Role Definition
| Field | Value |
|---|---|
| Job Title | Lexicographer |
| Seniority Level | Mid-Level |
| Primary Function | Compiles and edits dictionary entries for print and digital publication. Researches word usage through corpus analysis, writes and refines definitions, traces etymologies, selects illustrative example sentences, applies usage and register labels, and makes editorial judgments on word inclusion and exclusion. Works with tools like Sketch Engine and AntConc to analyse large text corpora. |
| What This Role Is NOT | NOT a computational linguist who builds NLP models. NOT a translator or interpreter. NOT a copywriter or content writer. NOT an entry-level editorial assistant who only proofreads. |
| Typical Experience | 3-8 years. Degree in linguistics, English, or related humanities. Often holds a master's in lexicography or computational linguistics. |
Seniority note: Senior/chief lexicographers who set editorial policy, manage teams, and make final decisions on culturally sensitive entries would score higher (likely Yellow). Entry-level corpus assistants doing primarily data extraction would score deeper Red.
Protective Principles + AI Growth Correlation
| Principle | Score (0-3) | Rationale |
|---|---|---|
| Embodied Physicality | 0 | Fully desk-based, digital work. No physical component whatsoever. |
| Deep Interpersonal Connection | 0 | Minimal human interaction. Primarily solo research and writing. Some peer review collaboration, but the value is linguistic precision, not the human relationship. |
| Goal-Setting & Moral Judgment | 2 | Significant editorial judgment: deciding which words merit inclusion, how to handle offensive or culturally sensitive terms, what connotations to capture, detecting and mitigating bias in definitions. These are judgment calls with no algorithmic answer — dictionary entries shape how people understand language. |
| Protective Total | 2/9 | |
| AI Growth Correlation | 0 | AI adoption neither increases nor decreases demand for dictionary editors. NLP tools consume lexicographic data but the relationship is one-way — more AI does not create more lexicographer positions. |
Quick screen result: Protective 2 + Correlation 0 = Likely Red Zone.
Task Decomposition (Agentic AI Scoring)
| Task | Time % | Score (1-5) | Weighted | Aug/Disp | Rationale |
|---|---|---|---|---|---|
| Corpus analysis and neologism detection | 25% | 4 | 1.00 | DISPLACEMENT | AI agents chain Sketch Engine, AntConc, and custom NLP pipelines to scan billions of words, extract frequency data, identify neologisms, and flag semantic shifts. The output IS the deliverable — the lexicographer who manually reads through concordance lines is being replaced by ML-driven pattern detection. |
| Definition writing and editing | 25% | 3 | 0.75 | AUGMENTATION | LLMs generate preliminary definitions from corpus data and existing entries. However, human lexicographers still lead — refining nuance, ensuring definitions capture real-world usage rather than training-data artifacts, and maintaining consistency with house style. AI drafts, human decides. |
| Etymology research | 10% | 3 | 0.30 | AUGMENTATION | AI can cross-reference historical corpora and etymological databases to trace word origins and synthesise information. But tracing disputed etymologies, evaluating competing theories, and making judgment calls on uncertain origins still requires human expertise. Human-led, AI-accelerated. |
| Example sentence selection/creation | 10% | 4 | 0.40 | DISPLACEMENT | LLMs generate plausible example sentences on demand. AI agents can search corpora for authentic usage examples, rank them by clarity and typicality, and present candidates. Human reviews for naturalness and bias, but the generation and selection workflow is largely automated. |
| Usage labeling, categorization, and cross-referencing | 10% | 4 | 0.40 | DISPLACEMENT | NLP classifies register (formal/informal/slang), identifies subject domains, and generates cross-references automatically. Sentiment analysis assigns connotation labels. The structured, rule-based nature of labeling makes it highly automatable. |
| Editorial judgment — inclusion/exclusion decisions | 10% | 2 | 0.20 | AUGMENTATION | Deciding whether a word has achieved sufficient currency to warrant inclusion, or whether an archaic term should be retained, requires cultural awareness, editorial policy interpretation, and accountability. AI can flag candidates; humans make the call. This is where dictionary editorial boards earn their authority. |
| Peer review, quality assurance, and bias detection | 10% | 2 | 0.20 | AUGMENTATION | Reviewing entries for accuracy, consistency, and potential bias — particularly in definitions of terms related to race, gender, disability, or cultural identity — requires human sensitivity. NLP can check formatting and flag inconsistencies, but detecting whether a definition inadvertently perpetuates stereotypes is human work. |
| Total | 100% | 3.25 |
Task Resistance Score: 6.00 - 3.25 = 2.75/5.0
Displacement/Augmentation split: 45% displacement, 55% augmentation, 0% not involved.
