AI Erased 40,000 Online Writing Jobs in Nairobi and What It Warns Every Gig Worker

An industry that supported as many as 40,000 people in Nairobi has been largely erased in about two years, according to New York Times reporting summarised by Techmeme on 6 September. The work was academic ghostwriting, producing essays and assignments for students in wealthier countries, and it paid well enough to sustain educated young Kenyans who had few other routes into professional English language work. Generative AI removed the demand almost entirely. At Metaintro, we treat this as the clearest case study available of what happens when automation hits a job category that exists only on a platform, and the mechanics of it apply well beyond Kenya.

What Exactly Collapsed in Nairobi?

A large, organised, and unusually well paid corner of the online gig economy. Kenyan reporting on the industry puts the scale at more than 40,000 people facing job losses as academic writing work disappeared. These were not casual side earners. Many were graduates using strong written English to earn considerably more than local entry level salaries in other fields.

The speed is the part that should alarm anyone in comparable work. The industry did not decline gradually over a decade in the way that manufacturing employment often does. It contracted to near nothing within roughly two years of consumer generative AI becoming widely available, because the substitute was immediate, free at the point of use and good enough for the buyer's purpose on day one.

What remains is instructive and grim. Writers report that the surviving work is largely editing AI generated papers so that they will not be flagged by the detection software universities now run, a practice they describe as humanising. The job did not vanish so much as invert. Where the work was once producing the text, it is now cleaning up after a machine that produces it faster, and that residual task pays less and requires the same person to compete with the tool that displaced them.

How Far Did Earnings Actually Fall?

Far enough to change what the work is for. Reporting on the industry describes monthly earnings for writers falling from roughly 900 to 1,200 dollars down to somewhere in the range of 500 to 800 dollars. Those figures should be read as reported ranges rather than as a survey, but the direction and rough magnitude are consistent across accounts.

The absolute numbers matter in context. In a market where formal graduate employment is scarce, an income above 900 dollars a month placed these workers well into the local middle class, funding family obligations, rent in Nairobi and further study. Halving that does not simply reduce discretionary spending. It removes the economic basis on which people made long term commitments, which is why the individual stories in the coverage centre on obligations that cannot be renegotiated as quickly as the income fell.

There is also a hidden cost that earnings figures do not capture. The remaining work is more precarious and less repeatable, because humanising an AI draft is a task the detection tools are actively trying to defeat and the platforms are under pressure to prohibit. Anyone building a plan around it is building on a category that is being deliberately closed, which is a materially worse position than simply being paid less for the same stable task.

Why Did This Category Break So Completely?

Because it had every characteristic that makes work vulnerable to a general purpose tool. The output was text, unregulated, delivered digitally, evaluated by a buyer who wanted speed and price over provenance, and mediated entirely through a platform rather than a relationship. Remove any one of those and the collapse would have been slower.

Compare it with work that has survived similar pressure. Nursing, electrical work and machine maintenance all involve physical presence, licensing and liability. Even in white collar categories, the roles that have held up tend to carry accountability that a buyer cannot outsource to a tool, which is why customer service exposure to AI differs so much from exposure in regulated professional advice.

The pattern is repeating in other markets that built employment on the same foundations. AI has forced the Philippines to rethink a call centre boom that employs vastly more people than Kenyan academic writing ever did, and business process outsourcing jobs in Krakow have been hit hard. Each of those economies encouraged young workers into English language digital services precisely because the work was accessible and scalable, which turns out to be the same reason it automates cleanly.

Which Online Gig Categories Look Most Exposed Next?

The ones where the buyer never meets the worker. Generic copywriting, basic translation, transcription, simple graphic production, elementary data entry and template driven research all share the essay writing profile of digital delivery, no licensing and price led purchasing. Fiverr losing a fifth of its buyers as AI absorbed small freelance jobs is the platform level measurement of that shift.

Platform dependence adds a second, separate risk that workers routinely underestimate. Even where demand survives, the platform itself may not. Amazon closing a gig work platform after twenty one years removed an income source from people whose skills were still valuable, and gig workers cut off with no appeal route face the same loss without any change in the underlying market at all.

Some digital gig work is moving in the opposite direction, which is worth knowing before concluding the whole category is finished. Indian gig workers earning money training robots and factory workers being paid to teach machines their own trade describe demand created by automation rather than destroyed by it. That work also has a limited life by design, since the point of the task is to make itself unnecessary, and treating it as a destination rather than a bridge is the mistake to avoid.

How Do You Tell if Your Own Category Is Next?

Score it against five questions, honestly. Is the output digital and deliverable without you being present. Is it unregulated, meaning no licence or professional body stands between the buyer and a substitute. Is the buyer choosing mainly on price and speed. Is the relationship mediated by a platform rather than held directly. And could a competent generalist with a good tool produce something the buyer would accept.

Five yes answers describe the Nairobi essay trade almost exactly. Four is still a serious warning. Two or fewer, and the immediate risk is much lower, though it does not vanish. The value of running this deliberately is that it separates anxiety from exposure, and most people who feel threatened by AI are either considerably safer than they fear or exposed in a different part of their work than the part they are worrying about.

The macro picture supports taking the exercise seriously rather than panicking. International Labour Organization work on non standard employment sets out how limited the protections are across platform and contract work worldwide, and World Bank research on digital jobs has documented how many developing economies have staked youth employment strategies on precisely these categories. Kenya's own digital skills push runs through bodies including the ICT Authority, which makes the policy exposure national rather than individual.

