@base <https://www.anthropic.com/research/labor-market-impacts> .
@prefix schema: <https://schema.org/> .
@prefix owl: <https://www.w3.org/2002/07/owl#> .

<#article> a schema:Article ;
  schema:headline "Labor market impacts of AI: A new measure and early evidence"@en ;
  schema:name "Labor market impacts of AI: A new measure and early evidence"@en ;
  schema:alternativeHeadline "Observed exposure combines LLM capability with real-world work usage"@en ;
  schema:datePublished "2026-03-05" ;
  schema:inLanguage "en" ;
  schema:url <https://www.anthropic.com/research/labor-market-impacts> ;
  schema:publisher <#publisher> ;
  schema:author <#author> ;
  schema:about
    <#observed-exposure>,
    <#theoretical-capability>,
    <#anthropic-economic-index>,
    <#onet-database>,
    <#task-level-exposure>,
    <#claude-usage-data>,
    <#automated-use-patterns>,
    <#coverage-gap>,
    <#computer-math-category>,
    <#computer-programmers>,
    <#customer-service-representatives>,
    <#data-entry-keyers>,
    <#bls-projections>,
    <#current-population-survey>,
    <#high-exposure-workers>,
    <#difference-in-differences>,
    <#young-worker-hiring>,
    <#job-finding-rate>,
    <#counterfactual-framework>,
    <#signal-from-noise> ;
  schema:abstract """Anthropic introduces observed exposure, a labor-market measure that combines theoretical LLM capability, Claude usage data, work context, automation weight, and task-time shares to track early AI exposure by occupation."""@en ;
  schema:articleBody """The report argues that AI labor-market effects are not yet visible as broad unemployment shocks, so better leading measures are needed. Observed exposure links O*NET occupational tasks, Anthropic Economic Index usage, and prior task-level LLM capability estimates. It finds that actual AI coverage remains far below theoretical capability, that more exposed occupations have weaker BLS projected growth, and that highly exposed workers tend to be older, female, more educated, and higher paid. The unemployment evidence is currently small and statistically weak, but hiring into exposed occupations for workers aged 22-25 shows a tentative slowdown."""@en ;
  schema:hasPart <#part-measure>, <#part-data>, <#part-gap>, <#part-occupations>, <#part-growth>, <#part-results>, <#part-updates> ;
  schema:mentions
    <#defined-terms>,
    <#argument-howto>,
    <#step-1>,
    <#step-2>,
    <#step-3>,
    <#step-4>,
    <#step-5>,
    <#faq-1>,
    <#faq-2>,
    <#faq-3>,
    <#faq-4>,
    <#faq-5>,
    <#faq-6>,
    <#faq-7>,
    <#faq-8>,
    <#faq-9>,
    <#faq-10>,
    <#faq-11>,
    <#faq-12>,
    <#faq-1-answer>,
    <#faq-2-answer>,
    <#faq-3-answer>,
    <#faq-4-answer>,
    <#faq-5-answer>,
    <#faq-6-answer>,
    <#faq-7-answer>,
    <#faq-8-answer>,
    <#faq-9-answer>,
    <#faq-10-answer>,
    <#faq-11-answer>,
    <#faq-12-answer>,
    <#author>,
    <#publisher> .

<#publisher> a schema:Organization ;
  schema:name "Anthropic"@en ;
  schema:url <https://www.anthropic.com/> .

<#author> a schema:Organization ;
  schema:name "Maxim Massenkoff and Peter McCrory"@en ;
  schema:url <https://www.anthropic.com/research/labor-market-impacts> .

<#codex> a schema:SoftwareApplication ;
  schema:name "Codex"@en .

<#defined-terms> a schema:DefinedTermSet ;
  schema:name "Defined terms for Labor market impacts of AI: A new measure and early evidence"@en ;
  schema:hasPart <#observed-exposure>, <#theoretical-capability>, <#anthropic-economic-index>, <#onet-database>, <#task-level-exposure>, <#claude-usage-data>, <#automated-use-patterns>, <#coverage-gap>, <#computer-math-category>, <#computer-programmers>, <#customer-service-representatives>, <#data-entry-keyers>, <#bls-projections>, <#current-population-survey>, <#high-exposure-workers>, <#difference-in-differences>, <#young-worker-hiring>, <#job-finding-rate>, <#counterfactual-framework>, <#signal-from-noise> ;
  schema:isPartOf <#article> .

<#observed-exposure> a schema:DefinedTerm ;
  schema:name "Observed exposure"@en ;
  schema:description """Anthropic's new measure of occupational AI exposure based on theoretical capability, observed usage, automation weight, work context, and task-time shares."""@en .

