For hundreds of thousands of workers across Southeast Asia, rural India, and beyond, the routine was the same: 200 micro-tasks a day — labeling images, flagging offensive content, answering survey questions for a few cents each. Log into a portal that knows you by a worker ID, not a name. On September 30, 2026, that portal goes dark permanently. Amazon announced in late August 2026 that AWS Mechanical Turk — the crowdsourcing marketplace Jeff Bezos once called “artificial artificial intelligence” — is closing after a 21-year run. Workers looking to rebuild their workflows may find AI-powered websites a useful starting point.
At its peak, MTurk connected over 500,000 registered workers across 190 countries with companies that needed human judgment packaged as an API call. Credit where it’s due: without MTurk’s labeled images and annotated text, a lot of the machine learning breakthroughs of the 2010s simply don’t happen.
If you’re currently on the platform, here’s what the timeline looks like:
- Sign-ups ended July 30
- The platform closes September 30
- Requesters get until October 30 to approve or reject submitted work — unreviewed HITs auto-approve after that window
- Approved worker earnings pay out on the normal schedule, so verify your payment settings in the Worker Portal now
One cascade effect worth noting for AWS users: the MTurk worker type is also being removed from SageMaker Ground Truth and Amazon Augmented AI on the same date.
“The AI industry has shifted away from anonymous open-crowd annotation toward managed labeling services with quality controls, audit trails, and domain expertise.”
Why the Gig Ate Itself
Two forces killed MTurk simultaneously — one from above, one from within.
From above, the data-labeling market professionalized. Platforms like Scale AI, Labelbox, and Prolific offered stricter quality controls and specialized workforces. Managed labeling became the enterprise standard. MTurk was cheap but messy — a Craigslist in a market that had decided it wanted LinkedIn.
Then came the kill shot from within. A 2023 study by researchers at EPFL found that 33–46% of MTurk workers completing text-production tasks were using generative AI to do so. The service marketed as genuine human intelligence was quietly filling with LLM outputs. Computer scientist Jaron Lanier had already warned that MTurk was built to “turn people into software components.” The irony landed faster than anyone expected: the components started outsourcing to machines.
Advocacy group Turkopticon documented what workers were experiencing throughout — low wages, dispute processes weighted against them, and little meaningful recourse when tasks were rejected. The shutdown removes a flawed but real income source. Amazon isn’t replacing it with anything for those workers — unlike Meta’s bet on building a skilled trades army as part of its broader workforce investment strategy.
Where Things Land Now
The open crowd is out; the specialists are in — and the clock is running.
The industry is moving toward curated, closed data pipelines and specialist human evaluators, not open crowds. The rise of the humanoid robot further underscores how automation is reshaping labor roles across industries. If you’re a researcher, Prolific or Qualtrics panels are the most direct alternatives. If you’re running ML pipelines through SageMaker, rebuild around private workforces or managed vendors before September 30 — the MTurk worker type disappears from Ground Truth and A2I on that same day, and any pipeline depending on it breaks immediately. For Turkers: the platform had serious structural problems, but losing accessible, flexible income is still a real loss, whatever Amazon’s understated press notice implies.
