Politics
How Facebook’s algorithm is amplifying one party over all others
An analysis of two dozen prominent Facebook pages shows the RSP appears in 54 percent of political posts—nine times those of Nepali Congress—often with AI-generated images.Daya Dudraj
In the month before Nepal’s parliamentary elections, popular Facebook pages with a combined six million followers posted about one political party far more than all others combined, an analysis by The Kathmandu Post has found.
The Post examined 4,754 posts published between December 23, 2025 and January 22, 2026, from 24 Facebook pages with hundreds of thousands to several million followers. Of those total posts, 2,447 were political content. The Rastriya Swatantra Party (RSP) content appeared in 1,324 posts (54.1 percent of political posts)—nearly nine times more than Nepali Congress and more than twice as much as Nepali Congress and Communist Party Nepal (UML) combined.
The pages—none officially affiliated with any political party—range from social media favourites like ‘Routine of Nepal Banda’ to regional pages like ‘Troll Pokhara’ (725,000 followers) and ‘Nepalgunj Gallery’ (178,000 followers). All primarily publish mixed content: local news, educational information, humour, satire and entertainment.
The analysis reveals a striking imbalance in how one political party dominates content during Nepal’s election season. Whether this reflects coordinated strategy, organic enthusiasm, or algorithmic preference for emotional content remains unclear. What the data shows definitively is that content favouring the RSP, its leaders Balendra Shah and Rabi Lamichhane, and the party’s blue bell symbol vastly outweighs coverage of the two traditional parties that have governed Nepal for decades. Additionally, almost all the material reviewed by the Post shows a focus on personality-driven and emotional imagery while ignoring policy debate.
What is also striking in this election cycle is that the traditional metrics of campaign strength—rally sizes, door-to-door canvassing, policy papers—may matter less than understanding how to harness social media platforms where millions of Nepalis now get their information, which they believe is fact-based news.
The 24 pages were selected based on three criteria: follower count, engagement levels, and regular posting activity. Only pages demonstrating high interaction through likes, comments and shares, with visibly active follower bases, were included.
For each page, the Post analysed the most recent 200 posts, and used AI-based tools to help categorise content, identify thematic patterns and detect political references. Posts were coded as political if they referenced parties, candidates, elections, or governance.
Posts mentioning RSP leaders, the bell symbol, or party messaging were coded as RSP-related. The methodology of this analysis identifies patterns in published content but cannot determine intent, coordination, or whether page administrators have connections to political parties.
Of 192 posts published by the page ‘Alternative for Nepal’ during the study period, 181 (94 percent) were political content. Of those political posts, approximately 60 percent focused on RSP, Balen Shah, or Rabi Lamichhane.
The page used the bell emoji in numerous posts.
Similarly, ‘Ktm Post’, with 147,000 followers, posted 161 items total: 111 about RSP, six about UML, three about Nepali Congress, one about other political parties, and 37 non-political posts.
‘The Nepali Favour’, a page focussed primarily on memes, posted 200 items during the study period. Of these, 135 (68 percent ) were politics-related, with “Balen Shah” and “Rabi Lamichhane” as the dominant keywords. The bell emoji and references to “RSP,” “candidacy announcement,” and various campaign locations appeared repeatedly.
‘Routine of Nepal Banda’, with nearly 5 million followers, posted 200 items between January 17 and 23. Of these, 99 (49.5 percent) were political content. Within the political posts, RSP-related content dominated, with “Balen Shah,” “Rabi Lamichhane,” and “RSP youth candidates” as the most frequent keywords. By contrast, Nepali Congress appeared in 11 posts (5.5 percent), UML in 10 posts (5 percent ), and NCP in 4 posts (2 percent).
“The current moment is highly emotional,” said Dovan Rai, a technologist who studies online behaviour and oversees R&D at Body & Data, a digital research firm. “People are making decisions based on their gut feelings, which are further amplified by algorithms.”
Rai compares algorithms to traders who “develop a kind of intuition from data” and “push whatever sells or goes viral even more aggressively.”
Research supports this observation. Studies of Facebook’s algorithm show that content generating strong emotional reactions—particularly outrage, inspiration, or anxiety—receives higher distribution than neutral information.
When Facebook’s internal documents were leaked by whistleblower Frances Haugen in 2021, the company’s own research showed its algorithm prioritises content provoking “angry” reactions, leading to what researchers called “more divisiveness and more polarisation,” as reported in the Wall Street Journal’s Facebook Files investigation in September 2021.
