Kumar Saikrishna Nayak
Elections used to be loud. Street-corner rallies, hoarse loudspeakers, campaigners going door to door, entire neighbourhoods gathered around a television for the evening debate. That version of democracy is dying, and most of us haven’t quite noticed the funeral. The real contest now happens somewhere quieter — inside a smartphone screen, in the space between a thumb and a feed. Political persuasion doesn’t shout at crowds anymore. It whispers to individuals, one at a time, and the whisper has been tested, refined, and timed with a precision no rally could ever match.
This isn’t speculation about some AI-run future. It’s already part of how political communication works in India and elsewhere with widespread mobile internet access.
From the Rally to the Feed
Campaigns describe their own methods as “smart” and “surgical,” and for once the marketing language is accurate. Nobody is addressing an anonymous mass anymore. Every voter is a profile being built in real time — every scroll, every pause, every like or skip can become another data point feeding a model of who you are and what might move you. Linger two seconds longer on a video about unemployment, and your feed may begin showing you more content like it, each one worded slightly differently, until what started as manufactured content starts to feel like your own settled opinion.
What makes this so effective is precisely that it doesn’t feel like anything is being done to you. A curated feed never announces itself as curated. It just feels instinctual — as if the device has learned to think alongside you rather than manipulate you. Two people can work in the same office, live in the same city, share the same friend group, and still inhabit completely different informational worlds. One sees hope where the other sees decline. One sees progress where the other sees collapse. The algorithm isn’t necessarily lying to either of them. It’s simply reflecting their inclinations at them, amplified, until the reflection becomes impossible to distinguish from the self that’s supposedly thinking.
Try this yourself, and it gets uncomfortable fast: spend one full day deliberately engaging with the opposite side of your politics. Like the arguments you find weak. Click into the topics you usually scroll past. Within hours, your feed may begin to recalibrate. The old version — the one that felt so personal, so obviously true — can start to look different, as if it never existed. That should bother you more than it probably does. If a belief can be reshaped in an afternoon just by changing what you click, how solid was it to begin with? Political conviction, it turns out, isn’t always bedrock. Often it’s just a data point that hasn’t been nudged yet.
A Global Infrastructure of Influence
None of this is a local phenomenon. In the United States, campaigns use predictive modelling, experimentation, and targeted advertising to identify voters and test which messages are more persuasive. In China, the digital infrastructure is subject to extensive state regulation, including rules governing algorithmic recommendations and the presentation of information. In the European Union, regulators have moved beyond drafting rules: new regulations governing the transparency and targeting of political advertising became fully applicable in October 2025, imposing significant restrictions on the use of personal data for political targeting.
Different governments, different justifications, same underlying playbook. None of it necessarily depends on coercion, because coercion is visible, and visible things get resisted. The real method can be subtler: build the environment so precisely that a person mistakes the nudge for their own free judgment. Nobody feels manipulated when they’re convinced the idea was theirs all along.
The Voting Booth, Revisited
There’s a specific kind of unease in standing inside a polling booth and asking yourself a plain question: was this decision actually made here, in this moment of privacy — or weeks ago, in a server room, by a system that had already modelled the outcome? When results appear close to pre-poll forecasts, “good analytics” is one explanation. Successful engineering is another. At a certain level of precision, the two can start to look identical.
This matters more, not less, in a country as vast and unequal as India. Behavioural targeting doesn’t treat every citizen the same way. Voters with less digital literacy, fewer alternative sources of information, or less social capital to question what their screen tells them may be more vulnerable to this quiet architecture of persuasion. A technology that presents itself as neutral can still deepen existing inequalities, simply by aiming most precisely at the people least equipped to notice they’re being aimed at.
The real danger isn’t that an algorithm shows you things you like. The real danger is what happens to the idea of the “undecided voter” — the citizen who walks into the booth still weighing options, still capable of changing their mind at the last second. When a system can predict, test, and reinforce a belief months before a single ballot is cast, the act of “deciding” can begin to feel less like a decision and more like a formality. The choice may have been shaped gradually, across a hundred small nudges disguised as ordinary browsing.
Democracy runs on the assumption that people arrive at their own conclusions, however imperfectly. What’s left of that assumption once the conclusion has been influenced by a system specifically engineered to produce a particular response?
These are not arguments for switching off our phones and going back to loudspeakers — that option no longer exists. But it is an argument for a specific kind of vigilance. Voters need more than digital literacy; they need the habit of noticing when a belief feels a little too comfortable, and the discipline to go looking for the discomfort of an opposing view on purpose. Ask, occasionally: who benefits from how easily I arrived at this conclusion?
Regulation has a role too — transparency in political advertising, real limits on behavioural micro-targeting, independent audits of how campaigns handle voter data. But regulation that struggles to keep pace with rapidly changing technology can keep losing the race by design. The European Union’s new political-advertising rules show that regulation can move from proposal to enforceable requirements, but the underlying technology continues to evolve.
Maybe the oldest safeguard is still the best one: doubt. The willingness to pause before a conviction settles in and ask where it actually came from. In an age where belief can be modelled and influenced before a voter ever reaches the booth, that small, stubborn refusal to be perfectly predictable might be democracy’s last working defence.
~~~
Kumar Saikrishna Nayak is a third-year B.A. (Hons.) History student at Motilal Nehru College (Evening), University of Delhi. He hails from Odisha.
