault. As attacks and counter-attacks took place between the two nuclear-power neighbors, an information war erupted in the online space, with its impact extending to Bangladesh as well.

In some cases, these false narratives glorified India’s military might, while in others, messages of Indian losses were used to express support for Pakistan. For example, old videos were circulated claiming India had attacked Pakistani military bases in Lahore, Karachi, or Sialkot, causing extensive damage. In response, similarly outdated footage (1, 2, 3) was used to claim Pakistan retaliated with attacks on various Indian locations. In some posts, old videos were shared, claiming heavy Indian casualties due to Pakistani strikes, with suggestions that India raised white flags while retrieving the bodies of its fallen soldiers.

Misinformation also circulated in the form of outdated videos claiming Pakistan had downed Indian fighter jets and drones (1, 2, 3). In some instances, footage from video games (1, 2) was shared as alleged proof of Indian aircraft being shot down. AI-generated images (1, 2) were also used to support similar claims. These visuals, including video game footage, old photos and videos, and AI images, also made their way into mainstream media reports (1, 2, 3, 4).

Another major topic of international misinformation was the Iran-Israel war. On June 13, after Israel launched a sudden strike on Iran, the two countries entered into an open conflict. Nearly all related misinformation that spread in Bangladesh favored Iran. As a Muslim-majority nation, religious sentiment plays a role in Bangladesh in shaping public sympathy toward countries like Palestine and Iran in their conflicts with Israel.

The central theme of such misinformation was to highlight the alleged scale of Israeli damage from Iranian attacks. For instance, a fire at Gaza’s power plant caused by an Israeli airstrike in 2014 was falsely circulated as evidence that Iran had destroyed an Israeli power station. Fires in a commercial building in Iraq, an airport in California, or a factory in Malaysia were misrepresented as consequences of Iranian attacks on Israel. Old videos—one from Gaza and another of mushroom harvesting in Nepal’s mountains—were falsely shared as footage of Israelis fleeing in fear of Iranian missile strikes.

Several AI-generated videos also surfaced in the context of the Iran-Israel war. These included visuals falsely depicting Iranian hypersonic missile launches, the destruction of Tel Aviv, or missile attacks on the headquarters of Mossad, Israel’s intelligence agency. One video, created using Google’s AI video tool Veo, showed an Israeli soldier and civilian pleading with Iran to stop the war. Another similar video claimed to show Israeli citizens marching and apologizing to Iran.

The rise in graphics card usage to spread misinformation

The trend of spreading false information by replicating the social media photo cards of news outlets—similar to the use of AI–has also been on the rise. In the first quarter of this year, 100 instances of such misinformation were recorded. In the second quarter, that number increased to 176 – 85% of them related to politics.

Compared to the first quarter, the second quarter also saw a greater use of videos in the spread of misinformation. In the first quarter, videos accounted for 42% of all misinformation. In the second quarter, this figure rose to 48%. Meanwhile, the use of images as a medium for spreading misinformation declined.

Methodology

This analysis is based on reports published between April and June 2025 by eight fact-checking websites (RumorScanner, BoomBD, NewsChecker, Fact Crescendo, Fact Watch, AFP Bangladesh, Ajker Patrika and Dismislab). In instances where multiple sites debunked the same misinformation, only one site’s fact-check report was considered as a unique sample to quantify subject-specific misinformation.

The unique fact-check reports were divided into 15 topics, including politics, sports, health, nature and environment, entertainment, fraud, disaster, religion, science and technology, law and order, education, economy & development, defense, policy & public affairs, and others.

Election related and AI-generated misinformation has been identified separately.