<#theoretical-capability> a schema:DefinedTerm ;
  schema:name "Theoretical LLM capability"@en ;
  schema:description """A task-level estimate of whether an LLM could make a task substantially faster, before accounting for real-world adoption."""@en .

<#anthropic-economic-index> a schema:DefinedTerm ;
  schema:name "Anthropic Economic Index"@en ;
  schema:description """Anthropic's usage-based research program for studying which economic tasks are performed with Claude."""@en .

<#onet-database> a schema:DefinedTerm ;
  schema:name "O*NET database"@en ;
  schema:description """The US occupational database that enumerates tasks associated with roughly 800 occupations."""@en .

<#task-level-exposure> a schema:DefinedTerm ;
  schema:name "Task-level exposure"@en ;
  schema:description """A task-level estimate of how much AI could affect work before aggregation to occupations."""@en .

<#claude-usage-data> a schema:DefinedTerm ;
  schema:name "Claude usage data"@en ;
  schema:description """Observed Claude traffic used to determine which theoretically exposed tasks are actually appearing in work-related settings."""@en .

<#automated-use-patterns> a schema:DefinedTerm ;
  schema:name "Automated use patterns"@en ;
  schema:description """Usage patterns that receive higher weight than augmentative use because they more directly indicate task substitution potential."""@en .

<#coverage-gap> a schema:DefinedTerm ;
  schema:name "Coverage gap"@en ;
  schema:description """The gap between tasks that LLMs could theoretically help with and tasks where professional AI usage is actually observed."""@en .

<#computer-math-category> a schema:DefinedTerm ;
  schema:name "Computer and Math category"@en ;
  schema:description """An occupational category where theoretical exposure is high and observed Claude coverage is reported at 33 percent."""@en .

<#computer-programmers> a schema:DefinedTerm ;
  schema:name "Computer Programmers"@en ;
  schema:description """The occupation identified as the most exposed in the report, with 75 percent coverage."""@en .

<#customer-service-representatives> a schema:DefinedTerm ;
  schema:name "Customer Service Representatives"@en ;
  schema:description """A highly exposed occupation linked to first-party API traffic and customer-service automation use cases."""@en .

<#data-entry-keyers> a schema:DefinedTerm ;
  schema:name "Data Entry Keyers"@en ;
  schema:description """A highly exposed occupation whose main source-document reading and entry tasks show significant automation."""@en .

<#bls-projections> a schema:DefinedTerm ;
  schema:name "BLS employment projections"@en ;
  schema:description """US Bureau of Labor Statistics projections from 2024 to 2034, used as an external comparison for observed exposure."""@en .

<#current-population-survey> a schema:DefinedTerm ;
  schema:name "Current Population Survey"@en ;
  schema:description """The survey data source used to compare worker characteristics, unemployment trends, and young-worker job starts."""@en .

<#high-exposure-workers> a schema:DefinedTerm ;
  schema:name "High-exposure workers"@en ;
  schema:description """Workers in occupations in the top observed-exposure quartile."""@en .

<#difference-in-differences> a schema:DefinedTerm ;
  schema:name "Difference-in-differences"@en ;
  schema:description """The framework used to compare unemployment gaps between more and less exposed workers after ChatGPT's release."""@en .

<#young-worker-hiring> a schema:DefinedTerm ;
  schema:name "Young-worker hiring"@en ;
  schema:description """The report's tentative signal that workers aged 22-25 are less likely to start jobs in highly exposed occupations."""@en .

<#job-finding-rate> a schema:DefinedTerm ;
  schema:name "Job finding rate"@en ;
  schema:description """The monthly rate at which young workers begin a new job in high- or low-exposure occupations."""@en .

<#counterfactual-framework> a schema:DefinedTerm ;
  schema:name "Counterfactual framework"@en ;
  schema:description """The report's approach for separating AI labor-market effects from business cycles and other confounding forces."""@en .

<#signal-from-noise> a schema:DefinedTerm ;
  schema:name "Signal from noise"@en ;
  schema:description """The goal of using repeated exposure and outcome measurement to identify genuine labor-market disruption."""@en .

<#part-measure> a schema:WebPageElement ;
  schema:name "Observed exposure as a new measure"@en ;
  schema:position 1 ;
  schema:about <#observed-exposure>, <#theoretical-capability>, <#anthropic-economic-index> ;
  schema:text """The report defines observed exposure as a task-to-occupation coverage measure that weights real work usage, automation, and task importance."""@en .

