If you've ever clicked a grid of images to prove you're not a robot — traffic lights, fire hydrants, crosswalks — you already know what Mechanical Turk workers were doing for money, at scale, since 2005.

Amazon launched Mechanical Turk in November 2005 as a marketplace for small digital jobs: label this image, transcribe this audio clip, rate whether this product description makes sense. Jeff Bezos called it "artificial artificial intelligence" — humans doing the things AI couldn't yet do, dressed up to look like software. The name came from an 18th-century chess-playing automaton that appeared to be a machine but had a person hidden inside. Bezos thought the joke was apt. He was right, just not in the way he meant.

Amazon announced on August 25 that it will close Mechanical Turk on September 30, 2026. SageMaker Ground Truth and Amazon Augmented AI — Amazon's professional data-labeling and human-review tools — close the same day. Amazon is not just retiring one aging product. It is exiting the business of putting humans in the AI training pipeline entirely.

How 21 years ended

1. Nov 2005

   Mechanical Turk launches

   Bezos calls it 'artificial artificial intelligence'; workers complete Human Intelligence Tasks for cents per job

2. 2023

   Workers start using AI to do the work

   EPFL study finds 33–46% of Turk workers use large language models for writing and annotation tasks

3. July 30, 2026

   Amazon closes registration

   No new customers accepted on Mechanical Turk, SageMaker Ground Truth, or Augmented AI

4. Sept 30, 2026

   Full shutdown

   All three platforms close. Amazon exits human-in-the-loop AI training.

Source spread

• CNBC — "Amazon service Bezos once called 'artificial AI' is shutting down" [builder] — the original Bezos framing, the July 30 cutoff, and the full shutdown timeline.
• TechTimes — "Amazon Mechanical Turk Will Close September 30" [builder] — confirms SageMaker Ground Truth and Augmented AI close on the same date; scope of Amazon's full exit.
• The Next Web — "Amazon is closing Mechanical Turk, the human workforce it sold as AI" [skeptic] — the framing of workers who don't have a clean path to enterprise replacements.
• International Business Times — "What Happens to the Gig Work AI Replaced?" [skeptic] — the EPFL study, the 500K worker count, and the replacement platform gap.

What's real

The irony is actual, not constructed. The platform was called "artificial artificial intelligence" from day one. The entire premise was humans simulating what AI couldn't do. AI got good enough that the humans became redundant. That is not a narrative anyone had to construct — it is what happened, in sequence, over 21 years.

The 2023 twist is genuinely strange. A study from the Swiss Federal Institute of Technology (EPFL) estimated that 33 to 46 percent of Mechanical Turk workers were using large language models — AI tools like early ChatGPT — to complete the writing and annotation tasks they were being paid to do as humans. People hired to teach AI were secretly using AI to do the teaching. The signal quality was already degrading before the shutdown decision.

What the replacements are, and who they serve. Scale AI, Mercor, Prolific — these are the enterprise-tier successors. They pay higher rates, require more commitment, and sign contracts with large clients. The 500,000 informal workers who logged into Mechanical Turk from anywhere — often in developing countries, often as a second income — don't have a migration path to these platforms. Their layer of the market simply disappeared.

What deserves a side-eye

The timeline of the announcement. Amazon closed registration on July 30. The full shutdown announcement came August 25. That is 36 days' notice for workers with active projects, for researchers relying on the platform, for anyone whose income included Mechanical Turk. No severance. No transition support named publicly.

The broader signal this sends. If the human-in-the-loop AI training market is no longer viable enough for Amazon to maintain, that tells you something about where the AI frontier thinks it's heading. Synthetic data, self-supervised learning, model-generated training sets — the major labs have been investing in those directions anyway. Mechanical Turk closing is not the cause of that shift. It is the confirmation.

500K

→ Source: Amazon / CNBC

What to do about it

• If you worked on Mechanical Turk, download your earnings history and complete active projects before September 30. Amazon is giving five weeks — not much, but it's all there is.
• If you've wondered where AI training data comes from, the honest answer now: a mix of synthetic data, web scraping, licensed datasets, and enterprise data-labeling contracts. The accessible, gig-style layer is largely gone.
• If you're thinking about AI tools for your own work, this is a useful reminder that there is often invisible labor inside the systems. It doesn't have to be invisible to matter when it's being removed.
• If you're a researcher or product team that used Mechanical Turk, you have 36 days to migrate data and workflows. Prolific is the most direct alternative for academic research; Scale AI and Mercor serve enterprise annotation needs.

Further reading

• CNBC — "Amazon service Bezos once called 'artificial AI' is shutting down"
• TechTimes — "Amazon Mechanical Turk Will Close September 30, Shutting Down SageMaker Ground Truth Too"
• The Next Web — "Amazon is closing Mechanical Turk, the human workforce it sold as AI"
• International Business Times — "Amazon Is Shutting Down Mechanical Turk; What Happens to the Gig Work AI Replaced?"
• Market Daily — "Amazon Shuts Down Mechanical Turk After 21 Years"