The United States has multiplied certifications to open the labor market beyond the university degree. The country has more than 1.1 million credentials in total, of which micro-credentials constitute a category with poorly coherent definitions. Employers have made few changes to their hiring after the removal of the diploma requirement. AI makes it harder to verify actual skills.
The essentials
- Multiplying signals of competence without standardizing them produces fragmentation rather than transparency: more than 1.1 million credentials, including 6,892 certifications in the strict sense, are offered in the United States, in a system where non-degree certifications are validated in a fragmented way (Third Way, 2026).
- According to the study cited by Third Way, 45% of companies that removed diploma requirements from their job postings made few changes to their actual hiring practices, though this study does not establish that this stems exclusively from a lack of comparable signals.
- Employers surveyed believe that generative AI makes it harder to evaluate actual skills and is associated with a reduction in entry-level hiring.
- Portable and verifiable proof of competencies can promote recognition of alternative training, which an interoperable infrastructure can support.
- Several trajectories could emerge for the next decade, producing very different results for social mobility.
The diploma persists because alternatives are unreadable
Hiring based on skills retains potential, but its implementation remains uneven and requires more standardized tools.
For about a decade, the discourse on “skills rather than diplomas” has gained traction in American businesses. A growing number of large technology companies have removed the diploma requirement for certain job categories. Governors have launched alternative apprenticeship programs. Online training platforms have proliferated, offering quick certificates on thousands of subjects.
The market responded to this demand: it produced more than 1.1 million credentials in total, including 6,892 certifications in the strict sense, with definitions, curricula, requirements and validation mechanisms that are often heterogeneous and insufficiently standardized. A certificate in “project management” issued by a public university in the Midwest, a private platform in San Francisco, and a local chamber of commerce have only the name in common. A recruiter may lack comparable information for many credentials, but comparison tools exist when data are published in standardized formats.
Third Way published a report on May 20, 2026, documenting this phenomenon: the expansion of the credentials market increases complexity without infrastructure for transparency and comparability. Employers receive too much information that is too heterogeneous to draw actionable conclusions from it. Faced with this informational noise, the university diploma remains used by a significant portion of employers as a filter because of the persistent fragmentation of alternatives.
The diploma survives as a signal of competence not because it is superior, but because alternatives are not yet readable at scale.
AI changes the conditions of verification
Generative AI has added a layer of complexity that the designers of micro-credentials had not anticipated. When a candidate submits a project, report, or analysis as part of an online certification, distinguishing what they produced themselves from what a tool generated for them remains difficult. Some employers consider that AI makes actual skills harder to evaluate, an issue set to intensify as tools improve.
The problem goes beyond individual deception. Students who use AI to complete certification exercises are using tools that their future employers expect them to master. The issue concerns their ability to direct the tool, interpret its results, and correct its errors. Current evaluation formats are often not designed to measure these abilities.
This gap also affects traditional certifications. Uncontrolled access to generative AI can compromise the interpretation of an online multiple-choice test as a measure of only the candidate’s competence. Evaluation formats that have survived decades of distance learning stumble over a new challenge: generative AI partly shifts evaluation toward verification of mobilization, judgment, and use of knowledge, without that information being universally available.
The wage bifurcation that AI is already producing in the labor market makes this competence verification problem all the more urgent: skills for managing AI tools are remunerated differently from competencies that can be substituted by them. Identifying which ones a candidate truly possesses becomes a high-stakes economic decision for the employer.
Experiments pointing the way
The picture is not uniform. Some states and some companies have begun building concrete responses, even if they remain fragmented.
The Trusted Career Profile, a digital portfolio of skills that aggregates training, certifications, and work experience in a standardized format verifiable by employers, was developed and released by HR Open Standards in collaboration with the T3 Innovation Network and ecosystem partners. The principle: create a common infrastructure for portability, rather than letting each certification issuer define its own format. Preliminary results suggest better readability for recruiters, but adoption remains limited at the state scale.
On the company side, sectors such as cybersecurity have developed relatively robust sector-wide certification frameworks, CompTIA, CISSP, which function as de facto standards recognized independently of the issuing institution. The lesson: portable certifications emerge when an entire sector agrees on what a competency means, not when each actor optimizes for its own interests.