The honest conclusion is that the countries which did the most to move young workers into online services were the most exposed when the tools arrived. That is not an argument against digital work. It is an argument for treating any single category as a stage rather than a destination, and for spending some of your best earning months building the thing that comes after it.

What Should You Do if Your Income Depends on a Platform?

Diversify the buyer before the tool arrives, not after. The single most damaging feature of the Nairobi collapse was concentration. Workers had one category of task, one type of client and often one platform. When the category went, everything went at once. Someone earning across two unrelated categories absorbs the same shock as a large pay cut rather than as a total loss.

Move toward work the buyer cannot verify without you. That means anything requiring local knowledge, a licence, physical presence, ongoing accountability or a relationship that survives the individual transaction. Practically, a writer who becomes the named specialist for a small number of recurring clients is in a far stronger position than one who competes anonymously on price for one off jobs, even when the underlying craft is identical.

Build the transition while the current income still exists. Upskilling for AI without going back to school and free and low cost routes to skills employers trust are both realistic from a Nairobi bedroom with a laptop. So is moving up the same value chain rather than out of it, since ten freelance skills and their current hourly ranges show which digital categories still command real rates.

Is There Any Route Back for the Writers Themselves?

There is, though it is narrower than the sector they lost. The underlying capability is genuine, and it is not the ability to write essays. It is a graduate level command of written English, research discipline, the ability to work to a brief and a deadline, and in many cases subject knowledge across several academic fields. Those are inputs to a number of jobs that are not disappearing.

The most direct adjacencies are AI adjacent rather than AI proof. Model evaluation, red teaming, data annotation for language tasks, prompt and output quality review, and localisation all require exactly the judgment these workers already have, and they are being bought at volume. It is uncomfortable that the most available work involves improving the systems that removed the previous income, and workers being asked to train the AI that will replace them is a real and legitimate grievance. It is also, for many people, the fastest available bridge.

Domestic and regional demand is the other route, and it is underrated because it pays less in dollar terms while being considerably more durable. Kenya has an active digital work policy through programmes such as Ajira Digital, and labour market data from the national statistics office is the place to check where formal employment is actually growing. Communications, education support, research assistance and technical documentation for African employers do not evaporate when a Californian tool ships.

What Does This Mean for Workers Outside Kenya?

That the warning is about structure, not geography. Gizmodo's account of the collapse framed it as a story about a specific and legally grey trade, and that framing is exactly what makes it easy to dismiss. The relevant feature was never that the work was ghostwriting. It was that the work was remote, digital, unlicensed, price led and platform mediated.

Millions of people in wealthy countries do work with that profile too. Freelance content production, junior research, basic design, entry level analysis and routine administrative support all match it. International Labour Organization work on platform employment documents how thin the protections are for this group globally, and the absence of notice, severance or consultation is the reason a demand shift lands as an immediate income loss rather than a managed transition.

The practical conclusion is not that digital work is a mistake. It is that the specific combination of one skill, one buyer type and one platform is the risk, and it is a risk that can be reduced deliberately while things are still going well. Older workers holding an underrated edge in AI upskilling is a useful corrective to the assumption that adaptation belongs to the young, and the Nairobi writers who have moved fastest were generally those who started before the work ran out rather than after.

Related Articles

• Fiverr Lost a Fifth of Its Buyers as AI Swallowed Small Freelance Jobs
• AI Is Forcing the Philippines to Rethink Its Call-Center Job Boom
• AI Devastates Krakow BPO Jobs
• Amazon Shuts a Long Running Gig Work Platform and What Its Workers Do Next
• When Gig Platforms Cut You Off With No Appeal, What Deactivated Drivers Can Do
• How India's Gig Workers Are Earning Money Training the World's Robots
• Companies Are Asking Workers to Train the AI That Will Replace Them
• How to Upskill for AI Without Going Back to School
• Free and Low-Cost Ways to Learn AI Skills That Employers Trust
• Why the Gig Economy Needs a Portable Worker Passport
• Older Workers Have the AI Upskilling Edge Tech Companies Keep Missing

People Also Asked

Q: How many jobs did Kenya's essay writing industry lose?

A: Reporting puts the peak workforce at around 40,000 people in Nairobi, with Kenyan coverage describing more than 40,000 facing job losses as demand collapsed. The contraction happened within roughly two years of consumer generative AI becoming widely available, which is far faster than most automation driven job losses.

Q: What work is left for those writers now?

A: Mostly editing AI generated papers so they are not flagged by university detection software, a practice writers call humanising. It pays less than the original work and is inherently unstable, because both the detection tools and the platforms are actively working against it. More durable adjacencies include model evaluation, data annotation, localisation and quality review of AI output.

Q: Which other gig work is at similar risk?

A: Categories where the buyer never meets the worker and the output is digital text or simple graphics. Generic copywriting, basic translation, transcription, template design and elementary data entry share the profile. Work involving licensing, physical presence, local knowledge or ongoing accountability has held up considerably better against the same tools.

Looking for your next opportunity? Losing a category of work is not the same as losing your capability, and research discipline, written English and subject knowledge are inputs to far more jobs than most freelancers realise. Metaintro matches writers, researchers and remote professionals to roles and contracts where those skills are still bought directly. Create a free profile and find where your work transfers next.