In 2005, Jeff Bezos launched a platform he called “artificial artificial intelligence” that once had more than 500,000 workers. Amazon is now shutting it down

The world of Artificial Intelligence is changing every day with newer, more capable AIs being launched by various companies. One of the prime example of that is the Amazon's Mechanical Turk.

When it was launched in 2005, Amazon founder Jeff Bezos even hailed it as “artificial artificial intelligence.” The phrase reflected the platform’s basic premise: when computers could not reliably complete a task, people could do it instead. In its prime, the platform had over half a million workers. But now Amazon is set to shut it down later his month, after about 20 years.

What is Mechanical Turk

Mechanical Turk is a crowdsourced work platform that for more than two decades connected people around the world with small digital tasks that computers once struggled to perform. The service will officially close on September 30, 2026.


It became an early example of how human workers could be integrated into digital systems at enormous scale. Workers, commonly known as “turkers,” completed tasks such as data labeling, audio transcription, surveys and image categorization. At its peak, the platform served more than 500,000 workers.

Its closure marks a significant shift in the relationship between human labor and artificial intelligence. The technology that once depended heavily on crowdsourced human input has advanced to the point where many of the tasks Mechanical Turk was created to handle can increasingly be performed by AI.

The name itself was borrowed from the 18th-century Mechanical Turk, a famous chess-playing automaton that appeared to operate independently but was secretly controlled by a human chess player hidden inside the machine. Amazon's version similarly relied on people working behind the scenes to perform tasks that software could not yet handle effectively.

The company initially envisioned Mechanical Turk as a way to help process and label the enormous quantities of information generated by Amazon's online marketplace. Over time, however, its role expanded into a much broader marketplace for small online jobs.

The platform's importance began to change as artificial intelligence developed rapidly. At the same time, specialized companies such as Scale AI, Mercor and Prolific built businesses around recruiting people specifically to help train and evaluate AI models.

Amazon also attempted to reposition Mechanical Turk as a source of human data annotation for its SageMaker machine-learning services. The company promoted access to workers around the clock, allowing businesses to obtain human review and labeling at relatively short notice.

But Mechanical Turk increasingly faced competition from newer platforms designed around the rapidly expanding AI industry.

Krista Pawloski, a data worker and organizer with Turkopticon, an advocacy group for data workers, said the platform had already been in decline as the wider AI-training industry moved forward. She argued that Amazon appeared to invest less in improving Mechanical Turk even as competing services gained ground.

That decline was particularly significant for workers who depended on the platform for income.

Pawloski began working on Mechanical Turk in 2008 while on maternity leave. What started as a source of supplemental income eventually became full-time work in 2012 after she lost her job and needed a flexible way to earn money while caring for her son, who has special needs.

She began organizing with Turkopticon in 2019. The organization initially focused on Mechanical Turk workers before expanding its advocacy to data workers more broadly.

For workers like Pawloski, Mechanical Turk represented something more than a collection of small online tasks. It provided flexible, remote employment that could be done from home and, in many cases, from a smartphone. For some workers, it became a full-time source of income.

The platform's decline also exposed an unusual problem created by the rise of AI: humans were increasingly being hired to produce data for artificial intelligence systems, while some workers were simultaneously using AI themselves to complete those assignments.

A 2023 academic study from Switzerland found that as many as 46% of Mechanical Turk workers surveyed were using AI models to complete tasks that were expected to be performed by humans. The finding raised questions about the reliability of crowdsourced data used to train AI systems — particularly when AI-generated material could end up becoming part of the training data for future models.

Mechanical Turk's shutdown therefore represents more than the end of an Amazon service. It closes a chapter that began when AI systems were far less capable and companies needed large numbers of people to perform tasks machines could not yet manage.

For hundreds of thousands of people who passed through the platform over the years, however, its legacy is more personal. Mechanical Turk offered a form of work that was flexible, remote and accessible, even if the individual jobs were often small.

A dedicated group of workers still relies on the platform as a livelihood. With its September 30 closure approaching, those workers are now confronting the question that sits at the heart of the changing AI economy: what happens to the human workers whose jobs helped build the systems now capable of replacing the work?