<#part-data> a schema:WebPageElement ;
  schema:name "Data sources and task coverage"@en ;
  schema:position 2 ;
  schema:about <#onet-database>, <#task-level-exposure>, <#claude-usage-data> ;
  schema:text """The measure combines O*NET task data, Anthropic Economic Index usage, and Eloundou et al. task-level LLM exposure estimates."""@en .

<#part-gap> a schema:WebPageElement ;
  schema:name "Gap between capability and adoption"@en ;
  schema:position 3 ;
  schema:about <#coverage-gap>, <#computer-math-category>, <#automated-use-patterns> ;
  schema:text """Actual professional Claude use remains a fraction of theoretical capability, with Computer and Math at 33 percent observed coverage versus much higher theoretical scope."""@en .

<#part-occupations> a schema:WebPageElement ;
  schema:name "Most exposed occupations"@en ;
  schema:position 4 ;
  schema:about <#computer-programmers>, <#customer-service-representatives>, <#data-entry-keyers> ;
  schema:text """Computer Programmers, Customer Service Representatives, and Data Entry Keyers rank among the most exposed occupations under the observed coverage measure."""@en .

<#part-growth> a schema:WebPageElement ;
  schema:name "BLS projections and worker characteristics"@en ;
  schema:position 5 ;
  schema:about <#bls-projections>, <#current-population-survey>, <#high-exposure-workers> ;
  schema:text """Higher observed exposure is associated with slightly weaker BLS projected growth and with a worker profile that is older, female, more educated, and higher paid."""@en .

<#part-results> a schema:WebPageElement ;
  schema:name "Initial unemployment and hiring evidence"@en ;
  schema:position 6 ;
  schema:about <#difference-in-differences>, <#young-worker-hiring>, <#job-finding-rate> ;
  schema:text """The report finds no systematic unemployment increase for exposed workers, but sees suggestive evidence of reduced hiring into exposed occupations for young workers."""@en .

<#part-updates> a schema:WebPageElement ;
  schema:name "Why the framework matters"@en ;
  schema:position 7 ;
  schema:about <#counterfactual-framework>, <#future-updates>, <#signal-from-noise> ;
  schema:text """Anthropic frames the work as a repeatable measurement system that can be updated as AI usage, capability, and employment data evolve."""@en .

<#argument-howto> a schema:HowTo ;
  schema:name "How the report measures early AI labor-market impact"@en ;
  schema:description """The report moves from task capability to observed work usage, then compares occupational exposure with employment projections, worker characteristics, unemployment, and hiring."""@en ;
  schema:isPartOf <#article> ;
  schema:step <#step-1>, <#step-2>, <#step-3>, <#step-4>, <#step-5> .

<#step-1> a schema:HowToStep ;
  schema:name "Start with task-level capability"@en ;
  schema:position 1 ;
  schema:text "Use prior estimates of which occupational tasks LLMs can theoretically speed up."@en ;
  schema:isPartOf <#argument-howto> .

<#step-2> a schema:HowToStep ;
  schema:name "Filter through observed work usage"@en ;
  schema:position 2 ;
  schema:text "Use Anthropic Economic Index data to identify which tasks appear in Claude usage in professional settings."@en ;
  schema:isPartOf <#argument-howto> .

<#step-3> a schema:HowToStep ;
  schema:name "Weight automation and task importance"@en ;
  schema:position 3 ;
  schema:text "Give automated uses more weight than augmentation and aggregate task coverage by time spent in each occupation."@en ;
  schema:isPartOf <#argument-howto> .

<#step-4> a schema:HowToStep ;
  schema:name "Compare exposure with projections"@en ;
  schema:position 4 ;
  schema:text "Test whether observed exposure aligns with BLS projected employment growth and worker characteristics."@en ;
  schema:isPartOf <#argument-howto> .

<#step-5> a schema:HowToStep ;
  schema:name "Track unemployment and hiring"@en ;
  schema:position 5 ;
  schema:text "Use CPS data to study unemployment gaps and young-worker job starts in exposed versus unexposed occupations."@en ;
  schema:isPartOf <#argument-howto> .

<#faq-1> a schema:Question ;
  schema:name "What is the report's main contribution?"@en ;
  schema:acceptedAnswer <#faq-1-answer> ;
  schema:isPartOf <#article> .
<#faq-1-answer> a schema:Answer ;
  schema:text "It introduces observed exposure, a measure that combines theoretical LLM capability with real-world Claude usage data."@en ;
  schema:isPartOf <#faq-1> .

