5 Findings Behind 100 Million Recruitment Data on AI Employment

In the first five months of 2026, the technology industry led by the United States laid off more than 100,000 people, reaching the level of the whole last year.

AI has become the biggest single reason. The Challenger lay - off report in April shows that the proportion of lay - offs due to AI in the United States has reached 26%, ranking first among all reasons.

On the other side of the coin, there is a sharp increase in AI recruitment against the trend. The overall recruitment market in the United States is still sluggish, but in the first quarter of 2026, the number of AI positions increased by 36% year - on - year, with a total of more than 55,000 (Veritone).

In the same market, lay - offs and expansions are happening simultaneously.

A global executive survey by HBR at the beginning of 2026 revealed the underlying logic: Companies are now laying off employees for the potential of AI, not for its actual performance.

This is the situation in the United States. What about China?

Let's use 100 million recruitment data to find out.

The Tencent Research Institute, in collaboration with the Institute of Population and Labor Economics of the Chinese Academy of Social Sciences, collected approximately 100 million public job information from six major domestic recruitment platforms from Q1 2024 to Q2 2025. Through keyword matching and multi - dimensional text mining, all positions with AI requirements were screened out for statistical comparison and analysis, resulting in the "Big Data Research Report on New Trends in AI Careers".

The report was officially released on May 23, 2026, at the 2026 Social Science Forum on "The Disciplinary Construction and Talent Cultivation of Digital Economics of the Chinese School" at the University of Chinese Academy of Social Sciences.

Here are the five most important findings.

Finding 1: The penetration rate of AI positions is less than 2%, but don't be deceived by this number

Among the 100 million recruitment data in the entire market, what is the proportion of positions that clearly require AI skills?

1.6% - 1.92%.

Less than 2%. It seems small, but this "low" is not unique to China. The global data from the Stanford 2025 AI Index Report shows: 3.2% in Singapore, 2.0% in Luxembourg, and 1.8% in the United States. China is in a reasonable range.

However, 2% may only be the surface. The data of public positions can only count the requirements "written in the recruitment descriptions". A large number of enterprises have widely used AI tools in actual work, such as using large models to write copywriting, generate code, process data, etc., but many have not written these skills into the recruitment requirements.

A survey of 20,000 AI users in 10 markets by the Microsoft 2026 Work Trend Index found that 49% of Copilot conversations involve cognitive work such as analyzing information and solving problems. For the corresponding positions, AI skill requirements are hardly mentioned during recruitment.

The hidden AI penetration is far deeper than the explicit one.

Although these positions accounting for less than 2% have a huge effect on career stratification. The following four findings will repeatedly prove one thing: Although there are few AI positions, they are rewriting the rules in terms of skill thresholds, rank structures, and income ladders.

The rewriting of work is something that everyone is facing or will face.

There is also a notable inflection point: The proportion of AI positions continued to decline throughout 2024, but suddenly rebounded to 1.8% in Q1 2025, almost returning to the highest level two years ago. The time node highly coincides with the release of DeepSeek. A similar phenomenon also occurred at the end of 2022. After the release of ChatGPT, the demand for AI skills jumped.

It can be seen that AI penetration is not advancing at a constant speed, but accelerating in pulses - Every technological and product breakthrough is a catalyst for job demand.

In addition, the high concentration in geographical distribution is also worthy of attention. The five major urban agglomerations (the Yangtze River Delta, the Pearl River Delta, the Beijing - Tianjin - Hebei region, the middle reaches of the Yangtze River, and Chengdu - Chongqing) have gathered 90% of the national demand for AI - skilled positions. The most significant increase comes from the Pearl River Delta. Due to the transformation demand of AI + manufacturing and the reduction of deployment thresholds by DeepSeek, the proportion of AI positions increased significantly in the first half of 2025.

Finding 2: The increment of AI positions is shifting from "tool - making" to "tool - using"

This is the steepest curve in the entire report.

Among AI - skilled positions, there are two major categories: "development skills" (people who create AI) and "application skills" (people who use AI). In the first quarter of 2024, development positions accounted for more than 80%, while application positions accounted for less than 20%. In the second quarter of 2025, development positions dropped to 65%, and application positions climbed to 35%.

Development positions still account for nearly two - thirds, but application positions almost doubled in one year.

What does this mean? AI recruitment is spreading from single - model development to a wider range of scenario applications. Enterprises are not only looking for people who can train models and adjust algorithms but also urgently looking for people who can embed ChatGPT, DeepSeek, Midjourney, etc. into business processes and directly solve problems.

According to Rework's analysis of LinkedIn's recruitment data in 2026, this trend is also confirmed. The demand for AI skills has increased by 142% in the past 12 months, and the growth rate of application - type positions has exceeded that of core development positions, such as prompt engineers and AI interaction designers.

There is also a "generalist" reshuffle within development positions. Among algorithm positions, the proportion of generalized positions marked only as "AI algorithm engineers" without specifying a specific direction has jumped from 14% to 24%, becoming the largest sub - category. At the same time, the proportion of "composite positions" that require both traditional deep - learning and large - model technologies has climbed from less than 16% to more than 25%, while traditional image algorithms, visual algorithms, etc. are shrinking.

The underlying logic behind this is that the emergence of large models has changed the way of AI development, shifting from "training dedicated models in specific fields" to "calling general large models + scenario adaptation". The scarcity of old - style specialists (who only know one algorithm) has decreased, and composite generalists who are proficient in both "large models + traditional AI" have become more valuable. Enterprises still need specialists, but they are no longer satisfied with narrow - field specialists who only know one direction.