Reinstatement check (Acemoglu): Partial. AI creates some new tasks — validating AI-generated definitions, auditing LLM outputs for bias, curating training data for language models. But these are absorbed into existing lexicographic workflows rather than creating net new positions. The "computational lexicographer" hybrid role at tech companies is growing, but these are closer to computational linguist positions than traditional lexicography.
Evidence Score
| Dimension | Score (-2 to 2) | Evidence |
|---|---|---|
| Job Posting Trends | -1 | Extremely niche profession — estimated fewer than 5,000 dedicated positions in the US. Traditional publishing lexicography positions are contracting as publishers reduce editorial headcount. Merriam-Webster, OUP, and HarperCollins have all restructured editorial teams. Some growth in "computational lexicographer" roles at tech companies (Google, Amazon, Apple) but these are NLP-adjacent, not traditional lexicography. |
| Company Actions | -1 | Major dictionary publishers have reduced editorial staff over the past decade. Collins and several other publishers rely increasingly on freelance contributors and automated corpus tools. No dramatic AI-driven layoff announcements specific to lexicographers, but steady attrition as digital-first dictionaries require smaller teams to maintain and update. |
| Wage Trends | 0 | Comparably reports $97,147 average (US), but this includes tech company computational linguist roles that inflate the number. Traditional publishing lexicographers earn $55K-$75K. Wages are stable but not growing in real terms. The wide salary range ($48K-$383K) reflects the publishing-vs-tech divide, not growth. |
| AI Tool Maturity | -1 | Production NLP tools (Sketch Engine, AntConc) have automated corpus analysis for years. LLMs now generate draft definitions, example sentences, and cross-references. Kocon et al. (2023) demonstrated LLMs creating dictionary entries automatically. Tools handle 50-70% of routine lexicographic workflow with human oversight — clearly in the "performing core tasks with human oversight" bracket. Anthropic observed exposure for Editors (SOC 27-3041): 24.6% — moderate, predominantly augmented. |
| Expert Consensus | 0 | Mixed. Leading lexicographers describe AI as "augmentation rather than replacement" — but this consensus refers to the senior editorial judgment role, not the full scope of mid-level work. AI is acknowledged as automating the data-intensive components. No consensus that the profession is growing; agreement that it is contracting and transforming. |
| Total | -3 |
Barrier Assessment
Reframed question: What prevents AI execution even when programmatically possible?
| Barrier | Score (0-2) | Rationale |
|---|---|---|
| Regulatory/Licensing | 0 | No licensing required for lexicographers. No regulatory framework governs dictionary compilation. |
| Physical Presence | 0 | Fully remote/digital work. No physical presence required. |
| Union/Collective Bargaining | 0 | No significant union representation in lexicography. Most positions are at-will or contract-based. |
| Liability/Accountability | 0 | Low stakes if a definition is inaccurate — no one faces legal consequences. Dictionary errors are embarrassing and reputationally damaging, but not criminal or civilly liable. |
| Cultural/Ethical | 1 | Some cultural expectation that dictionaries are authored by human experts, particularly for sensitive terms. The authority of a dictionary partly rests on the credibility of its editorial board. However, society is unlikely to reject AI-assisted dictionaries if the output is accurate — the barrier is moderate, not strong. |
| Total | 1/10 |
AI Growth Correlation Check
Confirmed at 0 (Neutral). AI adoption creates demand for linguistic data (training corpora, word lists, semantic annotations) but this demand is met by computational linguists and NLP engineers, not traditional lexicographers. The lexicographer's core product — dictionary entries — is consumed by AI systems, but this doesn't create a feedback loop that increases demand for human lexicographers. If anything, AI's ability to generate definitions reduces the need for human production capacity.
JobZone Composite Score (AIJRI)
| Input | Value |
|---|---|
| Task Resistance Score | 2.75/5.0 |
| Evidence Modifier | 1.0 + (-3 × 0.04) = 0.88 |
| Barrier Modifier | 1.0 + (1 × 0.02) = 1.02 |
| Growth Modifier | 1.0 + (0 × 0.05) = 1.00 |
Raw: 2.75 × 0.88 × 1.02 × 1.00 = 2.4684
JobZone Score: (2.4684 - 0.54) / 7.93 × 100 = 24.3/100
Zone: RED (Green >=48, Yellow 25-47, Red <25)
Sub-Label Determination
| Metric | Value |
|---|---|
| % of task time scoring 3+ | 80% |
| AI Growth Correlation | 0 |
| Sub-label | Red — AIJRI <25 but Task Resistance 2.75 >= 1.8 (not Imminent) |
Assessor override: None — formula score accepted. The 24.3 sits 0.7 points below Yellow, which is borderline. However, the evidence and barrier scores are both weak, and 45% of task time faces outright displacement. The Red classification is honest for the traditional publishing lexicographer.
