Kenya's Essay-Writing Industry Employed 40,000 People, and AI Ended It in Two Years
A New York Times report traces the collapse of Nairobi's contract essay-writing trade after ChatGPT. Surviving writers earn $500 to $800 a month, down from $900 to $1,200, mostly editing AI text to evade detectors.
The New York Times reported over the weekend on an industry that has almost completely disappeared: the contract essay-writing trade in Nairobi. At its peak it paid as many as 40,000 people to write coursework for university students in the United States and Europe. Within roughly two years of ChatGPT’s launch, most of it was gone.
The mechanism is not complicated, which is what makes it worth reading. Students who had been paying $40 to $70 for an essay ran the same request through a free chatbot and got something comparable in seconds. The Times profiles Teresios Bundi, who moved to Nairobi from a farming village in 2011, finished a public health degree in 2015, and then spent twelve years ghost-writing more than 2,500 essays, work that paid at least five times what public health jobs offered him locally. Writers who are still working report monthly income of $500 to $800, down from $900 to $1,200. The remaining jobs are mostly a new category: editing AI-generated papers so they do not trip the detection software universities now run.
What this case shows that the surveys do not. Most reporting about AI and employment deals in percentages and forecasts. This is one of the first industries where the before and after are both visible and the timeline is short enough to see clearly. Three details stand out. The work was outsourced knowledge work, which is exactly the category economists flagged as most exposed. It vanished quickly rather than eroding, because the substitute was not slightly cheaper but effectively free. And the jobs that survived did not survive by being better than AI, they survived by being about AI, cleaning up its output for a fee. There is also an obvious moral complication worth naming rather than tiptoeing around: this was an industry built on academic dishonesty, and losing it is not straightforwardly a loss. That does not change what it meant to the people whose rent it paid, and both things can be true.
What this means for you. If your work involves producing text to a specification for someone who does not much care who wrote it, this is the clearest available picture of what that exposure looks like in practice, and it is not a reason for panic so much as a reason to be honest about where your value sits. The parts of the Nairobi trade that survived were the ones requiring judgement about a specific institution’s rules, and that is the pattern to notice: what held up was context and accountability, not writing ability. If you manage or teach, it is also a reminder that detection software has already created its own small industry devoted to defeating it, which is worth factoring into any policy built on the assumption that detectors work.
Sources
Source: https://ground.news/article/for-years-kenyans-did-college-student-chores-then-came-the-ai_8f2998
Meta's Answer to Wasted GPU Time: Ask a Model Which Experiment to Run
Researchers from Meta FAIR, Oxford and UCL introduced AI Research Preference Models, which rank unexecuted machine learning experiments against each other rather than predicting scores, to decide where to spend compute.