More people may be more concerned about the changes in non - technical positions. The most prominent growth in AI demand is in consulting/analyst positions, where the AI penetration rate soared from 1.03% to 2.71%, and in design/creative positions, where it doubled from 1.29% to 2.74%. In addition, the AI penetration rates in management, education/research, and product/project positions are also relatively high. A common feature emerges: The first positions in non - technical fields to be penetrated by AI skills are knowledge - intensive positions, rather than physical labor and production operation positions.

Finding 3: China has not seen the "scissors gap" in the United States, and entry - level positions are still holding up

A cruel thing is happening in the US market. After the application of AI, the demand for senior positions continues to grow, but the demand for entry - level positions shows a continuous shrinking trend. A study by Lichtinger et al. based on tens of millions of LinkedIn resumes found that generative AI is a "seniority - biased technological change", and it has a much greater impact on junior employees than on senior employees. The data of Brynjolfsson et al. confirms this. Among high - exposure AI positions, the employment of early - career workers aged 22 - 25 has decreased by nearly 20%. A Fortune report used a more eye - catching title - "The Great White - Collar Recession".

Currently, China has not replicated this scenario.

Data shows that the proportion of senior positions in Chinese AI positions has slightly declined during the observation period, from 63.18% in the first quarter of 2024 to 60.87% in the second quarter of 2025. At the same time, although the proportion of entry - level positions fluctuated greatly in 2024, it has risen and fallen in the long - term, without showing a continuous shrinking trend like that in the United States.

The trend of the full sample including non - AI positions is more obvious. The proportion of senior positions has decreased from 22.64% to 15.40%, while the proportion of entry - level positions has remained stable at around 7%.

Why is it different between China and the United States? There are two possible reasons.

First, the cost structure is different. The median annual salary of entry - level positions in the United States is $50,000 - $85,000, and using AI for replacement has a relatively clear cost - reduction benefit. The cost of entry - level labor in China is relatively low. After calculating, enterprises may find that equipping entry - level employees with AI to improve efficiency is more cost - effective and reliable than direct replacement.

Second, the industrial stages are different. The United States is still in the arms race of underlying large - model R & D, and the demand for senior scientists and architects remains high. China is accelerating the shift to scenario - application implementation, and this kind of demand naturally favors application - type and engineering - type talents, which supports the stability of entry - and mid - level positions.

But an important limitation should be added here: The "resilience" of entry - level positions may be temporary.

The World Economic Forum (WEF) constructed four scenarios for employment in 2030 in January 2026. There is only one key variable: Whether the AI skill adaptation speed of the labor force can keep up with the growth speed of AI capabilities. The resilience of entry - level positions in China is based on a premise - "AI empowerment is more cost - effective than AI replacement". However, the capabilities of large models are still increasing exponentially, and the usage cost is rapidly decreasing. Once the critical point is passed and the premise is rewritten, "empowerment" is likely to turn into "replacement".

In a nutshell: The United States is using AI to replace entry - level employees, while China is using AI to enhance entry - level employees. But how long this enhancement can last depends on the evolution speed of models and the cost - reduction curve. This is a time window, not a stable state.

Finding 4: AI positions prefer "high - educated + experienced" people, but the signaling effect of academic qualifications is declining

The requirements for academic qualifications and experience in AI positions are much higher than the market average.

Academic qualifications: Among all positions, 24% require a "bachelor's degree or above". What about AI positions? 71%. The proportion of AI positions requiring master's or doctoral degrees is nearly 12%, far exceeding the proportion of less than 1% of the master's and doctoral groups in the national population. Statistical tests show that the probability of setting a bachelor's degree threshold for AI positions is 4 - 8 times that of non - AI positions.

Experience: 79% of AI positions require work experience, 22 percentage points higher than the whole market. The average years of experience required are 1.59 - 2.19 years higher than the market benchmark.

Employers are not looking for "AI beginners" but for "industry veterans + AI skills". The additional 1 - 2 years of experience requirement is a recognition of the value of the practitioners' original domain knowledge and judgment. You are not necessarily needed just because you know AI, but because you have accumulated non - codifiable judgment in a certain field, and AI can become your amplifier. Experimental research has also confirmed this (Ajuzieogu, 2025). Experienced financial analysts combined with AI tools can improve the quality of investment advice by 31%.

The 2025 OECD report on enterprise AI adoption points out a more fundamental structure: Senior positions require "developing new algorithms", while entry - level positions require "using AI tools". The so - called "AI skill requirements" refer to completely different things at different ranks.

However, there is a loosening signal: The proportion of AI positions requiring a "bachelor's degree or above" reached a peak of about 80% in the third quarter of 2024 and then began to decline, dropping to about 70% in the second quarter of 2025. A study by PwC covering nearly 1 billion recruitment advertisements on six continents also found that in positions with the highest AI penetration, employers' requirements for formal academic qualifications are declining the fastest.

Since the industrial era, the traditional career development logic has been "good academic qualifications → good positions → high income". Now, there is a parallel path. AI skills are becoming an "alternative qualification". It is not replacing academic qualifications themselves but replacing the signaling function that academic qualifications originally undertook: proving that you have learning ability, tool - migration ability, and problem - solving ability. If you don't have a high - level diploma but can use AI to innovate and solve real problems, the market is also willing to give you a chance.

Finding 5: The salaries of AI positions are not only higher but also more resistant to decline

Throughout the observation period, the average monthly salary of AI positions has always been 7,000 - 9,500 yuan higher than that of non - AI positions, and the premium ratio has remained above 40%.