The pace of AI capability change
The ILO 2025 refined index is the most-current per-occupation generative-AI exposure measure currently in the public domain. It reflects AI capability as of the 2025 update window. AI capability is not stable; the gradient is updated annually, and the calculator refreshes when the source updates. The next ILO update is expected in late 2026.
The lag between exposure and displacement
Exposure measures task-level technical feasibility for AI. Displacement is the labour-market outcome (workers laid off, roles eliminated). The two have not moved in tandem in 2024-2025. Brookings 2025 (No AI Jobs Apocalypse, For Now) finds aggregate-labour-market data does not yet show mass displacement of high-exposure occupations. This could change as AI deployment matures; the lag dynamic is not well-modelled in any public data.
Within-occupation variation
The calculator outputs an occupation-family-level gradient. A marketing manager at a SaaS company and a marketing manager at a regional non-profit are the same O*NET-SOC code and receive the same gradient. Within-occupation variation (employer, geography, seniority, role specialisation) is real and sometimes larger than between-occupation variation. The calculator cannot resolve this.
The four-band ceiling
The ILO 2025 index outputs four bands (Low, Moderate, High, Very High). Two occupations both in the High band may differ substantially in actual exposure within the band. The site does not invent within-band precision; the band is the resolution of the source data.
The decade-projection cycle
BLS Employment Projections are released every two years and cover a decade. The 2024-2034 projection cycle was released in September 2025. Rapid 2024-2026 AI-deployment shifts may not yet be incorporated. The next BLS projection cycle is expected late 2026.
The aggregate-vs-AI-specific concern
BLS projections cover demographics, demand, policy, and technology together. They do not isolate AI as a discrete disruption variable. The calculator uses BLS for the growth panel (where AI is one factor among many) and ILO 2025 for the exposure gradient (where the measure is AI-specific). The split is intentional, but the BLS number is not an AI-specific projection.
The frontier-AI-vendor optics
Brookings 2024's task-level rubric is built on OpenAI's task-completion data. Brookings 2024 itself flags this. The calculator inherits the limitation; the ILO 2025 refined index acts as a triangulation against the Brookings task layer. Using a frontier-AI-vendor's own work as primary methodology for an AI-jobs calculator is a calibrated compromise; the alternative (using no task-level data) would give a less-useful output.
Why this site is publishing it anyway
The calculator is the most-defensible per-occupation reading of the AI exposure question that the public-domain primary sources support. It is calibrated, not precise. The site exists because the alternative (vendor-published single-percentile calculators with no methodology) is structurally worse. The honest claim is that this tool is more transparent, more sourced, and more empowering than the alternatives. It is not a claim of certainty.
For pre-empted critiques, see /how-to-argue-with-this/. For the full methodology, see /methodology/.