A skills taxonomy is a structured, hierarchical list that classifies skills into categories and subcategories, so everyone uses the same name for the same skill. A skills ontology goes further: it maps the relationships between skills, jobs, and people, such as which skills are related, which are prerequisites for others, and which jobs require which combinations.
Put simply, a taxonomy tells you what a skill is and where it sits; an ontology tells you how it connects to everything else. Most modern recruiting technology uses both, with the taxonomy as the vocabulary and the ontology as the map that lets software, and AI agents, reason about fit.
Skills Taxonomy vs Skills Ontology at a Glance
| Skills taxonomy | Skills ontology | |
|---|---|---|
| Structure | Tree or hierarchy (category, subcategory, skill) | Graph or network of linked concepts |
| Core question | “What do we call this skill, and what group does it belong to?” | “How does this skill relate to other skills, jobs, and people?” |
| Relationships | Mostly parent and child | Many types: related to, requires, is part of, is adjacent to, is essential or optional for |
| Typical use | Standardizing job requisitions, profiles, and reports | Matching, adjacent-skill discovery, internal mobility, AI reasoning |
| Maintenance | Periodic updates by a governing team | Continuous updates, often data-driven and machine-assisted |
| Public example | Lightcast Open Skills (three-level hierarchy) | ESCO (occupations and skills linked as machine-readable concepts) |
What Is a Skills Taxonomy?
A skills taxonomy is a controlled vocabulary for skills, organized in levels. Lightcast’s skills taxonomy, one of the most widely used, lists more than 34,000 skills in a three-level hierarchy of category, subcategory, and skill, and it can be browsed for free.
Taxonomies solve a naming problem. Without one, “customer service,” “client support,” and “customer care” appear as three different skills in your ATS, your job ads, and your analytics. A taxonomy collapses those into one standardized entry, which makes search, reporting, and job distribution far more reliable. This is the same principle behind job title normalization, applied to skills instead of titles. For related terms, see the agentic recruiting glossary.
The limit of a taxonomy is that it is mostly one-dimensional. It can tell you that Python belongs under “programming languages,” but not that someone who knows Python and statistics is likely to learn machine learning quickly.
What Is a Skills Ontology?
A skills ontology is a knowledge model that defines skills and the relationships among skills, occupations, and other entities. The European Commission’s ESCO framework is a good public example: it covers 3,039 occupations and 13,939 skills across 28 languages and describes itself as a dictionary whose concepts, and the relationships between them, can be understood by computer systems. ESCO is published as linked open data and can be downloaded in SKOS, a W3C format for representing thesauri, classification schemes, and taxonomies. In ESCO, an occupation is not just a label: it is linked to the skills considered essential or optional for it.
That relationship layer is what makes an ontology useful for hiring. With it, a system can:
- Recognize that a candidate with warehouse inventory experience has adjacent skills for a logistics coordinator role
- Suggest skills to add to a job description based on what similar roles require
- Expand a search beyond exact keyword matches to candidates with related skills
- Identify internal employees who are a short step away from an open role
Where Do O*NET and Other Frameworks Fit?
Most organizations do not build skills frameworks from scratch. They start from public or commercial frameworks and extend them.
- O*NET (US). The ONET-SOC taxonomy includes 1,016 occupational titles, 923 of which have detailed data, plus 56,495 job and alternate titles. ONET also catalogs more than 2,000 detailed work activities that describe what people actually do in each occupation. O*NET began as a taxonomy of occupations, but its linked descriptors of skills, knowledge, and activities give it ontology-like depth.
- ESCO (EU). Multilingual, relationship-rich, and machine-readable, as described above.
- Lightcast Open Skills. A large, frequently updated skills taxonomy built from job posting and profile data.
- Proprietary frameworks. HR tech and talent intelligence vendors, including professional networks, maintain their own skills graphs and often map them to O*NET or ESCO.
Why the Difference Matters for Skills-Based Hiring
Skills-based hiring promises bigger, more diverse talent pools. LinkedIn’s Economic Graph research found that a skills-first approach expands talent pools by nearly 10 times on average, and grows Gen Z candidate pools by more than 10 times. Adoption intent is high: in TestGorilla’s 2025 survey, 85% of employers said they use skills-based hiring and 53% said they had dropped degree requirements (vendor data).
Practice lags behind intent. A 2024 study by the Burning Glass Institute and Harvard Business School found that despite widespread announcements of dropped degree requirements, the change affected fewer than 1 in 700 hires. One reason is infrastructure: removing a degree requirement is easy, but evaluating skills consistently requires a shared skills language (a taxonomy) and a way to reason about equivalent and adjacent skills (an ontology).
The pressure is not going away. The World Economic Forum’s Future of Jobs Report 2025 projects that nearly 40% of workers’ core skills will change by 2030, and 63% of employers name skills gaps as the main barrier to business transformation. A static list cannot keep pace with that change; a relationship model can at least show where skills are shifting and which adjacent skills bridge the gap.
How Skills Ontologies Power AI Recruiting Agents
AI recruiting agents depend on structured skills data. When an agent writes a job ad, targets an audience, screens applicants, or rediscovers past candidates, it needs to know that two differently worded skills mean the same thing (taxonomy) and that a candidate’s skills are close enough to a role’s requirements (ontology). Large language models help interpret free text, but grounding them in a governed skills framework makes matching more consistent, more explainable, and easier to audit for fairness.
In Joveo’s agentic recruiting platform, skills and job data inform decisions across the funnel: which markets and publishers to advertise in, how to describe roles so the right candidates respond, and how to screen and rediscover talent. See how AI recruiting agents work and the data sources behind labor market intelligence.
Do You Need a Taxonomy, an Ontology, or Both?
- Start with a taxonomy if your skills data is inconsistent across job ads, requisitions, and your ATS, and your first goal is clean reporting and search.
- Add an ontology when you want matching beyond keywords, internal mobility recommendations, or AI agents that can reason about adjacent skills.
- Anchor both to a public standard such as O*NET or ESCO, so your data can be compared against labor market sources and exchanged with other systems.
- Govern it. Assign ownership, set an update cadence, and audit matching outcomes for adverse impact, just as you would any selection procedure.
For the broader strategy, see what skills-based hiring is and how it works.
Frequently Asked Questions
What is the main difference between a skills taxonomy and a skills ontology?
A skills taxonomy classifies skills into a hierarchy of categories. A skills ontology defines the relationships between skills, jobs, and people, such as related, required, or adjacent skills.
Is O*NET a taxonomy or an ontology?
ONET is built on an occupational taxonomy (ONET-SOC), but because it links occupations to detailed skills, knowledge, abilities, and work activities, it is often used as the foundation for ontology-style matching.
What is an example of a skills ontology?
ESCO, the European Commission’s classification of occupations and skills, is a widely cited example. It links 3,039 occupations to 13,939 skills in 28 languages and publishes those relationships as machine-readable linked open data.
What is a skills graph?
A skills graph is a practical implementation of a skills ontology, usually stored as a network of nodes (skills, jobs, people) and edges (relationships). The terms are often used interchangeably.
Why do recruiters need a skills taxonomy?
A taxonomy standardizes how skills are named across job ads, applications, and systems, which makes searching, reporting, and matching reliable. Without one, the same skill appears under many names and data cannot be compared.
How does a skills ontology help AI in recruiting?
It gives AI a structured map of how skills relate, so matching, screening, and candidate rediscovery can go beyond exact keywords while remaining consistent and auditable.
