ILLUSTRATION: ADAM MAIDA
A NEPALI NEWSROOM published a true story about a popular politician. Within days it had lost more than 100,000 followers and was scrambling to repair the damage. Nobody at the publication had made an editorial mistake. The algorithm had simply decided the story was unwelcome, and an audience trained to react instantly did the rest.
The episode involved Onlinekhabar, one of Nepal’s largest digital outlets, after it published an article examining Balendra Shah’s conduct in relation to the national flag, back when he was still mayor of Kathmandu and building the online following that would later carry him to the prime minister’s office.
Shah’s supporters organised an unfollow campaign on Facebook so effective that the outlet’s page lost over 100,000 followers in a short span, while his backers flooded other platforms, including mainstream media comment sections, to drown out the story with content favourable to him.
Onlinekhabar, by its own later admission, had to actively work to limit the reputational damage. A single article had been effectively punished by a coordinated, algorithmically amplified response before any editor could decide how to follow up on it.
That dynamic, once a curiosity, is now closer to how Nepal’s news cycle actually operates. The clearest evidence sits in a Kathmandu Post analytical report from February 2026, which examined 4,754 posts across 24 of Nepal’s highest-follower Facebook pages over one month, from late December 2025 to late January 2026, in the run-up to the general election.
It found that 54.1% of all political posts were heavily favourable to the Rastriya Swatantra Party (RSP)—more than double the combined share for the Nepali Congress and the UML put together. This was not because the RSP had twice as much to say. IT expert Rajiv Subba put it plainly a week before the vote: “The dominance of one party on social media reflects the influence of algorithms.” The RSP had built a technical team that understood how Facebook’s and TikTok’s recommendation systems actually worked, and the resulting advantage looked less like superior messaging than superior engineering.
The mechanism that mattered most was a hashtag trend that began, of all places, in Indonesia. #NepoBabies, originally aimed at a controversial attempt to install a president’s underage son as vice-president, migrated to Nepal through TikTok and Instagram reels and was rapidly localised into a format contrasting politicians’ children’s luxury lifestyles—one viral clip showed a Christmas tree decorated entirely with Louis Vuitton boxes—against ordinary Nepalis’ daily struggles. The trend arrived pre-packaged from a foreign platform’s recommendation engine, was adopted and amplified by Nepali TikTok and Instagram users, and only then became something editors had to cover, react to and eventually follow rather than lead.
The official verdict on what this did to Nepali democracy came, remarkably, from the government’s own commission of inquiry. The Karki Commission, formed to investigate the violent September 8th–9th 2025 protests that toppled the previous government, submitted its report to then-interim prime minister Sushila Karki in March immediately alongside the historic election.
Its conclusion named the force directly: “Not only in Nepal’s Gen Z movement but also in the current electoral atmosphere, these two technical concepts—filter bubble and echo chamber—played and continue to play the greatest role.” A government inquiry into the deadly protest movement found that the main factor was not how any single news outlet reported the story, but how it spread across social media platforms.
That conclusion has its critics, and the criticism matters because it cuts against an obvious story of algorithmic omnipotence. UML leaders told the House of Representatives in April that “invisible forces—AI, algorithms, Goebbels-types—established the champions of nationalism, democracy, republicanism, federalism, peace, and prosperity as dangerous villains,” blaming the algorithm for their party’s defeat rather than examining what RSP did better.
Some analysts are more restrained, pointing out that virality and votes are not the same currency: candidates who dominate Facebook and TikTok engagement have, in past Nepali elections, sometimes underperformed at the ballot box relative to quieter rivals with stronger grassroots organisation. Algorithms can manufacture attention. They cannot, by themselves, manufacture trust, and Nepal’s 2026 result was a landslide built on real grievances that the platforms amplified rather than invented.
What is difficult to dispute is the redistribution of power this has produced inside Nepali journalism itself. Editorial boards used to decide what counted as news and where it sat on the front page. Increasingly, that decision is made upstream, by recommendation systems whose criteria no Nepali editor can see or appeal, with the newsroom’s job reduced to reacting fast enough to whatever the feed has already decided matters.
Balendra Shah’s own April public asset declaration recorded Rs14.6m in cash generated from social media channels like Facebook and YouTube, a detail that captures the inversion neatly: the country’s most powerful political figure built his career monetising the same platforms that now set the terms his coverage must operate within.
Advocate Babu Ram Aryal has been pushing for a data protection law that would force some transparency onto these systems. Nepal still has no such law. Until it does its newsrooms will keep discovering, one viral backlash at a time, just how much editorial authority they have already lost. ■