The European Union built the European Qualifications Framework, a common grid with eight levels that allows comparison of qualifications issued in different countries. This framework has not solved all mutual recognition problems, but it has created a shared language that employers and training systems can use. The European framework offers the possibility of linking American micro-credentials for transatlantic mobility, a path that remains exploratory, but that points in a direction.
The question of institutional capacity to orchestrate these transitions is raised here acutely: building an infrastructure for portability of skills raises the question of governance and the capacity to define standards without capture by private interests.
The public architecture that the market will not produce on its own
The certification market has demonstrated its ability to produce diversity. It has not demonstrated its ability to produce portability. These two objectives are not naturally compatible: each certification issuer has an interest in their own certification being valued, not in it being easily comparable to that of a competitor.
The Third Way report notes the lack of common language between employers and certification issuers and calls for building an infrastructure that makes credentials transparent, comparable, and aligned with their value in the labor market.
These two elements form a public infrastructure in the precise sense of the term: a common good whose value increases with the number of users, and that no one has an interest in financing alone. The United States has built similar infrastructures in other domains, accounting standards, communication protocols, property registries. The challenge for certifications is analogous.
It is not without risks. Any common dictionary of competencies reflects choices about what matters and what does not. Large technology companies, universities, and private training organizations each have different interests in the definition of these standards. The governance of the infrastructure is as important as its technical design.
Scenarios for 2030 depending on the path taken
Several trajectories could emerge for the next decade, producing very different results for social mobility.
In the first, federal action could focus on the adoption of data standards and transparency of certifications, potentially improving the comparability of proof of competencies. Certifications issued by different actors would become more comparable if they referenced the same definitions. A worker from Ohio who obtains a certification in robot-assisted welding can have it recognized in Texas or Wisconsin without additional procedure. The cost of transition between jobs could fall. The salary premium of the university degree could shrink for occupations where certified competencies are verifiable.
The social elevator through alternative training becomes measurable.
In the second trajectory, fragmentation of the current system persists. Online training platforms continue to multiply and issue their own certifications. Large employers build their own internal tests. Small and medium-sized businesses could maintain the diploma as a default filter in the face of the complexity of evaluating multiple types of certifications. Alternative training would remain a more restricted market.
The risk of polarization that automation poses for workers without recognized qualifications also materializes in this scenario, but through the certification system rather than the labor market directly.
Three signals will make it possible to distinguish the actual trajectory in the coming years: the share of employers who accept micro-credentials without requiring a university diploma in parallel; the adoption rate of portable skill portfolios like Indiana’s Trusted Career Profile beyond the state that created it; and the evolution of the salary premium between graduates and non-graduates with equivalent measured competencies. If these three indicators converge toward the second trajectory by 2028, the political cost of correction will be higher.
Education at a Glance 2026 deals primarily with teacher shortages and indicators of education systems; the portability of qualifications appears as an obstacle to professional mobility in several member countries. The United States is not alone in this situation, but the dispersion of its federal system makes the problem more acute than in Europe, where national frameworks already offer a first level of standardization.
Actions accessible to employers right now
Waiting for a federal ontology to reform hiring practices is not a strategy. Some American companies have begun to act without waiting.
Several experiments combine partnerships with training organizations to define certifications and practical evaluations during recruitment to verify skills beyond the title. IBM has published its competency frameworks for several technical fields, allowing candidates to understand precisely what the company expects and trainers to adapt their programs. This model creates predictability for all parties.
AI can also be part of the assessment solution, not just the problem. Adaptive assessment tools, which adjust questions in real time according to responses, make delegation to an external tool less effective and measure reasoning rather than memorization, are beginning to be deployed by platforms like Khanmigo for education and by specialized recruitment companies. These formats are not perfect, but they represent a more robust direction than static multiple-choice tests in the face of generative AI.
Two questions remain open, that neither employers nor training organizations can solve alone. The first concerns the trade-off when competency definitions conflict between sectors. The second concerns ensuring that the portability infrastructure remains accessible to the least qualified workers, and not only to those who already have the means to navigate a complex certification market.
Sources
- Third Way / National Governors Association, Building the Map: A Federal Policy Agenda to Fix America’s Credentialing System (2026)
- OECD, Education at a Glance 2026 (Organisation for Economic Co-operation and Development)
- National Governors Association, Credentialing Briefing, January 2026
