Amazon announced on August 25, 2026, that it will permanently close Mechanical Turk, its crowdsourced digital work platform, effective September 30. The decision ends a 21-year run for a service that occupies an unusual place in tech history: it was built specifically to paper over the gaps in what early computers and algorithms couldn’t do, using human labor instead.

“We regularly evaluate our programs, tools and services and make adjustments based on those assessments,” Amazon wrote in a notice posted to the Mechanical Turk website. “Following an assessment, we’ve made the decision to close AWS Mechanical Turk, effective September 30, 2026.” Amazon founder Jeff Bezos gave the platform its most memorable description years ago, calling it “artificial artificial intelligence” — a nod to how it quietly inserted human judgment into processes that looked automated from the outside.

What Mechanical Turk Actually Did

Launched in 2005, Mechanical Turk connected businesses and researchers with people willing to complete small digital jobs known as Human Intelligence Tasks, or HITs — things like labeling images, transcribing audio, moderating content, and answering surveys, typically paid just a few cents per task. At its peak, the platform counted more than 500,000 registered workers. Over time, Mechanical Turk became closely tied to Amazon Web Services’ broader machine learning offerings, marketed as a source of round-the-clock human review and data annotation to support products like SageMaker.

The platform’s name is itself a reference to an 18th-century hoax: a chess-playing machine built by Wolfgang von Kempelen that appeared to defeat human opponents, including Napoleon Bonaparte, but secretly concealed a human chess player operating it from inside. Amazon revived that concept for the internet age, building a system where humans performed tasks that appeared automated but weren’t — mirroring, in reverse, the same illusion the original machine relied on.

Why It’s Closing Now

The timing carries a certain irony. Mechanical Turk existed because early AI and automated systems couldn’t reliably perform basic classification, transcription, and recognition tasks on their own. Two decades later, AI systems have become capable of doing exactly that kind of work, and specialized data-labeling competitors — including Scale AI, Mercor, and Prolific — have emerged to serve the AI training industry more directly than Mechanical Turk ever did. Amazon had already signaled the platform’s decline last month by cutting off new customer signups, a move workers on online forums correctly interpreted as a warning sign.

Krista Pawloski, a data worker and organizer with the advocacy group Turkopticon, told CNBC that Mechanical Turk had been in decline for years, with Amazon investing fewer resources into improving it as competing platforms drew workers away.

A Twist: Workers Were Already Using AI to Do the Work

Perhaps the clearest sign of how thoroughly AI has reshaped this corner of the internet is a 2023 academic study cited in multiple reports on the shutdown, which found that as many as 46% of surveyed Mechanical Turk workers had used AI tools to help complete their assigned tasks. In other words, a platform built to supply verified human judgment had, in practice, become partially populated by AI-assisted output anyway — blurring the exact distinction the service was designed to preserve, and undermining its value as a source of clean, human-generated training data.

What This Means for the Workers Left Behind

For the workers still active on Mechanical Turk, the closure carries real, immediate consequences — some have relied on the platform as a full-time source of flexible remote income for years, according to Pawloski. Amazon has pointed affected workers and requesters to an FAQ page explaining how to prepare for the September 30 shutdown, and most will need to look toward newer, AI-focused data-labeling platforms for similar work going forward. Beyond the individual impact on Mechanical Turk’s user base, the shutdown marks a symbolic milestone in how completely AI has taken over a category of work that, not so long ago, only humans could reliably do.