<#faq-2> a schema:Question ;
  schema:name "Why is observed exposure different from theoretical exposure?"@en ;
  schema:acceptedAnswer <#faq-2-answer> ;
  schema:isPartOf <#article> .
<#faq-2-answer> a schema:Answer ;
  schema:text "Observed exposure asks which feasible tasks are actually seeing work-related, often automated, AI use."@en ;
  schema:isPartOf <#faq-2> .

<#faq-3> a schema:Question ;
  schema:name "What data sources does the measure combine?"@en ;
  schema:acceptedAnswer <#faq-3-answer> ;
  schema:isPartOf <#article> .
<#faq-3-answer> a schema:Answer ;
  schema:text "It combines O*NET tasks, Anthropic Economic Index usage, and task-level exposure estimates from Eloundou et al."@en ;
  schema:isPartOf <#faq-3> .

<#faq-4> a schema:Question ;
  schema:name "What does the report say about current AI coverage?"@en ;
  schema:acceptedAnswer <#faq-4-answer> ;
  schema:isPartOf <#article> .
<#faq-4-answer> a schema:Answer ;
  schema:text "It says actual coverage remains only a fraction of what is theoretically feasible."@en ;
  schema:isPartOf <#faq-4> .

<#faq-5> a schema:Question ;
  schema:name "Which occupational category example is highlighted?"@en ;
  schema:acceptedAnswer <#faq-5-answer> ;
  schema:isPartOf <#article> .
<#faq-5-answer> a schema:Answer ;
  schema:text "Computer and Math occupations are highlighted, with Claude currently covering 33 percent of tasks in that category."@en ;
  schema:isPartOf <#faq-5> .

<#faq-6> a schema:Question ;
  schema:name "Which occupations rank among the most exposed?"@en ;
  schema:acceptedAnswer <#faq-6-answer> ;
  schema:isPartOf <#article> .
<#faq-6-answer> a schema:Answer ;
  schema:text "Computer Programmers, Customer Service Representatives, and Data Entry Keyers are among the most exposed."@en ;
  schema:isPartOf <#faq-6> .

<#faq-7> a schema:Question ;
  schema:name "How does exposure relate to BLS projections?"@en ;
  schema:acceptedAnswer <#faq-7-answer> ;
  schema:isPartOf <#article> .
<#faq-7-answer> a schema:Answer ;
  schema:text "Occupations with higher observed exposure have somewhat weaker projected employment growth through 2034."@en ;
  schema:isPartOf <#faq-7> .

<#faq-8> a schema:Question ;
  schema:name "Who is more likely to be in exposed occupations?"@en ;
  schema:acceptedAnswer <#faq-8-answer> ;
  schema:isPartOf <#article> .
<#faq-8-answer> a schema:Answer ;
  schema:text "The report says exposed workers are more likely to be older, female, more educated, and higher paid."@en ;
  schema:isPartOf <#faq-8> .

<#faq-9> a schema:Question ;
  schema:name "Does the report find higher unemployment from AI?"@en ;
  schema:acceptedAnswer <#faq-9-answer> ;
  schema:isPartOf <#article> .
<#faq-9-answer> a schema:Answer ;
  schema:text "It finds no systematic unemployment increase for highly exposed workers since late 2022."@en ;
  schema:isPartOf <#faq-9> .

<#faq-10> a schema:Question ;
  schema:name "What early labor-market signal does the report find?"@en ;
  schema:acceptedAnswer <#faq-10-answer> ;
  schema:isPartOf <#article> .
<#faq-10-answer> a schema:Answer ;
  schema:text "It finds suggestive evidence that hiring of workers aged 22-25 into exposed occupations has slowed."@en ;
  schema:isPartOf <#faq-10> .

<#faq-11> a schema:Question ;
  schema:name "Why focus on unemployment?"@en ;
  schema:acceptedAnswer <#faq-11-answer> ;
  schema:isPartOf <#article> .
<#faq-11-answer> a schema:Answer ;
  schema:text "The report treats unemployment as the outcome most directly tied to potential economic harm for displaced workers."@en ;
  schema:isPartOf <#faq-11> .

<#faq-12> a schema:Question ;
  schema:name "How will the framework be used later?"@en ;
  schema:acceptedAnswer <#faq-12-answer> ;
  schema:isPartOf <#article> .
<#faq-12-answer> a schema:Answer ;
  schema:text "Anthropic plans to revisit the analysis as AI usage, model capability, and labor-market outcomes evolve."@en ;
  schema:isPartOf <#faq-12> .