The pattern the Post identified in Nepal mirrors findings from elections globally. Investigations into Facebook’s role in elections in the Philippines, India, and Myanmar documented how personality-driven, emotionally charged political content systematically outperforms policy-focussed material. In Myanmar, UN investigators found that Facebook’s algorithm had amplified hate speech and misinformation that contributed to violence against Rohingya Muslims, leading the company to acknowledge in 2018 that it hadn’t done enough to prevent its platform from being used to “foment division and incite offline violence”.
The content the Post analysed did not focus on hate speech. Instead, they featured emotional narratives: leaders visiting hospitals, meeting with grieving families, surrounded by supporters. The content avoided policy discussion in favour of personal stories and symbolic imagery—precisely the material that Facebook’s algorithm, according to extensive research, is designed to amplify. Many posts featured AI-generated images of RSP leaders.
When journalist Dinesh Sitaula, who had been accompanying RSP candidate Balendra Shah on his Achham trip, died on February 1, multiple pages quickly posted images appearing to show Shah mourning. At 6:06 pm, one page posted Shah offering tributes. An hour later, another showed him standing before a funeral pyre, apparently in tears. An hour after that, a third image depicted Shah and RSP President Rabi Lamichhane placing garlands on a body during funeral rites.
Even though all three images already visibly look fake, the Post ran them through Hive Moderation, an AI detection tool, and all three of them showed a 99.9% likelihood of being AI-generated.
Similarly, on January 15, the page ‘Miss Pabi 41’ (95,000 followers) posted an image showing popular YouTuber Bhojraj Thapa lying ill in a hospital bed, with Lamichhane standing beside him with folded hands, apparently crying, while Shah held Thapa’s hand. “Rabi and Balen reached early in the morning to visit Bhojraj Thapa,” the caption read. “Everyone, please write ‘Get Well Soon.’”
The post drew hundreds of comments, most writing “Get Well Soon.” That image was also AI-generated.
Several pages also shared demonstrable misinformation, including AI-generated images and false claims that said Nepali Congress President Gagan Kumar Thapa had been arrested, or that CPN-UML leader and former prime minister KP Sharma Oli and former home minister Ramesh Lekhak had been taken into police custody. Other posts showed the public chasing Nepal Communist Party leader Pushpa Kamal Dahal.
Deepak Adhikari, a fact-checker and journalist who monitors social media, confirmed he has observed this pattern. “These same pages are at the forefront of spreading misinformation,” he said.
The most influential page in the network, ‘Routine of Nepal Banda’, has a history connected to RSP leader Balendra Shah, now the party’s candidate challenging CPN-UML Chairman Oli in Jhapa-5.
On September 2, 2023, Shah, then Kathmandu’s mayor, posted an incendiary status on Facebook after the vehicle his wife was in was reportedly stopped by traffic police for a routine check. “For today, let it pass,” Shah wrote. “But if the government stops any KMC vehicles from tomorrow onwards, I will set fire to Singha Durbar.”
The following day, RONB posted congratulations to Shah on the birth of his daughter. The platform was accused of posting the message to deflect from the controversy about Shah’s violent threat to burn down the government’s administrative headquarters. As criticism mounted, RONB founder Victor Paudel announced he was shutting down the site. He revived it several days later.
Even though it has faced previous accusations of political bias or posting status updates without any context, it remains defiant with a large youth following—including many who support Shah’s candidacy and want to see him beat Oli in the hotly contested race in Jhapa.
On September 6, 2025, Routine of Nepal Banda posted information about a vehicle hitting a girl. The post read: “Just in: In Harisiddhi, Lalitpur, a provincial minister’s car hit a young child and fled, due to which the public has been blocking the road since morning.”
Thirty-seven minutes later, another post was published. On a photo of then prime minister KP Sharma Oli, it read: “When a car hits the public, leaders don’t stop and walk away, and the prime minister says it's ‘normal’ and brushes it aside: About the recent accident in Harisiddhi, Prime Minister KP Sharma Oli said, ‘A car ordinarily touched a child. She has been taken to the hospital. There was no such intention.’”
Earlier this month, a four-year-old died after being hit by a vehicle with an RSP sticker in Rautahat. In that incident too, RONB posted something. “Sad news: A young girl playing on the road was hit by a car in Rautahat and died. Further investigation is ongoing. Source: District Police Office, Rautahat”
But unlike the September incident, instead of posting another update in a separate post, it edited the original post to add information. Twenty-two minutes after the first post, it was edited to add: “Further investigation is ongoing. Update: When we inquired at the District Police Office Rautahat, the accident-causing vehicle was found empty and had an RSP sticker on it, and now the police have taken the driver into custody.”