Assessor Commentary
Score vs Reality Check
The 24.3 score places this role just below the Yellow Zone boundary (25), which is borderline but defensible. The task decomposition reveals a role split roughly 45/55 between displacement and augmentation — but the augmented tasks (definition writing, etymology, peer review) are themselves being compressed by AI assistance, meaning the human hours required per entry are shrinking even where humans remain in the loop. The 1/10 barrier score is the critical factor: there is virtually nothing preventing AI from executing lexicographic work. No licensing, no physical presence, no liability, no union protection. The only barrier is a modest cultural preference for human-authored dictionaries, which is eroding as AI-generated content becomes normalised. If barriers were higher, this would tip into Yellow.
What the Numbers Don't Capture
- Market contraction vs technology: The lexicography profession was already shrinking before AI. The shift from print to digital dictionaries in the 2000s-2010s reduced headcount dramatically. AI is accelerating an existing decline, not initiating one. The evidence score may understate the cumulative pressure because the pre-AI contraction is treated as baseline.
- Function-spending vs people-spending: Tech companies (Google, Apple, Amazon) invest heavily in linguistic resources for voice assistants and NLU — but this spending goes to computational tools, datasets, and NLP infrastructure, not to human lexicographers. The market for lexicographic products is growing (more languages, more AI training data needed); the market for human lexicographers is not.
- The "computational lexicographer" escape hatch: A genuinely different role is emerging — part lexicographer, part data scientist, part NLP engineer — at tech companies. These roles pay $90K-$130K+ and are growing. But they require Python, ML frameworks, and corpus engineering skills that traditional humanities-trained lexicographers often lack. The transition is more like a career pivot than a natural progression.
Who Should Worry (and Who Shouldn't)
If you work at a traditional dictionary publisher writing and editing entries — you are the most exposed version of this role. Publishers are doing more with fewer people, and AI tools are the primary reason. Your workflow of reading concordance lines, drafting definitions, and selecting examples is precisely what LLMs and corpus tools automate. 2-4 year window before significant headcount reduction.
If you are the person who decides what goes in the dictionary — the senior editorial judgment role — you are safer than the label suggests. Deciding whether to include a controversial neologism, how to define a politically charged term, or when to retire an archaic entry requires cultural awareness and institutional accountability that AI cannot replicate. But these decisions occupy perhaps 10-15% of a mid-level lexicographer's time.
If you have computational skills — Python, NLP, corpus engineering — you have an exit ramp into computational linguistics or NLP engineering roles that are Yellow or Green Zone. The lexicographer who can build corpus pipelines and evaluate LLM outputs is far more employable than one who only writes definitions.
The single biggest separator: whether you are a dictionary entry writer or a linguistic authority. The entry writers are being automated. The authorities who shape editorial policy and make culturally consequential decisions are transforming into AI supervisors — but there are very few of those positions.
What This Means
The role in 2028: A radically smaller profession. Major dictionaries will maintain small editorial boards (5-15 people) who oversee AI-generated entries, make inclusion/exclusion decisions, and handle culturally sensitive definitions. The production lexicographer who writes 400-600 entries per year is replaced by AI systems that generate thousands of draft entries per day with human curation. Tech companies will employ "computational lexicographers" who bridge linguistics and NLP — but these are effectively a different profession.
Survival strategy:
- Learn Python, NLP, and corpus engineering. The computational lexicographer hybrid role is the viable path. Sketch Engine and AntConc skills are not enough — you need to build pipelines, evaluate ML outputs, and work with structured data (XML, JSON).
- Move into AI content quality and bias auditing. Your expertise in nuanced language, definition accuracy, and sensitivity review is directly transferable to evaluating AI-generated text for bias, accuracy, and cultural appropriateness.
- Pursue senior editorial leadership. The editorial board role — setting policy, making inclusion decisions, handling media and public accountability — is the human stronghold. Position yourself for these scarce but protected positions.
Where to look next. If you're considering a career shift, these Green Zone roles share transferable skills with lexicography:
- Children's Librarian (AIJRI 49.3) — Research skills, language expertise, and community education transfer directly. Strong interpersonal and physical presence barriers provide protection.
- Editor-in-Chief / Managing Editor (AIJRI 49.4) — Senior editorial judgment, content strategy, and team leadership. The strategic end of editorial work resists automation far better than production.
- Speech-Language Pathologist (AIJRI 55.1) — Deep linguistic analysis expertise transfers to clinical language assessment. Requires additional clinical training but leverages your core strength — understanding how language works.
Browse all scored roles at jobzonerisk.com to find the right fit for your skills and interests.
Timeline: 2-4 years for significant displacement in traditional publishing. The technology is already deployed — Sketch Engine, LLM-based definition generation, automated corpus analysis — and the profession was already contracting. Computational lexicography at tech companies has a longer runway but serves a fundamentally different labour market.
