Mechanical Turk Is Closed: What It Means for AI Workers
Amazon ended the marketplace behind much of AI's human labeling work on September 30, 2026
Short Answer
Amazon has closed Mechanical Turk. Its notice on the site reads, "Following an assessment, we've made the decision to close AWS Mechanical Turk, effective September 30, 2026." As of this writing that date has passed. Amazon had already stopped accepting new customers, reported as July 2026, and has given no reason beyond a general assessment of its services. For the people who did tasks on the platform, the closure ends a source of paid work that fed data labeling for AI systems. This page explains what the service was, what Amazon has and has not said, and what the closure means for workers, researchers and requesters.
We wrote this from the news reports linked below. We did not interview workers, and where a figure comes from a single source we say so. Amazon's own FAQ is the authority on wind-down dates, so check it for anything time-sensitive.
What Mechanical Turk was
Mechanical Turk is a marketplace for small online tasks. A company or researcher, called a requester, posts a task, such as labeling an image, transcribing a clip or answering a survey. A worker completes it for a small payment. The work is called a human intelligence task. The model made it easy to buy human judgment in bulk, which is why it became a standard source of labeled data for building machine-learning systems and a standard way to run online studies.
The name is the point of an old irony. The original Mechanical Turk was an eighteenth-century chess-playing automaton with a person hidden inside. Amazon's service reversed the idea in plain view: a buyer sees a machine-like interface, and behind it are people doing the work. That is also why the platform is a useful example for this site's subject. The AI systems people use are built, in part, on human labor that the end product rarely credits.
What Amazon has said, and what it has not
The notice on the Mechanical Turk homepage states the September 30, 2026 closure and adds, "We regularly evaluate our programs, tools, and services and make adjustments based on those assessments." The Next Web, citing CNBC's reporting, says the closure banner was reported on August 25, 2026, and that Amazon had stopped accepting new customers the month before. It quotes an Amazon spokesperson, Montana MacLachlan, saying the company "remains focused on delivering the best possible experience for all its Amazon and AWS customers." According to that report, Amazon has published no other statement and the FAQ is the only guidance it has offered.
Several things are therefore unknown from the public record we found:
- The reason. Amazon cites an assessment and nothing more. Some coverage notes that Amazon also runs a competing data-labeling service, SageMaker Ground Truth. That is context, not an explanation, and we are not claiming it is the cause.
- How many people used it. The most-quoted figure, more than 500,000 workers across 190 countries, comes from Amazon's product documentation and, per the report, is not a recent count. It tells you the scale the service advertised, not how many workers were active this year.
- What happens to workers' pending pay. Reports describe a requester wind-down with approval and bonus windows and a period of transaction history access, but different outlets give different dates. We are not repeating a specific wind-down date; read Amazon's FAQ.
What the people affected have said
The Next Web quotes Krista Pawloski of Turkopticon, a worker-run organization that tracks requesters, saying some requesters, including insurance and travel companies, had about 35 days from the announcement to find alternatives. That is one person's account of requester timelines, and we have not verified it against Amazon's schedule. It is useful mainly as a sign of how short the notice was for the businesses that depended on the platform.
Tech Policy Press titled its own piece on the subject "Mechanical Turk is Closing. The Workers Who Built AI Are Still Here." We read that title as the argument that matters: the platform is closing, but the demand for human data work is not. Companies that built models on crowd labor still need people to check outputs, label edge cases and rate responses. What changes is who employs those people, on what terms and with what visibility.
Why the closure matters for the AI conversation
There are three reasons we treat this as an AI-accountability story rather than a product shutdown.
- It shows where the labor sits. Public talk about AI emphasizes models and chips. A marketplace like this one is a record of the human task layer underneath, and its closure removes one of the few public places that layer was visible.
- It moves the work, it does not end it. The need for labeling and evaluation continues. The risk for workers is that work moves to vendors and private contracts where pay rates, conditions and worker identities are harder for outsiders to see.
- It affects researchers too. Academic surveys and behavior studies relied on cheap access to participants. Researchers now have to move their studies elsewhere, and results from different platforms are not always comparable.
We are deliberately not putting a wage figure on any of this. The sources we read did not give a reliable pay rate, and a number without a source would be the kind of claim this site avoids.
If you are a worker, a researcher or a requester
- Workers: read Amazon's FAQ for the wind-down terms, download and keep your own records of completed tasks and earnings, and check whether a task you did is still waiting on approval. Worker-run groups such as Turkopticon track platforms and requesters.
- Researchers: note in your methods which platform your participants came from, and avoid pooling data from different recruitment sources without checking whether the populations are comparable.
- Requesters: the closure also removes the Mechanical Turk worker type from related workflows, according to coverage of Amazon's FAQ, so check any pipeline that depends on it.
For the wider picture of how AI shifts work, see our pages on AI layoffs and whether AI will replace your job.
How to read the coverage of this closure
Reports on the closure differ on small points, and it helps to know which kind of claim you are reading. A statement quoted from Amazon's own notice, such as the September 30, 2026 date, is firsthand. A figure taken from Amazon's documentation, such as the worker count, describes what Amazon advertised, not necessarily what is true today. A comment from a worker organization is an informed account from people close to the work, but it is still one account. And a date reported by a trade outlet for a requester deadline may or may not match Amazon's FAQ. When two outlets disagree, we have left the figure out rather than pick one.
Sources and method
This is an analysis of published reports. We did not test the platform or speak to workers.
- The Next Web, Amazon is closing Mechanical Turk, the human workforce it sold as AI (closure notice, spokesperson quote, Turkopticon comment, documentation figure)
- Tech Policy Press, Mechanical Turk is Closing. The Workers Who Built AI Are Still Here (title only reviewed for this page)
- AI Weekly, Amazon to Close Mechanical Turk on Sept. 30 After Two Decades
Frequently asked questions
▸ Is Amazon Mechanical Turk shutting down?
▸ Why is Amazon closing Mechanical Turk?
▸ How many people worked on Mechanical Turk?
▸ What happens to the AI data-labeling work now?
Latest related briefings
AI Layoffs Boomerang: Companies Now Rehiring Workers They Let Go
AI layoffs are coming full circle as companies rehire workers they replaced with automation. What does this mean for your job security?
Read analysis JOBS LABORAI Layoffs Threaten Jobs as Consulting Salaries Soar
As AI layoffs loom for many, top consulting firms hike starting salaries to $200K, widening the gap between high earners and workers facing automation.
Read analysis JOBS LABORAI Layoffs in 2026: 15 Companies Cut Jobs, Workers Hit Hard
In 2026, at least 15 major companies have blamed AI for job cuts, directly impacting thousands of workers and raising new concerns about job security.
Read analysis