Meta failed to identify 25 percent of political ads in the DUCSU election

Many have viewed this year’s Dhaka University Central Students’ Union (DUCSU) election as a test for the country’s democracy and student politics. It was also a major test for the role of online campaigns on social media platforms in a competitive election. The DUCSU election ultimately ended peacefully, but did the social media platforms pass the test? One area of social media’s role in the election is ensuring transparency in ad-based online campaigns. This allows voters to know who is running the ad and who is paying for it. However, neither YouTube nor Google publish information on political ads in Bangladesh, and  TikTok does not allow political ads at all. Among the major companies, only Meta, as part of its transparency policy, publishes information on electoral or political ads in Bangladesh. In this report, Dismislab examined the area of ensuring transparency in election ads on Meta’s platforms (Facebook and Instagram).

In the DUCSU election, many candidates ran campaigns using advertisements on Facebook. On September 7 alone, Dismislab found 231 campaign ads by searching for “DUCSU” in Meta’s Ad Library. The candidates and their supporters promoted these advertisements from their profiles or pages and spent money for them.

When an advertisement regarding politics or elections is placed on Facebook or Instagram, Meta verifies the advertiser’s identity and displays a small text on the ad – “Paid for by” – thus declaring the advertiser . This also allows ordinary users to easily know the advertisement cost. The purpose of this Meta policy is to increase transparency and prevent secret funding or fake campaigns.

Regular advertisements are only available in the Meta Ad Library as long as they are active. But political advertisements are kept in the library for seven years for the sake of transparency. The advertiser can voluntarily provide a “disclaimer” or identification information for the political ad. If they do not, Meta’s automated system identifies the ad as political and asks the advertiser to provide the information. If someone does not provide the information, Meta removes the political ad for a lack of transparency.

However, an analysis of the 231 DUCSU election campaign ads found in the Ad Library on September 7, just two days before the high-stake election, revealed that Meta failed to identify 25 percent of them as political or election-related ads. The investigation showed that while one candidate’s ad was removed for lacking a disclaimer, another candidate continued to run ads without any information unimpeded; one ad from the same candidate was removed, but others were promoted without a disclaimer.

As a result, voters were able to find out the amount of money spent on election ads on Meta for some candidates, but could not find out the same for others. According to Md. Pizuar Hossain, a Senior Lecturer at the Department of Law at East-West University and a PhD researcher at Monash University in Australia, this creates an uneven playground for candidates, where some get an advantage over the other. He attributed this to the weakness of Meta’s automated system for detecting political ads.

The “Paid for by” disclaimer shown on Meta’s political advertisements contains some mandatory information regarding the advertiser’s identity:

• Full name of the funder (individual or organization)
• Registered address
• Phone number
• Website
• Email

By clicking on a political ad, any user can go to the library and see:

• The start and end dates of the ad campaign
• The estimated amount of money spent
• How many times the ad was shown
• Whom the ad was targeted at (age, gender, region)
• Multiple versions of the ad, if any

Rules for some, not for others

Of the 231 election campaign ads , 66 percent were marked as political. This means these ads contain disclaimer information. The advertisers provided the transparency-related information themselves or after being identified by Meta. The advertisements were promoted between August 21 and September 7. The amount of money spent on these is also available there.

Meta identified 9 percent of the ads as political and removed them because the advertisers did not provide transparency-related information. But the platform failed to identify 25 percent of the election ads as political, and they ran unimpeded without transparency information.

Dismislab’s analysis shows that even when the language of the ads was the same, some were identified as political, while others were not. Arafat Chowdhury, an independent candidate for the DUCSU general secretary post, had two of his ads (1, 2) removed by Meta for lacking a disclaimer. However, five ads run from three pages associated with Sheikh Tanvir Bari Hamim, the general secretary candidate of the Jatiyatabadi Chhatra Dal-backed panel, were not removed by Meta despite lacking a disclaimer.

An election ad for member candidate Abir Hasan was removed for not providing information, but at least three of his ads ran on voting day without any disclaimer (1, 2, 3). An ad for vice-president candidate Umama Fatema was also removed by Meta. Her other election ads, however, did include transparency information.

The highest number of ads without a disclaimer, a total of five, were seen running from the page of Mustaqeem Mahmud Rahim, a candidate for the International Affairs Secretary post. Additionally, three ads each were seen running from panels named Swatantra Shikkharthi Oikya (Independent Students’ Unity) and Sachaton Shikkharthi Sangsad (Conscious Students’ Parliament) – none of which had any disclaimer. Meta’s automated system failed to identify these ads as political, and so these ads or related information are no longer in the Ad Library.

