Our research corpus has grown — from 4,422 papers at the last knowledgebase update to 13,150 papers now (OpenAlex + arXiv, 1984–2026). 11,721 of them have a full abstract and are indexed in depth. What the growth means, in four findings.
The field is exploding
7,041 papers in the 12-month window since September 2025 alone — more than half of the entire corpus is younger than twelve months. The previous 12-month window brought 4,343 papers. Within the corpus period, the annual publication rate has increased more than twentyfold from 2022 to 2026.
The evaluation gap is deepening
Despite the growth, the central finding from our
evidence-gap analysis holds — only
sharper: just 2.2 % of all papers carry the subcategory "Evaluation &
Benchmarks". And 9 of the 20 thinnest yet fastest-growing research cells in
our gap analysis (thinness × momentum) are */evaluation cells. Research
builds programs, curricula, and frameworks — but almost nobody measures
whether they work.
Compliance tripled
The regulatory pressure shows up in the numbers: the "Compliance & AI Act"
category grew from 214 to 652 papers year-over-year (+205 %). Article 4
of the EU AI Act — the obligation to ensure AI literacy across the workforce
— is the main driver. Interestingly, the compliance/evaluation combination
remains one of the thinnest cells, at 4 papers. Much is being written about
compliance, almost nothing about its measured effectiveness.
The largest areas
| Category | Papers |
|---|---|
| AI Literacy Construct | 1,146 |
| AI Literacy Pedagogy | 1,139 |
| Learning Design | 946 |
| Assessment | 942 |
| Compliance | 909 |
Foundational research still dominates — but the applied areas (learning design, assessment, compliance) are growing fastest.
What this means for training
Anyone offering AI-literacy training today is arguing from a field that is 97.8 % unevaluated. That is not a weakness of trainers — it is the central gap of research itself. That is exactly why our assessment instrument and the readiness mini-audit combine Art. 4 compliance with program evaluation: organizations need the measurement layer that research owes them.
The old posts citing 4,422 papers remain online — they were the respective snapshot, and they show how fast this field moves.