Mechanical Turk was Amazon’s platform for paying people pennies to do small tasks that computers struggled with, from labeling data and transcribing text to survey work and research studies. Amazon now says it is shutting the service down on September 30. The comments landed on a simple explanation. MTurk was built for broad, low-skill human judgment at internet scale, and that is exactly the layer LLMs and bots are now good enough to attack from both sides. Buyers can automate many former MTurk jobs directly. Workers and intermediaries can also use AI to fake the human work, which makes the output hard to trust.
That trust problem came up more than nostalgia. People who had run real workloads through MTurk described it as usable only with heavy scaffolding around it. You needed duplicate submissions, consensus checks, worker filtering, and custom task management just to get acceptable accuracy. Paying more mostly made tasks complete faster, not better. Others said quality had already been degrading before the
LLM boom because workers were incentivized to race through microtasks and attention checks. Several commenters argued that this leaves little room for a horizontal marketplace of anonymous generalists. The remaining demand for human labor is shifting toward narrower, higher-trust work such as medical review, specialist labeling, moderation, and
model evaluation, where credentials, vetting, latency, and oversight matter more than raw task volume.
A few people with direct exposure said Amazon had effectively stopped caring long before the announcement. One large requester said customers were informed at the same time as workers and that the AWS manager responsible for
AMT had moved to
Bedrock and
SageMaker model evaluation years ago, leaving almost no team behind. That matched smaller signs of neglect like stale site updates and rough integrations. The strongest practical read was not that “human-in-the-loop” work is disappearing. It is being absorbed into vertical services, in-house pipelines, and AI-era QA systems that look nothing like the old open bazaar of penny tasks.