Analysis
Amazon will close Mechanical Turk entirely on September 30, 2026, CNBC reported, ending one of the internet's longest-running crowdsourced labor platforms after Amazon stopped accepting new customers on it in July. Existing customers were able to keep using the service in a maintenance-only mode between the two dates, with no new features added since the wind-down began.
What Mechanical Turk Was
Launched in 2005, Mechanical Turk matched businesses with a distributed workforce willing to complete small digital jobs -- officially called Human Intelligence Tasks, or HITs -- for pay typically measured in cents rather than dollars: labeling images, transcribing audio, answering short surveys, moderating content. Jeff Bezos gave the platform its enduring nickname at launch, calling it "artificial artificial intelligence," a description that captured the model precisely -- humans standing in for tasks that, in 2005, no algorithm could reliably perform on its own.
“## Why It's Dying Now That framing is exactly what makes the shutdown notable two decades later.”
Why It's Dying Now
That framing is exactly what makes the shutdown notable two decades later. Krista Pawloski, an organizer with Turkopticon, a data-worker advocacy group, told CNBC the platform had been in visible decline for years before this final closure. The core reason: the AI industry Mechanical Turk helped bootstrap has since built purpose-made replacements. Companies like Scale AI, Mercor and Prolific now run more structured, better-paid data-labeling and human-feedback operations specifically built around training frontier AI models -- work that requires more specialized skill and higher-quality output than Mechanical Turk's flat-rate microtask model was ever designed to deliver.
- Amazon Mechanical Turk -- launched 2005, stopped new signups July 30, 2026, fully closes September 30, 2026
- Scale AI -- structured data-labeling and RLHF platform serving major AI labs, one of Mechanical Turk's direct successors
- Mercor, Prolific -- newer entrants recruiting and vetting human workers specifically for AI model training and evaluation
The Counterweight
The risk in reading this shutdown as a clean AI-replaces-humans story is that it overstates how automated the underlying work has actually become -- Scale AI, Mercor and Prolific still rely on large human workforces, just organized differently and paid more, which is a labor-market shift, not a labor-elimination one.
The Irony Worth Naming
There's a real irony in the timing that's easy to miss: the AI boom Mechanical Turk's human workforce helped train -- image labeling, content moderation, early machine-learning datasets all ran through MTurk workers for years -- ultimately built AI models capable enough to reduce demand for the exact platform that helped train them, while simultaneously spawning a new, more specialized layer of human-labor companies to do adjacent work AI still can't handle alone. Mechanical Turk didn't lose to AI directly; it lost to better-organized, more purpose-built human-labor marketplaces that AI's rise created demand for.
What to Watch
The open question for the broader data-labeling economy is whether Scale AI, Mercor and Prolific's more structured, higher-paid model represents a permanent improvement in how AI training labor gets organized, or just the current phase before those companies face their own version of Mechanical Turk's fate as AI models increasingly self-improve with less human feedback per unit of capability gained.