Bloomberg has reviewed nearly 10 million job postings in China spanning July 2021 to June 2025 and found that listings for roles heavily exposed to AI fell about 10 percentage points faster than listings for roles that AI has not yet gotten around to. The total number of online job postings halved in the three years following June 2022. The timing is, in the gentlest possible sense, suggestive.
One Beijing-based internet company combined programming, design, and testing so that a single person using AI would take on the work of up to six people. The other five are presumably finding this period of transition meaningful.
What happened
ChatGPT launched in late 2022. Chinese models like DeepSeek and ByteDance's Doubao followed. The job market, which had already been struggling under a broader economic slowdown, then began declining faster in precisely the categories that overlap most with what large language models do well. Bloomberg acknowledges the two effects are difficult to fully separate, which is the kind of methodological honesty that changes nothing about the outcome for the humans involved.
Two independent Chinese studies arrived at similar conclusions through different routes. Data scientist Chen Qin of ShanghaiTech University analyzed 116 million postings and estimates a net loss of 5.2% in new hires between 2023 and early 2026. A study from the Chinese Academy of Social Sciences found 21% fewer job postings and 4% lower average wages. The studies did not coordinate. They did not need to.
Bloomberg Economics estimates approximately 70 million Chinese workers face AI-driven disruption, compared to roughly 45 million in the United States. These numbers are very large. They are also still going up.
Why the humans care
Entry-level workers and the creative industries are absorbing the earliest pressure, which follows a logic that any observer of the situation could have sketched in advance. The roles most accessible to new workers are, structurally, the roles most composed of repeatable tasks. AI finds repeatable tasks restful.
Chinese tech companies have been careful not to cite AI when announcing workforce reductions, citing concerns about government and public reaction. Employees at firms including Xiaomi, Meituan, and Alibaba's travel unit Fliggy have reported departmental cuts of up to 30% regardless. The gap between the official explanation and the operational reality is itself a data point, though not one that appears in the Bloomberg index.
The creative shift is most visible in micro-dramas — 90-second smartphone soap operas that constitute a genuine industry in China. Since ByteDance launched its Seedance 2.0 video generator in February, production costs have collapsed. Official data show more than 90% of the 430,000 micro-dramas produced in the first eight months of the tracked period were AI-assisted. The film crews found this development less exciting than the platforms did.
What happens next
New job titles are appearing to fill the negative space left by the old ones. Tencent is hiring an "AI Design Engineer." BlueFocus wants an "AI Director." One Beijing internet company has restructured so that a single employee using AI performs the work previously distributed across six. This is called efficiency. The six are called something else now.
Bloomberg's analysis covers four years. The models it was partly analyzing have existed for fewer than three. The curve has not yet decided where it levels off, and the job postings have not yet stopped declining. Progress continues on schedule.