Exactly 50 minutes later, it was edited for the third time to read: “The driver has been taken into police custody.”
Routine of Nepal Banda was criticised for softening its post when writing about an incident involving RSP while presenting an opinionated criticism when the incident involved another party.
RONB told the Post that claims of selective posting to favour any candidate or party are untrue.
“We are not selective in the information we post and we have not backed any party or candidate,” Victor Paudel, RONB’s founder, said. “The reason Balen’s posts may seem more visible is because they draw significantly higher engagement, while posts about others do not.”
Not all pages examined by the Post avoided substantive political commentary. ‘KMAG Nepal’, which claims to publish “general news and updates on daily activities” and is registered under the name of Lakshya Sapkota, posts content with clear ideological framing to its 103,000 followers. The account shows that at least one of its admins is based in Australia.
Of 200 posts examined between January 1 and 22, 129 posts were directly related to politics. These posts frequently discussed institutional reform, rule of law, open market thinking, individual freedom, and informed voting. The page gave particular prominence to Gagan Thapa of Congress, portraying him as “the most reformist, liberal, and forward-looking leader within Nepali Congress” and “a figure who can enable change within old parties, as a representative of the young generation.”
However, references to RSP and its leaders still dominated the page’s political content, with “Balen Shah” and “Rabi Lamichhane” as the most frequently mentioned figures.
None of the posts analysed included substantive discussion of policy positions, party manifestos, or comparative analysis of governance proposals.
Instead, the posts focused on personal stories, emotional appeals, and symbolic imagery. The bell emoji, RSP’s election symbol, appeared across multiple pages, often without explicit partisan messaging but in contexts that associated the symbol with hope, change, or reform.
The Post found no evidence that the two dozen pages are coordinated by RSP or acting under the party’s direction. None of the pages has disclosed formal connections to the party, nor are any of the page administrators publicly identified with RSP.
However, the collective amplification of similar messages in these pages raises some questions: Does RSP benefit from an organic ecosystem of supporters who independently create favourable content? Does Facebook's algorithm simply favour the emotional, personality-driven content that RSP's brand embodies? And can traditional parties compete without adopting similar tactics?
Adhikari, the fact-checker and journalist, said a “narrative war” is underway on social media through AI-generated images and videos that could influence election outcomes. “Although there has been no systematic study on this in Nepal, and it may be too early to draw firm conclusions, the style and intensity of campaigning through social media is clearly increasing,” he said.
Traditional political parties say they are working to adapt to the digital landscape, but with limited resources and largely volunteer teams.
Min Bahadur Shahi, head of the UML’s publicity department, said the party focuses on short-form video for younger voters. “At present, we are producing TikTok and short-form videos to communicate with the Gen Z generation,” he said. “Our plan is to engage them on issues related to anti-corruption and global indices.”
Shahi emphasised that UML’s digital efforts are organic. “Whatever we are doing in the digital campaign is entirely voluntary,” he said. “We do have friends with knowledge of digital and social media, but they are not a paid team. We have not paid Meta, nor have we spent money on boosting any posts.”
Nepali Congress operates a 15-person media team handling video editing and news writing, according to Media Centre Coordinator Jiban Bhandari. “We work on content production by creating various photos, videos, and clips based on the party's agendas,” Bhandari said. The centre compiles district news and distributes it to media outlets nationwide.
“Our main focus is to turn the party's activities, plans, and both past and present work into content,” he said. “In terms of social media, our primary focus is on Facebook, but we also have a strong intellectual community on TikTok and X.”
The Nepali Communist Party has prioritised countering misinformation and AI-generated content, said party spokesperson Agni Sapkota. And like UML, NCP relies heavily on volunteers. “Most of our colleagues are working on a voluntary basis,” Sapkota said. “We make up for the lack of resources through hard work.”
The contrast between what traditional parties describe and what the Post’s analysis shows is striking. Despite UML’s focus on “Gen Z” and Congress’s 15-person team, their content appeared in 19% of political posts across the analysed pages—less than one-fifth of RSP’s presence.
RSP has formed digital campaign committees at the central, constituency and candidate levels to run its digital campaign, party spokesperson Manish Jha said.
“There are two or three teams working from the centre, while election mobilisation committees have been formed at the district level,” Jha said. “Candidates themselves are also involved. Those contesting under the first-past-the-post system are running massive campaigns, and proportional representation candidates are also actively engaged.”
Jha said the party has not paid any social media pages to post content on its behalf. “We have not entered into any contracts with popular pages like Routine of Nepal Banda to carry our content,” he said. “Rather, they have supported us on their own. Nothing has been done through any formal party process. It has all been spontaneous, though there may have been occasional requests.”
