Amazon sunsets Mechanical Turk to new customers after 21 years

Amazon's Mechanical Turk, the crowdsourcing platform that helped train a generation of AI models, will stop accepting new customers next month — a quiet end for a service whose name itself was a 250-year-old hoax about artificial intelligence.

Amazon Web Services will close Mechanical Turk to new customers effective July 30, 2026, according to a notice on the platform's website. Existing users can continue using the service, but AWS said it does "not plan to introduce new features," effectively placing the platform on life support.

"Someone at Amazon is going to decide keeping the MTurk servers running is a waste of time and resources and pull the plug entirely," one user wrote on Reddit after the announcement, reflecting a sentiment that the platform's decline had been underway for years.

Launched in November 2005 — a year before AWS's core cloud infrastructure services — Mechanical Turk was one of the first online marketplaces where people were paid small sums to perform tasks that resisted full automation: completing CAPTCHA challenges, identifying sentiment in sentences, and later, annotating images and text for AI training. The name referenced an 18th-century chess-playing machine that was secretly operated by a human hidden inside — an irony that became increasingly pointed as the platform's relationship with AI grew more complicated.

The AI data pipeline that ran on pennies

At its peak, Mechanical Turk was the hidden workforce behind countless AI products. Companies used it to label training data for computer vision models, clean text datasets for natural language processing, and validate search results. In 2018, AWS formally integrated the platform with SageMaker, its machine learning service, positioning Turk as the human-in-the-loop for neural network training.

But the economics that made Turk attractive — tasks paying cents per assignment — also made it vulnerable. A 2023 analysis found that between 33 percent and 46 percent of workers on the platform were using large language models to complete their tasks, creating a snake-eating-its-own-tail dynamic where AI was being used to fake the human labor that was supposed to train AI.

The quality concerns, combined with Amazon's development of SageMaker GroundTruth — a more automated data labeling service — made Turk's continued relevance questionable. AWS has been closing worker accounts without detailed explanation, according to posts on the Mechanical Turk subreddit, further shrinking the available labor pool.

What replaces the Turk

Amazon now directs customers toward SageMaker GroundTruth and third-party crowdsourcing services for data annotation needs. The shift reflects a broader industry move away from micro-task marketplaces toward managed labeling services that combine automated pre-labeling with human review — a model that offers faster turnaround and more consistent quality.

The retirement also closes a chapter on one of the gig economy's earliest platforms, predating Fiverr and Freelancer by years. Mechanical Turk was never a major revenue driver for Amazon — the company does not disclose its financials separately — but it played an outsized role in the development of modern AI systems by providing cheap, scalable human judgment.

For the AI industry, the platform's sunset raises a question that the original Mechanical Turk hoax would have appreciated: if the humans are being replaced by the very AI they were hired to train, what was the point of the marketplace in the first place? Amazon's answer, it seems, is to move on.

This article is for informational purposes only and does not constitute investment advice.