In a non-representative sample of 200 institutions, 86.5% report using AI in at least one area, but only one in four has formalized a strategy for doing so. This UNESCO survey of 200 institutions across 19 countries measures a gap that is not merely organizational: it is the gap between what a diploma still certifies and what it will certify in ten years. Institutions that have not defined their position risk seeing their certification become less legible in the eyes of the job market.
The essentials
- 87% of surveyed institutions use AI, but only 26% have a formal strategy: the tool precedes governance (UNESCO IESALC, 2025).
- The real risk for universities is that AI redefines expected competencies before they have redefined what they evaluate.
- Leading institutions, a handful of pioneering universities in Brazil, Mexico, and Argentina, are betting on oral assessment, situated projects, and transparency of use rather than prohibition.
- For young people without diplomas already in precarious positions, a devaluation of the university diploma would worsen an already documented inequality.
- The window to act is short: models advance faster than the usual cycles of curriculum reform.
87% usage, 26% strategy: what the gap reveals
The survey was conducted in 2025 among 200 higher education institutions in 19 countries in Latin America and the Caribbean, and its results were published in 2026. It covers reported uses by teaching staff, students, and administrative services, as well as formal mechanisms put in place by leadership.
The 87% usage figure is not surprising. The tools are free, accessible on smartphones, and their adoption requires no institutional authorization. What is more striking is the disproportion: among institutions using AI in the sample, the majority have no formal AI strategy. No revised academic integrity policy, no teacher training on detection or pedagogical integration, no formalized reflection on what the tool changes about assessment.
This gap is not specific to Latin America. But the region experiences it with particular intensity for two reasons. The first stems from demographic pressure: Latin American universities have massified access over the past two decades, with cohorts of students often more precarious, less equipped with study methods, and more exposed to the temptation to delegate to AI what universities are slow to teach. The second stems from resources: public institutions in the region have narrower budgetary margins to recruit pedagogical specialists or finance curriculum experiments than their North American or European counterparts.
The gap between usage and strategy is therefore not an administrative delay. It is a diagnosis of institutions’ capacity to steer a change they did not initiate.
AI transforms assessment without institutional decision
A law course, a sociology assignment, an internship report in management: three exercises that, since 2023, can be written in minutes by a correctly prompted language model. Teachers know this, students know this. Institutions have not yet decided what this changes about the value of the exercise.
Cheating existed before ChatGPT, but the situation is now structural. When a written output can be generated by AI without control of the process, its validity as autonomous proof of acquired knowledge must be reassessed. If the assessment mechanism does not adapt, the diploma’s credibility as a certificate of competence could weaken.
The 74% of institutions without a strategy pose a concrete problem. The absence of a framework does not eliminate usage: practices develop without the institution having defined what it wants to certify. Some teachers ban AI, others encourage it, others ignore it. In the same university, a student can spend four years in a contradictory environment, never knowing what their diploma attests to in the eyes of the job market.
The fundamental risk lies there. A prolonged divergence of practices without a common framework risks blurring the diploma’s readability as a signal. Employers then compensate by adding their own tests, valuing alternative certifications more or probationary internships more. This movement is already visible in tech recruitment in North America. It could extend to other sectors and regions if universities do not regain control.
Pioneers redefining what they evaluate
A few institutions in Latin America have begun to answer this question without waiting for national regulation to compel them. Their approaches converge around three axes.
The first is the shift toward oral assessment and process. Universities in Brazil, notably in engineering and humanities programs, have increased the share of defense presentations and individual interviews, where deep understanding cannot be delegated to a model. AI can write the report; it cannot defend methodological choices before a jury.
The second axis is transparency of use as a competency in its own right. Some programs in Argentina and Mexico ask students to document how they used AI in their work, which prompts, what verifications, what modifications. The objective is not to track usage but to make it visible and reflective. In doing so, these programs evaluate something real: the capacity to work with a tool while retaining one’s own judgment.
The third axis is teacher training. According to the survey, the main obstacle is lack of financial resources, followed by insufficient staff expertise and infrastructure limitations. Teachers who do not know what AI does well and poorly cannot design assessments that account for it. A few universities have launched short internal training programs focused on pedagogy rather than technique.
These initiatives remain minority. But they show that the path exists and that it does not require resources beyond reach. What it requires is a decision: choose what the diploma still certifies and build the assessment accordingly.
The governance challenge UNESCO points to without resolving
The IESALC survey is not limited to statistical findings. It formulates several recommendations, three of which deserve to be weighed.
