"Invisible labor" is not a metaphor — it is a measurable category of work that the U.S. statistical system has, by design, left out of GDP, productivity figures, and most workforce policy. The largest piece of it is the 53 million Americans the AARP's Caregiving in the U.S. 2020/2024 series counts as unpaid family caregivers, providing an estimated $600 billion of unpaid care annually, the great majority of it performed by women. The next largest is the 2.5 million paid domestic workers (nannies, housekeepers, home health aides) that the Economic Policy Institute estimates lack overtime protection under the Fair Labor Standards Act. Together with content moderators, data labelers, gig-economy "ghost workers," and digital piece-workers, these workers collectively keep the visible economy functioning — and almost none of them appear in the standard "future of work" debate.
The argument of this piece is specific. Invisible labor is invisible by policy choice, not by accident. The New Deal–era exclusions of domestic and agricultural workers from the National Labor Relations Act, the structure of the Bureau of Labor Statistics' surveys, the legal classification of platform workers as independent contractors — each of these is a decision that could be reversed. The cost of reversing them is bounded and the benefits, where measured, are sizable. The reason the reversal hasn't happened is political, not economic.
The data we don't usually see
The OECD has, for decades, attempted to quantify unpaid household and care work in its member countries. Its 2023 estimates, summarized in the OECD Family Database, put the value of unpaid care work at roughly 8–16% of GDP depending on valuation method — comparable to or larger than entire industrial sectors that get extensive policy attention. In the U.S., the BLS American Time Use Survey shows that adults perform an average of roughly 3 hours per day of unpaid household and care work, with women performing about 1.3 hours more than men.
For paid invisible labor, the numbers are starker. The National Domestic Workers Alliance and the EPI's joint analyses of domestic workers find median annual earnings well below the federal poverty line for a family of three, with roughly two-thirds of workers lacking access to employer-provided health insurance. The 2023 NDWA "Domestic Workers Chartbook" documents that approximately 90% of U.S. domestic workers are women and a majority are immigrants or workers of color.
Content moderation — the digital version of invisible labor — has been studied increasingly carefully since Mary Gray and Siddharth Suri's 2019 book Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (Houghton Mifflin Harcourt). Gray and Suri documented that the cleaning of the visible internet relies on a workforce — typically based in the Philippines, Kenya, India, and other low-wage geographies — that views, classifies, and removes hundreds of thousands of pieces of disturbing content daily, often at piece rates below $2 per hour. The 2022 Time investigation into Kenyan workers training OpenAI's content-moderation systems for less than $2 an hour brought the practice into mainstream view but has not, so far, produced wage or working-condition reforms commensurate with the harm.
Why the digital age made this worse, not better
The common assumption is that technology should reduce invisible labor — automation replaces routine work, leaving more humane tasks for humans. The empirical reality is more complicated. Several categories of invisible labor have grown in absolute terms in the digital era, not shrunk.
First, platform-mediated piece-work. Mechanical Turk, Scale AI, Surge, and dozens of similar platforms now coordinate millions of workers performing micro-tasks: image labeling, transcription, content moderation, AI-output evaluation. A 2024 ILO report estimated the global digital labor platform workforce at roughly 435 million workers, the great majority earning at or below minimum-wage rates in their geographies, with negligible employment protection.
Second, AI training labor. Reinforcement learning from human feedback (RLHF) — the technique that turned base language models into ChatGPT — is built on the work of human raters and red-teamers, typically employed through subcontractors and often outside the legal frameworks that would apply to direct employees. This is, structurally, the most invisible labor of all: workers whose efforts are embedded in products used by hundreds of millions, with neither attribution nor durable employment.
Third, the offshored back office. The customer-support agents who handle U.S. banking, telecom, and tech-product calls from Manila, Hyderabad, and Bogotá; the medical transcriptionists who clean U.S. radiology reports overnight; the freelance graphic designers and copywriters on Upwork and Fiverr competing on global rates — each of these constitutes a layer of labor that is essential to a "visible" U.S. product or service and almost entirely absent from how Americans think about the U.S. workforce.
The interventions with the strongest evidence
Bring excluded workers under existing labor law
The single most consequential policy lever is extending FLSA overtime protection, OSHA workplace-safety jurisdiction, and NLRB organizing rights to currently excluded categories. The Domestic Workers Bill of Rights model — first enacted in New York in 2010 and now in roughly a dozen states — provides a template. Empirical evaluations of the New York implementation by the Center for American Progress have documented improved working conditions without measurable employment loss.
Algorithmic transparency for platform workers
The EU Platform Work Directive (2024) requires that any algorithmic system materially affecting platform workers' conditions be subject to human oversight, transparency, and worker consultation. State-level equivalents in California, New York, and Illinois are emerging. NWLB's flagship analysis of where this fight is settling is at The Gig Economy Settlement →.
Care-work investment
The most under-discussed leverage point in the U.S. economy may be care-work compensation. CMS data show the median home health aide wage at roughly $16/hour in 2024, well below most retail and warehouse alternatives. The result is a chronic labor shortage in the category the BLS projects to add the most jobs of any over the next decade. Public investment to raise wages and credential workers in care occupations has, in early state-level experiments (notably Washington's Initiative 1163 and several Medicaid demonstration projects), produced both better worker retention and better patient outcomes. We cover this in depth at The Caregiver Workforce →.
Statistical visibility
This sounds modest but matters: the BLS does not produce a routine, granular series on platform work, domestic work, or unpaid care work. The 2017 BLS Contingent Worker Supplement was the last major federal effort, and even that undercounted platform workers. Restoring and expanding that survey, with adequate sampling of immigrant and home-based workers, would do more for invisible-labor policy than most direct interventions because it would make the data unmistakable.
What firms can do that doesn't wait for policy
Several practical moves are available to firms that want to take this seriously without waiting for legislative change. Publish supply-chain labor practices, including for contractors and content-moderation vendors. Pay direct rates to AI-training workers rather than allowing layers of subcontracting to obscure them. Bring critical "invisible" functions back onto the payroll where they can be properly compensated and credentialed. Each of these is doable; the firms that do them tend to be more durable employers because they understand their actual cost structure.
Invisible labor isn't invisible because of technology. It's invisible because the U.S. statistical, legal, and tax systems were built to leave it out. Each of those choices can be reversed. None of them yet has been.
What is at stake
The conversation about invisible labor matters not because we expect rapid policy change — we don't — but because the visible economy depends on it in ways that are now beginning to break down. The shortage of nurses, home health aides, and CNAs is the single largest workforce constraint facing the U.S. healthcare system. The dependence of leading AI systems on under-compensated content moderators and human raters is a reputational and operational risk for the firms involved. The unpaid care load that falls disproportionately on women remains the largest single explanation for the persistent gender pay gap in high-wage occupations, per Claudia Goldin's Nobel-recognized work on this question.
Bringing invisible labor into view doesn't, on its own, fix any of this. But it makes the policy and operational choices visible — and visible problems are the only ones that get solved.
Updated May 21, 2026. This piece was substantively rewritten as part of NWLB's 2026 editorial refresh.
