Amazon's Mechanical Turk was once one of the most quietly influential platforms on the internet. Since 2005, it has connected businesses with a global pool of workers who completed small digital tasks for tiny payments: tagging images, transcribing audio, completing surveys, checking the tone of a sentence. The tasks were simple. The payments were often just a few cents. But at scale, it worked.
The announcement that Mechanical Turk will close to new customers on July 30, 2026 is not a surprise to anyone who has watched the platform closely. AWS added it to its "Services in Maintenance" list, which is the company's internal signal for products it plans to retire. Existing users keep access for now, but Amazon has confirmed there will be no new features, and the service will stop accepting jobs for its AI data-labelling tool, SageMaker, as well as everything else.
The real story here is how Mechanical Turk became a victim of the thing it helped build. Its second wind came in 2018 when Amazon positioned it as a way to feed human-labelled data into AI training pipelines. Companies building AI systems needed humans to look at thousands of examples and mark them up: this photo contains a dog, this review is negative, this sentence is rude. Mechanical Turk was cheap and fast.
Then AI got good enough to undercut that work. A 2023 analysis found that between 33% and 46% of Mechanical Turk workers were using AI tools to complete their tasks. People hired to provide human judgement were outsourcing to the very systems that judgement was supposed to train. That compromised the value of the data, and it made it hard for anyone buying that data to trust what they were getting.
At the same time, quality issues that had plagued the platform for years worsened. Workers' accounts were being closed with little notice or explanation. Fraud and bots were common. Researchers who had used it for surveys found it increasingly unreliable. The market responded by moving to better-run alternatives.
The data annotation industry, the broader market that Mechanical Turk once helped define, is actually growing fast. The global market for data labelling tools was worth around $3.2 billion in 2025 and is projected to reach $34 billion by 2035. That money is flowing toward more structured services: platforms like Scale AI, Appen, Surge AI, and Labelbox, which combine human reviewers with stricter quality controls and, increasingly, AI-assisted pre-labelling that sends only genuinely ambiguous cases to humans.
Amazon itself has a direct replacement. SageMaker Ground Truth is the company's managed data-labelling service, and it can route tasks to private teams, outside vendors, or third-party specialists listed in the AWS Marketplace. The move away from Mechanical Turk is in part a push toward that cleaner, more controlled offering.
For most business operators, this is a background story rather than an operational alert. If you have never used Mechanical Turk directly, nothing changes for you. But the broader signal matters. The era of cheap, low-oversight crowdsourced labour for AI tasks is ending. The companies that still rely on human-in-the-loop data review, whether for content moderation, quality checking, or AI training, will increasingly need to work with structured professional services rather than open marketplaces. Those services cost more per task, but they come with accountability and data governance that open platforms never could provide.
Mechanical Turk's slow fade is also a useful reminder about what happens when a platform degrades without a clear plan. Amazon let the quality and trust decay over years before making any formal announcement. Businesses that built workflows on top of it had no warning. When you depend on a third-party platform for any operational task, the question is always what happens if it stops. In this case, the answer was a slow decline that left users to figure it out themselves.