What happened
At its peak, more than 40,000 people in Nairobi were writing essays for university students overseas. The New York Times, reporting from Nairobi, has documented how that industry shrank to almost nothing within a few years of ChatGPT’s release, in an account summarised by The Next Web and Business Standard on 6 September.
The reporting, by Adam Satariano and Paul Mozur with Edwin Okoth, follows Teresios Bundi, who moved from a farming village to Nairobi in 2011, finished a public health degree in 2015 and then spent twelve years ghost-writing. He produced more than 2,500 essays at $40 to $70 each — several times what public health work paid locally. He has since closed the business and works for a German development agency helping young Kenyans find work in the digital economy.
Writers who remain report monthly earnings of $500 to $800, down from $900 to $1,200. One operator who employed 100 writers has shut down.
The timing is the story
Kenya adopted a ten-year national digital plan in 2022 that aimed to move its graduates into online outsourcing work. OpenAI released ChatGPT in November of the same year.
The country produces more than 100,000 university graduates a year, youth unemployment runs above 25%, and roughly 80% of work in the economy is informal. Online freelancing was the plan for absorbing educated young people who could not find formal jobs at home, and the essay trade — legally grey, socially unadmitted, but well paid by local standards — was one of the largest employers within it.
What makes this case unusual is that the substitution was total rather than partial. The customer was a student who wanted a passable essay by a deadline. A chatbot supplied exactly that at zero marginal cost, and the question the reporting records students asking — why pay a freelancer in Kenya when ChatGPT will do it — has no good answer.
What is left
The work that survives is downstream of the model rather than in competition with it. Writers described editing AI-written papers so they would not trip the detection software universities now run.
That is a smaller, worse-paid and more precarious version of the same job, and it exists only for as long as detection tools stay imperfect and universities keep paying for them.
The reporting also notes a broader trend line: AI systems’ completion rate on a measured set of tasks rose from 2.5% in October 2025 to 16% in July 2026. Transcription work in Kenya, a related sector, began declining to automation as early as 2017.
Why it matters beyond Kenya
Most labour-market arguments about AI are forecasts. This is a completed case with a named country, a dated national plan and a countable workforce, and it points at a specific vulnerability: economies that positioned themselves as the low-cost supplier of routine cognitive work sold into rich countries have the least insulation when that work becomes free.
The thing to watch is whether Kenya’s digital strategy is revised, and what replaces essay work in it. John Tanui, a principal secretary, continues to argue the country can be an outsourcing hub. The industry that was supposed to demonstrate it is gone.
