In a September 8, 2026 analysis, the Bipartisan Policy Center found that US online job postings mentioning AI skills rose 8% in the fourth quarter of 2025, climbed another 47.5% by April from the start of 2026, added 27% by August and stood 165% above their level a year earlier. The analysis draws on Lightcast data and counts advertised skills, not people hired.
A separate September workforce report from iCIMS also found employers adding AI requirements across industries, although AI-related postings represented 4% of US hiring demand in its Lightcast-based analysis. Together, the findings point to fast growth from a limited base and a widening occupational reach—not a comparable surge in completed hiring.
The timeline shows acceleration, not the size of the market
The headline increase describes the change in postings containing an AI skill from one year to the next. It does not mean that AI requirements now appear in most advertisements, nor does it measure the share of workers whose daily responsibilities involve AI.
That distinction is central to interpreting the pace of change. A relatively small category can produce a large growth rate when employers begin adding terms such as generative AI, prompt engineering or machine learning to job descriptions. The continued gains through August nevertheless make a one-month anomaly less plausible: advertised requirements advanced over several measurement points.
The safest conclusion is that employers are revising job descriptions and screening criteria quickly. The data cannot establish whether every listed capability is mandatory, how consistently employers define it or whether the requirement survives from the initial advertisement to the final hiring decision.
The occupational map now reaches beyond technology roles
Technical occupations remain the most concentrated part of the market. Data engineering, data science and model-development roles require capabilities that are directly tied to building, deploying or maintaining AI systems, so their skill profiles differ from jobs where AI is one tool inside a broader workflow.
The Lightcast Global AI Skills Outlook separates the market into skills, occupations and locations rather than treating AI work as a single occupation; it identifies writers as a notable nontechnical group for generative-AI requirements and also finds AI-related value in established fields such as architecture and law. That structure helps explain how demand can extend beyond tech without turning every affected position into an AI specialist role.
Industry and occupation should also be kept separate. A finance, healthcare or manufacturing employer may recruit a technical specialist, while a technology company may seek a lawyer, writer or operations manager who can work with AI-enabled processes. Sector-level growth alone therefore cannot show what an individual employee will be asked to do.
Human and operational skills remain part of the requirement
The emerging pattern is additive rather than a wholesale replacement of existing expertise. The September data places AI capabilities alongside communication, management, leadership, problem-solving, automation, operations and workflow management. Those combinations suggest that employers often want AI applied within a business process, not detached from it.
For workers, the evidence supports prioritizing capabilities in context. A writer may need to evaluate generated material and preserve accuracy; an accountant may need to understand where an automated workflow fits within controlled financial processes; a technical specialist may need deployment and data skills. These are occupational interpretations of the posting pattern, not proof that any particular tool or certificate produces better hiring outcomes.
Domain knowledge remains the stable part of the signal. AI terminology may change quickly, but advertisements that combine it with judgment, communication or process management indicate that employers still value the ability to apply tools to the underlying work and assess the result.
Postings cannot be converted into a count of hires
An online advertisement records an employer’s stated demand at one stage of recruitment. A listed skill may be required or preferred, one advertisement may cover several openings, and a vacancy may be edited, duplicated, cancelled or filled through another channel.
Posting data also reveals neither the number of applicants nor how they progressed through screening, interviews and offers. It cannot show whether a person started work, remained in the role or used the advertised capability after joining. A rise in mentions is therefore a labor-demand signal, not a hiring total.
The available evidence establishes that AI language is spreading rapidly through US job advertising and appearing in both specialist and established nontechnical work. What remains unresolved is the conversion from advertisements to offers, employee starts and sustained workplace use; answering that will require outcome data broken down by occupation and industry rather than additional posting counts alone.
Also read: