Amazon Mechanical Turk Is Closing to New Customers on July 30 - The 'Human-Powered' Marketplace That Fed AI Winds Down

Amazon Mechanical Turk, Amazon’s crowdsourcing marketplace, will stop accepting new customers on July 30, 2026. The official site now carries a notice stating: “Amazon Mechanical Turk will be closed to new customers, effective July 30, 2026. Existing users will not be impacted by this change.”¹

Launched in November 2005, Mechanical Turk is a marketplace where requesters post small tasks and a distributed pool of workers completes them for pay⁴. For more than two decades it served as the mechanism for having humans handle work that machines could not, and in particular it quietly underpinned the creation of AI training data. Now that platform is entering wind-down mode just as AI itself is advancing rapidly.

Officially a “Service in Maintenance”

AWS’s official “Services in Maintenance” documentation now lists “Amazon SageMaker AI – Mechanical Turk” with an announcement date of June 30, 2026². Services on this list no longer accept new onboarding; existing customers can continue to use them with ongoing support, but no new functionality will be added. AWS also recommends that customers explore alternatives — in practice, this is the stage before retirement².

In statements to the press, AWS described the move as the result of “careful consideration” and said it “continues to invest in security and availability improvements for Mechanical Turk, but we do not plan to introduce new features.”³⁴ The company has not explained why it chose to stop accepting new customers at this point⁴.

The Original “Artificial Artificial Intelligence”

The name Mechanical Turk comes from an 18th-century chess-playing “machine” that was in fact operated by a human chess player hidden inside³. The service productized the same idea: work that machines could not yet do would be handled, behind the scenes, by people. Requesters posted tasks and workers around the world completed them for small payments⁴. In the early years, typical jobs included solving CAPTCHAs and judging the sentiment of a sentence — tasks automation could not yet handle³.

The machine learning boom marked a turning point. Beginning around 2018, Amazon repositioned Mechanical Turk as a way for companies to have humans annotate data used to train neural networks, integrating it with its SageMaker machine learning platform³⁴. Humans tagged what appeared in images, and that ground-truth data trained AI systems — foundational work for the AI industry.

AI Ended Up on the Other Side of the Tasks

Ironically, it was AI itself that undermined that role. A 2023 analysis estimated that between 33% and 46% of workers on the platform were using large language models (LLMs) to complete their tasks³. On a platform whose value proposition was collecting human judgment, much of the work was being delegated to AI — eroding the premise of “distinctly human data.”

At the same time, many of Mechanical Turk’s typical tasks — text classification, sentiment judgments, simple image descriptions — are now things current LLMs can do directly. The demand structure for a human micro-task marketplace appears to have shifted substantially over the past few years.

Notably, this is not just about Mechanical Turk. The same AWS maintenance list includes “Amazon SageMaker AI – Ground Truth,” the data-labeling service, and “Amazon SageMaker AI – A2I (Augmented AI),” which routes AI decisions to human reviewers, both with the same June 30, 2026 date². A whole tier of human-annotation and human-review services is winding down together, suggesting AWS is restructuring this part of its portfolio.

What It Means for Current Users and Researchers

Existing customers can continue using the service as before after July 30¹. But with no new features planned³ and AWS recommending alternatives², organizations that have built Mechanical Turk into their workflows should start thinking about a medium-term migration path.

The most affected groups are development teams that relied on human judgment for AI model evaluation and data creation, and researchers who used Mechanical Turk to recruit participants for surveys and behavioral experiments. Academic research in particular has long treated it as a standard participant pool, so losing the ability to create new accounts is not a trivial change. Meanwhile, demand for human review of AI output has not gone away — human oversight remains important for problems like hallucination. How the supply of human-generated data shifts toward specialized vendors and in-house teams will be the thing to watch.

The original notice is available on the Mechanical Turk official site, and the full list of services in maintenance is in AWS’s official documentation.