Responding to questions about why emotionally charged content and AI-generated misinformation appear frequently in posts about RSP, Jha said the party produces content aligned with public sentiment but does not promote false information.
“This falls under campaign strategy. In a 360-degree campaign strategy, there is a saying—‘you can sell emotion without being emotional.’ That is strategy,” Jha said. “We have not distorted facts. We are working within a certain discipline.”
Jha did not directly address the Post’s findings of AI-generated images showing RSP leaders at events that never occurred, or false posts claiming rival politicians had been arrested.
What is also unclear so far is how much of the social engagement and online support for the one party translates into votes at the ballot box. The Post interviewed several voters in Kathmandu about their social media use and political information sources in the past few months.
Nirajan Khadka, a 48-year-old shopkeeper from Anamnagar, hasn’t read any party manifestos or watched candidate debates. He’s formed opinions from hours of Facebook scrolling each morning.
“Renu will definitely win in Chitwan-3,” he said, predicting the race featuring the daughter of Nepali Communist Party coordinator Dahal. “And Oli will lose in Jhapa-5.”
Khadka represents a growing segment of urban voters who rely on social media for political information. “I usually check horoscopes first thing in the morning and then news from Nepal and the world,” he said. “These days, it’s been entertaining to scroll through candidates’ gimmicks during their campaign.”
But he doesn’t see variety. “Sometimes there are photos saying so-and-so was arrested. Sometimes photos saying someone went to visit a hospital,” he said. “Maybe because I look at such things a lot, the same people keep repeating.”
Prakash Bishwakarma, a 28-year-old hotel business operator in Thapagaun, sees the same pattern. “I do see candidates from other parties too, but most of what I come across is from the ‘bell’ candidates,” he said, referring to the RSP’s election symbol. “To be fair, their content is also quite entertaining.”
For Purna Bahadur Basnet, a 56-year-old cosmetics shop owner, the imbalance frustrates him. “These days, RSP candidates keep reappearing on my feed and I wish there was a way to remove them,” he said.
Meta, Facebook’s parent company, has faced growing scrutiny over how its platform influences elections globally, particularly in countries besides the United States. At the end of 2024, the year that saw a record number of elections around the world, Meta said it has a dedicated team responsible for its election integrity efforts, including members from its intelligence, data science, engineering, research, content and public policy, and legal teams.
“We ran a number of election operations centres around the world to monitor and react swiftly to issues that arose, including in relation to the major elections in the US, Bangladesh, Indonesia, India, Pakistan, the EU Parliament, France, the UK, South Africa, Mexico and Brazil,” Meta said in its note.
Critics have noted that Meta’s resources for election monitoring remain heavily concentrated in Western democracies and large markets, leaving smaller countries like Nepal with minimal oversight.
Nepal’s Election Commission says it has been closely monitoring social media platforms for content that violates the election code of conduct. When such material is identified, the commission formally writes to the relevant authorities seeking its removal.
The commission recommends action to the Cyber Bureau of Nepal Police and the Nepal Telecommunications Authority, after which the bureau reaches out to the platforms or institutions concerned to take down the flagged content,” said Sita Pun Shrees, assistant spokesperson for the Election Commission.

According to Shrees, as of February 20, the Cyber Bureau has proposed the removal of 290 pieces of content found on social media. Of these, 107 have been taken down. Most of the removed content was on TikTok, while the number of Facebook posts taken down remains negligible. Of the 24 pages examined by the Post, not a single one had taken down the disputed content.
Shrees expressed dissatisfaction with Meta’s lack of cooperation for the upcoming election. During the 2022 elections, she told the Post, the social media platform had created a dedicated portal to facilitate the reporting and removal of content violating the code of conduct. “This time, we have not received the same level of support,” she said. “We write to them requesting the removal of specific content, but there is often little response.”
The primary challenge for officials is identifying individuals operating Facebook accounts. “If we can identify the account owners or administrators, we can directly ask them to remove the content,” Shrees said. “But Meta has not been cooperative in helping us trace account holders.”
The Post contacted Meta four times between February 2 and February 20 via email to ask about Facebook’s policies governing political content during Nepali elections, whether it monitors coordinated inauthentic behaviour in Nepal, and whether it has identified violations related to these pages. Meta did not respond to our emails.
With less than a month to go before one of the most consequential elections in the country’s history, voters like Khadka are not unique. Their morning scroll routine is now the electoral battleground. And on that battlefield, data shows, one party’s bell is ringing louder than all others.




8.65°C Kathmandu















