evidoria

← Back to browse

Good practice Imported

AprendES — El Salvador's World Bank-Financed AI-Tutor Pilot Claims PISA Scores Rivaling Germany and Sweden, but Selection Bias Clouds the Comparison

El Salvador · San Salvador · See the El Salvador profile · See the San Salvador profile

Evidence: Descriptive / self-reported Top 56% 47/100 · Ask Evidence Copilot about this practice

A World Bank-financed AI-tutor pilot in 171 Salvadoran public schools posted PISA-for-Schools results the government says rival Germany and Sweden, but the sample was self-selected by infrastructure and willingness, not randomised, and the national PISA 2025 showed far smaller ga

1,198 students
Students tested in the PISA-for-Schools pilot comparison (June 2026)
171 schools
Pilot schools participating in the PISA-for-Schools comparison (June 2026)
1,000+ schools
Public schools reached by AprendES (September 2026)
501 USD million
World Bank financing committed (approved March 2026)
7,073 students
Students tested in the nationally representative PISA 2025 (2025)
278 schools
Schools in the nationally representative PISA 2025 sample (2025)
AprendES — El Salvador's World Bank-Financed AI-Tutor Pilot Claims PISA Scores Rivaling Germany and Sweden, but Selection Bias Clouds the Comparison

Details

Maturity
Scaling
Promoter
Ministerio de Educación, Ciencia y Tecnología de El Salvador (MINEDUCYT) / World Bank
Period
2025-2031
Keywords
AI tutoring, PISA assessment, World Bank financing, education modernization, public schools

Context

AprendES ('Acelerando el Aprendizaje en El Salvador con Innovación Digital') is a US$501 million World Bank-financed education modernisation project approved in March 2026, run by El Salvador's Ministry of Education, Science and Technology (MINEDUCYT) under Minister Karla Trigueros.

Objectives

Bundle devices, connectivity, digital platforms, teacher and principal coaching, pedagogical materials and an AI tutoring tool that generates exercises, answers questions and adapts to each student's level, with a government ambition to reach all public schools within 18 months.

Activities

The pilot began in mid-2025 in a small set of schools and had expanded to more than 1,000 public schools by September 2026. In June 2026 the OECD's PISA-for-Schools instrument was administered to 1,198 volunteer students across 171 participating schools, chosen for infrastructure readiness, connectivity and willingness to participate rather than at random.

Results

President Bukele announced that participating pilot schools scored above the national average in reading, mathematics and science, comparable to Germany's and Sweden's PISA 2022 averages. The government's own nationally representative PISA 2025 exercise (7,073 students across 278 schools) showed scores stable in mathematics and reading, with improvement only in science — a far more modest picture.

Conclusions

No independent evaluation has isolated what share of the pilot schools' gain, if any, is attributable to the AI tutor specifically versus the other bundled inputs or to the non-random selection of better-resourced, willing schools. Bukele himself acknowledged the results 'do not prove that every Salvadoran school has been transformed,' and the World Bank called them 'encouraging' while cautioning that scaling to the full system remains 'an immense effort.'

Implementation

Indicative cost
Very high (> €5M) — US$501 million in World Bank financing approved March 2026, covering devices, connectivity, platforms, coaching and the AI tutor across a planned 2025-2031 rollout.
Time to results
Long (> 3 years) — Pilot began mid-2025; expanded past 1,000 schools by September 2026; government target of reaching all public schools within 18 months of that point.
Staffing & skills
Ministry of Education, Science and Technology (MINEDUCYT), led by Minister Karla Trigueros, World Bank (financing and technical oversight)

Conditions for success

  • Bundling devices, connectivity, teacher/principal coaching and pedagogical materials alongside the AI tutor, rather than deploying AI alone
  • A government commitment to a phased pilot-to-national rollout timeline

Common failure modes

  • A self-selected, non-random school sample (chosen for infrastructure readiness, connectivity and willingness) confounds any comparison to national performance
  • No independent evaluation isolates the AI tutor's specific contribution from the rest of the bundled inputs
  • The nationally representative PISA 2025 results were far more modest than the touted pilot comparison

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful