Mamdani’s Win Shows How Human Contact Can Defeat the Algorithm and the Chatbot

In the final weeks of New York’s mayoral race the streets hummed with familiar promises about housing, transport and trust. Yet somewhere beyond the rallies and the posters another kind of campaign was quietly under way. Voters sitting in cafés and kitchens were turning to chatbots, asking simple questions about policies and personalities, and getting confident answers. What most did not realise was that the machines had already started shaping the conversation.

The greatest modern threat to democratic choice is not the fake video or the troll farm. It is the polite helpful chatbot that narrows choice without intending to.

The regulator’s early warning

Weeks before the Dutch general election the Netherlands Data Protection Authority told voters bluntly: do not trust chatbots for voting advice. After twenty one thousand tests the agency found that popular models kept pointing to the same two parties while ignoring the rest. Deputy Chair Monique Verdier called their replies unreliable and clearly biased. The fault was not in malice but in design: systems trained on stale data and uncertain correlations.

Across the Atlantic the New York Attorney General issued a similar warning a year earlier. Chatbots, her office said, were often wrong even about polling hours and locations. If a machine cannot tell you when to vote, how can it tell you whom to vote for?

Together these warnings expose the new architecture of influence. As voters migrate to conversational AI for guidance, unseen design choices begin to shape what the public thinks it knows.

The New York example

On 4 November 2025 Zohran Mamdani, a 34 year old assembly member from Queens and a self described democratic socialist, defeated Andrew Cuomo and Curtis Sliwa to become the city’s first Muslim mayor and its youngest in a century. His campaign rested on housing justice, transport affordability and relentless community contact.

Yet the digital air around him felt weighted toward the past. Search engines and chatbots still summarised New York politics in the language of older names. No official audit of chatbot prompts has been released, so there is no proof of misconduct. But by parity of reasoning with the Dutch findings, there was a foreseeable risk that general-purpose chatbots would emphasise the better-known candidates and bury the insurgent.

It is not a conspiracy, only inertia disguised as intelligence.

The persuasive companion

Large language models can now match and sometimes surpass human persuasiveness in short conversations. That power now lives inside billions of private exchanges. People ask these systems for comfort, for career advice, even for moral reassurance. Each response is fluent, plausible and softly directive.

Multiply that across millions of users and you have something more intimate than a social-media feed: a digital counsellor that can shape world views by tone alone. Surveys already show that more than half of under-thirty year olds have sought personal or professional advice from AI tools. The scale of interaction gives these models psychological leverage that no previous medium ever held.

The analogue counter attack

Mamdani’s campaign succeeded because it refused to live entirely online. His volunteers knocked on doors, gathered in union halls and spoke in markets and mosques. The work was old-fashioned, human and slow. Yet in a year dominated by algorithms, the warmth of a handshake proved more persuasive than a thousand automated messages.

References

Netherlands DPA 2025 chatbot advisory; New York AG 2024 consumer alert; EU AI Act Article 6 and Recital 62; CMA foundation model market study 2025; peer reviewed research on AI persuasion; Pew and AP-NORC surveys on AI use; historic ICO and US Senate Cambridge Analytica inquiries.