The first concerns national policies. Many countries in the region do not yet have a regulatory framework for AI in higher education. UNESCO recommends that education ministries build national guidelines, not to standardize practices but to set a floor of requirements—academic integrity, transparency, training—below which no institution should operate. This is a reasonable recommendation, but the timeframes for producing regulations in the region often exceed cycles of technological evolution.
The second recommendation concerns interinstitutional cooperation. The survey reveals that institutions that have developed a strategy often did so in networks, sharing experiences and pedagogical tools. A regional platform for exchanging best practices, which IESALC could facilitate, would accelerate the diffusion of approaches that work without imposing a single model.
The third recommendation concerns equity. Better-resourced private institutions advance faster. Public institutions, which serve more low-income students, and specialized institutions are the most vulnerable to these transformations. The gap in AI preparedness between public and private institutions risks aggravating existing inequalities. This is a risk the survey names but which public policies in the region have not yet addressed.
The analogy with what Europe is building to regulate AI in the workplace is instructive: collective bargaining made it possible to impose oversight where companies were advancing without governance. Universities could draw inspiration from this model to ensure that AI adoption does not happen at the expense of the most vulnerable.
The value of the diploma in 2035 depends on decisions made between now and then
The 2035 horizon corresponds to two complete cycles of curriculum reform in most higher education systems. Decisions made or not made between now and then will influence the value of the university diploma for generations to come.
Two trajectories are emerging, neither of which is inevitable.
In the first, institutions that acted early build new competency frameworks. The rise of AI prompts the design of more authentic assessments, notably including critical judgment, reasoning, and sometimes oral formats. The diploma then certifies an enhanced competency rather than text or calculation production alone. Employers learn to read these new signals, and the diploma regains or maintains its readability. This trajectory requires institutional will, resources to train teachers, and cooperation between institutions to avoid fragmentation of standards.
In the second trajectory, institutions vacillate. Practices diverge, employers lose confidence in the diploma signal, and alternative certifications—competency badges, on-demand tests, digital portfolios—fill the void. This movement already exists in certain tech sectors in North America. If it extends to other sectors, the university diploma does not disappear, but it ceases to be the sole credible signal. Students with means multiply certifications; others rely on an increasingly contested diploma.
Current gaps in AI preparedness risk aggravating structural inequalities in higher education.
What would allow a shift from one trajectory to another is less national regulation than a movement by institutions themselves. The signals to watch are clear: the proportion of institutions that revise their academic integrity policy to explicitly integrate AI; the evolution of the share of oral and process-based assessments in curricula; and the speed at which ministries publish guidelines—not prohibitions, but frameworks.
The good news, if one can formulate it, is that Latin America is not behind on a problem others would have solved. The region is experiencing the same acceleration as Europe or Southeast Asia, with the same gap between adoption and governance. It has the same window to act, and its pioneers show that action is possible at constant means. What is missing is less money than the will to decide.
The signal recruiters are beginning to send
Companies recruiting diploma holders in Latin America are not yet in open rupture with universities. But several documented signals merit attention.
In certain tech and consulting sectors, several companies in Latin America have added practical tests to their recruitment processes, not to replace the diploma but to complement it. The reasoning is explicit in several HR manager interviews reported by regional business press: the diploma normally attests to the successful completion of a program and its requirements, without automatically guaranteeing that the person masters all the competencies the program claims to teach.
This movement remains marginal. It does not concern sectors that recruit in volume—education, health, public administration—where the diploma remains the dominant criterion. But in certain sectors where AI occupies a central place in daily work, pressure on universities to clarify what they certify comes from the market as much as from regulators.
For students entering this market in five or ten years, the question is already concrete: what to do with a diploma whose readability is being negotiated without yet knowing in which direction? The honest answer is that they have every interest in taking AI seriously as a competency, not in using it to circumvent learning, but in learning to work with it in a visible and reflective way, regardless of their institution’s policy. This capacity, more than the diploma alone, is what employers are already seeking to measure, and what universities acting early have begun to certify.
Sources
- UNESCO IESALC, AI Implementation in Higher Education in Latin America and the Caribbean (2025)
- UNESCO IESALC, ILIA 2025 (Informe sobre la Inteligencia Artificial en la Educación Superior en América Latina y el Caribe)
- Journal d’un Progressiste, Young people without diplomas are losing ground while public money finances the others
- Journal d’un Progressiste, Collective bargaining: Europe opens oversight of AI in the workplace
