AI Wiped Out Kenya's Essay-Writing Industry, Erasing 40,000 Jobs in Two Years

The most concrete accounting yet of AI’s labor impact came out Friday, and it should end the argument about whether AI job displacement is hypothetical. The New York Times’ report on Kenya’s essay-writing industry documents how an estimated 40,000 writers in Nairobi, the “shadow scholars” who produced college essays for students in the US and UK, watched their industry shrink to almost nothing within roughly two years of ChatGPT’s release. Rates fell, assignments dwindled, and livelihoods that required degrees disappeared.

What the Collapse Actually Looked Like

The mechanics matter for understanding what AI displacement looks like at ground level. This was skilled work: the writers held degrees, researched, and wrote to deadline for essay mills serving foreign students. ChatGPT did not take their clients directly at first; it took their clients’ need for them, because the end customer could generate the essay themselves. Work thinned, per-piece rates collapsed, and the middle of the market vanished, leaving only the low-margin residue that AI could not yet do. The Times profiles writers like Teresios Bundi, whose income was essentially erased. This is the displacement pattern to internalize: AI eliminates the middle of a task market first, compressing rates for everything that remains.

Why This Case Is the Cleanest Signal We Have

Most AI labor-impact analysis drowns in projection. The Kenya case is unusually clean evidence: a discrete, measurable workforce, a precise shock date (ChatGPT’s release), and a visible before-and-after, corroborated by prior documentation of the industry in The Guardian and the documentary The Shadow Scholars. The uncomfortable finding is the speed: two years from thriving industry to near-zero, with no transition support, in a market where those workers had no regulatory protection.

For anyone whose work involves producing text, code, or analysis on demand, the lesson is not that AI will eliminate the profession. It is that the commodity layer of every knowledge profession disappears first and fastest, and the durable value sits in the parts a model cannot do: accountability, relationships, judgment under ambiguity, and verification. The Kenyan writers’ mistake was not their skill level; it was selling exactly the layer that got automated.