Md. Pizuar Hossain of East-West University told Dismislab, “When Meta identifies an ad as political, it faces some additional conditions. For example, it takes time to get approval to run the ad, a ‘Paid for by’ disclaimer must be shown, and this reduces the ad’s reach and effectiveness. But the candidates whose ads are not identified as political can easily reach many people without any obstacles. As a result, the competition is not equal.” He believes that the way Meta’s ad policy is enforced “becomes a greater influencing factor than the actual popularity and the election promises of a candidate.”

Concern over its impact on national elections

The question is, does a university-centric election campaign fall within the scope of Meta’s political advertising policy? Meta’s policy on ads about social issues, elections, and politics states that it applies to any election, referendum, or ballot initiative. Activities like encouraging people to vote or disseminating voting-related information also fall under its purview.

Previously, Meta applied this policy in university student union elections in India, Nepal, and Sri Lanka and removed ads that lacked disclaimers. The same scenario has been seen in election campaigns for the Harvard Graduate Students Union and the Cambridge Students’ Union.

Historically, Dhaka University and DUCSU elections are extremely important in Bangladesh’s national politics. The interim government’s Home Affairs Adviser stated: “Elections like DUCSU, JACSUs will be considered as models for the national election.” Although the scale and nature are different, many see the university election as a potential reflection of the upcoming national election.

Pizuar Hossain says, “If this kind of inconsistent flagging is repeated in the national election, it will damage public trust in the electoral [campaign transparency] process and its impartiality.” According to him, the relevance of these ads is not limited to the university context. “Meta should consider and realize that a university parliament election can also influence national-level elections.” 

Meta’s weakness in identifying political advertisements is not new. Earlier, a research by Digitally Right showed that before the 2024 national parliament elections, the company “flagged” or identified many irrelevant ads as political, while failing to identify ads from many politicians and parties. (Dismislab is primarily a research initiative of Digitally Right, and Pizuar Hossain was also involved in the 2024 study.)

How many ads for whom

An analysis of the 231 DUCSU-related ads shows that the highest number of ads on Meta were run in favor of the panel supported by Chhatra Dal, the student wing of Bangladesh Nationalist Party (BNP). A total of 76 ads were run in their support. Forty-nine ads were run on behalf of various independent candidates. In addition, 35 ads were seen running for the Swatantra Shikkharthi Oikya (Independent Students’ Unity) panel, 21 for the “Oporajeyo 71, Odommo 24” panel, and 14 for the Oikkoboddho Shikkharthi Jote (United Students’ Alliance) panel.

It is worth noting that these ads were found on September 7, from a search on just one day. The ads that ran before or after this date but were not identified as political are not included in this count. This is because if a regular ad is not active, it is not stored in the Ad Library. Consequently, they cannot be found later.

Two-thirds of the total ads, or 149, were placed by the candidates themselves from their own accounts. Fourteen were run from pages opened in the name of panels, and 68 were placed by supporters. Among the candidates, general secretary candidate Arafat Chowdhury was at the top in terms of ad campaigns on Meta. He ran 19 ads. He was followed by Anid Hasan, the candidate for the Literature and Cultural Affairs Secretary post. Thirteen ads were run from his page. Mustaqeem Mahmud Rahim, Naim Hasan Hridoy, and Umama Fatema each ran five ads.

In addition to the candidates and panels, various supporters were also seen running campaigns through ads in favor of a candidate or panel. The most ads, 21 in total, were run by a page named ‘Amader DUCSU’ (Our DUCSU). Furthermore, a page named ‘Amar DUCSU’ (My DUCSU) ran 11 ads. After that, 9 ads were run by the ‘BNP Media Cell’. All three of these pages campaigned in favor of the panel supported by Chhatra Dal. In addition, a page named “Kishorkantho 2.0” ran campaigns in favor of the “Oikkoboddho Shikkharthi Jote” (United Students’ Alliance) supported by Islami Chhatra Shibir. The number of their ads was 4.

Regarding campaigning on online/social media, the DUCSU election code of conduct states that such campaigns can be conducted in a lawful, positive manner. No activities that are prohibited by conventional law can be carried out. Personal attacks, character assassination, spreading rumors, or false information must be avoided.

Methodology

For this study, all relevant ads were collected by searching the keyword “DUCSU” in the Meta Ad Library on September 7, 2025. From among them, only those seeking votes for candidates of various panels were selected. Surveys, information-based posts, and news-like ads published around the DUCSU election were excluded.

Each selected ad was analyzed to see which ads had the “Paid for by” disclaimer, which ones did not, and which ads were removed by Meta for not having a disclaimer. We also analyzed which candidates from which panel were promoted through these ads and what types of page or profile were used for the campaign:

• From candidates’ pages/profiles
• From panels’ pages/profiles
• From supporters’ pages/profiles