The Amazon Mechanical Turk closure process begins on 30 July 2026, when the platform will stop accepting new customers, according to an official announcement on the service’s website. Existing users may carry on as normal, but Amazon Web Services has confirmed it will introduce no new features, a posture the company describes as the result of ‘careful consideration.’
AWS has formally classified Mechanical Turk as a service ‘moving to maintenance,’ a category distinct from the more terminal ‘sunset’ designation it applies to services with fixed end-of-operations dates. The platform is not dead yet, but it is no longer being built.
Twenty Years of Tasks Nobody Else Wanted
Mechanical Turk launched in 2005 as a marketplace for what Amazon called ‘human intelligence tasks’: small, repetitive jobs that computers of the era could not reliably handle. Workers, known informally as Turkers, were paid fractions of a dollar to complete CAPTCHA challenges, categorise images, or identify the emotional tone of a sentence.
The platform generated genuine controversy throughout its life. Critics argued the pay rates amounted to sub-minimum-wage piece work, and labour ethicists spent years debating whether crowdsourcing platforms owed their workforces the same protections as conventional employees. Mechanical Turk also surfaced, if briefly, in the early stages of the Facebook-Cambridge Analytica scandal, where crowdsourced tasks played a role in the data collection chain.
From 2018, Amazon repositioned the service more explicitly as an AI infrastructure tool, billing it as a way for companies to annotate training data for neural networks as part of its SageMaker suite. The pitch was straightforward: human reviewers could label the data that machine-learning models needed to improve. The Mechanical Turk homepage now points customers towards Amazon SageMaker Ground Truth, which it describes as a service to ‘easily build and manage your own data labeling workflows and workforce,’ and Ground Truth Plus, a ‘turnkey data labeling service that provides an expert workforce and manages it on your behalf.’
The Amazon Mechanical Turk Closure and the LLM Paradox
The deeper irony is that the technology Mechanical Turk helped build eventually hollowed it out. A 2023 study by researchers at École Polytechnique Fédérale de Lausanne (EPFL) estimated that between 33% and 46% of workers on the platform were using large language models to complete their assignments.
The researchers reached that estimate through a combination of keystroke detection and synthetic text classification, applied to an abstract summarisation task drawn from prior academic literature. They fine-tuned the e5-base language model to distinguish between summaries written by humans before the experiment and those generated by prompting ChatGPT with the original task instructions, as DeepLearning.AI’s The Batch reported. The analysis covered 46 summaries from 44 workers.
The finding raised two problems simultaneously. If a substantial portion of the workforce was outsourcing its work to LLMs, the human-generated training data that companies believed they were purchasing was partly machine-generated after all. And if LLMs could complete the tasks well enough to pass review, the case for paying humans to do them at scale became harder to make.
That loop has a certain historical symmetry. The original 18th-century Mechanical Turk was itself a deception: a chess-playing automaton that concealed a human operator inside its cabinet, the machine performing as a machine while a person did the actual thinking. Amazon’s version ran the trick in reverse, presenting human labour under a technology brand, and eventually found that the humans had begun outsourcing their own role to software.
The name, it turns out, was always the most honest thing about it.
Online commentary following the announcement has been blunt. One Reddit user argued the platform had effectively died ‘years ago,’ driven out by bots and fraud, and predicted that whoever controls the server budget at Amazon will eventually decide the remaining infrastructure is not worth maintaining. AWS’s ‘maintenance mode’ classification does nothing to contradict that timeline. The question is not whether Mechanical Turk will be fully decommissioned, but when someone at the company decides the servers are no longer worth the electricity bill.
