Ghost Work: How India’s Annotation Army Guards AI Giants in US and China
Estimates suggest that by 2030, India’s annotation market could exceed $7 billion and employ more than a million. It is anchored not in Bengaluru or Hyderabad but in non-metros like Patna, Jaipur, Luc
Though annotators are paid modest sums (often a few cents per image), the algorithms they train create fortunes for Silicon Valley firms and Chinese AI giants. The gap between who builds the data and who profits from it is widening.
Image: Shutterstock
In a cramped room in Patna, Bihar’s capital city, a young graduate hunches over a screen, clicking boxes around traffic lights, crosswalks and bicycles. Thousands of kilometres away, in Jaipur, a homemaker earns extra income by labelling medical scans pixel by pixel. These are not glamorous jobs, nor do they come with titles that carry weight. Yet, without this quiet, monotonous labour, the algorithms that power self-driving cars in California or diagnostic AI (artificial intelligence) tools in London would remain blind.
India has become one of the world’s fastest growing hubs for data annotation, the unglamorous but essential task of teaching machines how to “see” and “understand”. Every AI model, no matter how advanced, is built on millions of labelled examples: A cat tagged as a cat, a cancerous tumour outlined in an X-ray, a Hindi phrase aligned with its English translation. The workers behind this industry form an invisible backbone, an economy of micro tasks that underpins trillion-dollar technologies abroad.
Estimates suggest that by 2030, India’s annotation market could exceed $7 billion and employ over a million annotators. Unlike India’s celebrated IT corridors in Bengaluru or Hyderabad, this new economy is anchored in places like Patna, Jaipur, Lucknow and Ranchi—cities dismissed as the “non-metros”. With only a laptop and a broadband connection, workers in these towns plug directly into the global AI supply chain.
The work itself is deceptively simple. A medical annotation job might involve drawing precise boundaries around organs in MRI scans to train a model in automated diagnostics. An automotive project might ask annotators to mark pedestrians across thousands of dashcam images, so self-driving algorithms can distinguish a child from a lamppost. Hours are spent clicking, tagging and correcting tiny strokes of human judgement that accumulate into datasets vast enough to train AI. It is tedious, repetitive and often invisible, yet utterly indispensable.
For India, this surge in annotation has two faces. On one side, it represents opportunity: A pathway for underemployed graduates, homemakers and gig workers in Tier II and III 3 cities to participate in the global digital economy. For many, it is their first step into tech work, accessible without advanced degrees or relocation to expensive metros. On the other side, it raises uncomfortable questions about labour and value.
Though annotators are paid modest sums (often a few cents per image), the algorithms they train create fortunes for Silicon Valley firms and Chinese AI giants. The gap between who builds the data and who profits from it is widening.
Companies like Scale AI, CloudFactory and Indian firms such as iMerit and Playment have become intermediaries in this ecosystem, channelling global demand into distributed Indian labour pools. Their marketing language speaks of “democratising opportunity” and “building the AI workforce of the future”. But the reality, for many workers, is hours of monotonous clicking for pay that rarely matches the glamour of the technology they fuel. The industry has begun to be described as “ghost work”, an invisible layer of labour that gives machines their intelligence while remaining unseen in the AI success stories.
As this sector grows, the stakes become clear. Will annotation remain an invisible gig economy of low pay and repetitive work, or can it evolve into a structured industry with career paths, skills training and protections? With projections of a million annotators by the end of the decade, India faces a choice: Allow this to remain hidden labour powering someone else’s algorithms, or recognise it as a critical digital industry in its own right.
For now, the young graduate in Patna continues to draw boxes around images they will never see again. The AI model will learn, the global tech firm will advance, and the worker will move to the next file in the queue. It is a quiet revolution, an economy of invisible clicks shaping the intelligence of machines that will never know their names.
The writer is a grade XII student at the Dhirubhai Ambani International School in Mumbai