Schools Are Racing to Teach AI While Reading Scores Keep Falling

Last updated: July 1, 2026

On June 18, 2026, the European Commission and the OECD released the most ambitious AI literacy blueprint ever built for K-12 education: 19 competences across four domains, shaped by more than 2,000 educators, policymakers, and researchers from over 100 countries. Its first domain asks students to “critically evaluate AI-generated information” and identify potential bias in AI outputs. That is a reading comprehension skill. In the United States, 40 percent of fourth graders cannot clear the National Assessment of Educational Progress Basic reading benchmark — the worst result since 2002. A blueprint for AI literacy that lands in a classroom where kids cannot read proficiently is a roof without walls.

TL;DR

The European Commission and OECD launched the AILit Framework on June 18, 2026: 19 competences across four domains for AI literacy in K-12, developed with 2,000-plus stakeholders from over 100 countries.
The framework’s first domain, Engage with AI, asks learners to critically evaluate AI-generated content and identify bias — skills that require reading comprehension, working memory, and processing speed.
In the U.S., 40 percent of fourth graders scored Below Basic on NAEP reading in 2024, the largest share since 2002; 69 percent scored below Proficient.
Reading science (Geary, Dehaene, Wolf) and the IDA 2025 multi-system definition establish that foundational cognitive processing skills must be explicitly built — no technology layer delivers them automatically.
If your child has foundational processing gaps, ask whether AI tools in their classroom are building evaluative skills or compensating for ones never built.

Europe and the OECD launched a 19-competence AI literacy blueprint for every K-12 student. Before your child’s school rolls it out, there is a more urgent question to answer.

Common questions

Does it matter if this framework is from Europe? My child goes to school in the US.

It matters more than it appears. The AILit Framework feeds directly into the PISA 2029 Media and AI Literacy assessment — the same international test the US participates in. That means within three years, AI literacy competences will shape how US students are benchmarked against international peers. The US has historically adopted frameworks that gain PISA traction. These competences will reach American district curriculum decisions even without the EU name on the cover.

What foundational skills does AI literacy actually require?

The AILit Framework’s first domain, Engage with AI, asks learners to critically evaluate AI-generated content, identify potential bias, and assess source reliability. Those are reading comprehension and critical thinking tasks. Managing AI requires self-regulation and metacognition. Shaping AI requires sustained written argument and civic reasoning. All of these rest on working memory, processing speed, phonological awareness, and auditory processing — the same multi-system foundation the IDA 2025 definition identifies as essential for reading development.

My child struggles with reading. Should I be worried about AI tools in their classroom?

Not automatically — but pay attention to what the AI is doing. A tool that provides scaffolding while a child builds decoding skills is different from one that reads all text indefinitely. The first supports skill-building; the second replaces the expectation the skill gets built. If you want to understand your child’s processing picture, a screener gives you a starting point today. Keep in mind: a screener is not a diagnosis. If your child might need formal accommodations such as an IEP or 504 plan, or you suspect a vision, hearing, or medical cause, a professional evaluation is the route to those supports.

Is the AILit Framework saying AI should replace reading instruction?

No — the framework is explicit that AI cannot replace human judgment, care, or relationships, and it trains students specifically to recognize when not to use AI. The concern here is not about the framework’s intent but about how “AI literacy” gets implemented in schools that have not addressed foundational literacy gaps. The framework is a competence map. It does not close the NAEP reading gap. That work still has to happen first.

The EU and OECD built a 19-skill AI literacy roadmap for every K-12 child. Its first ask: critically evaluate what AI tells you. 40% of US 4th graders cannot clear the basic reading benchmark. That is the real gap. Tweet Now Follow Learning Success

What the EU and OECD built

The AI Literacy (AILit) Framework for Primary and Secondary Education launched on June 18, 2026 during a flagship event in Brussels organized by the European Digital Education Hub. The initiative is a joint product of the European Commission and the Organisation for Economic Cooperation and Development, with support from CodeAI (formerly Code.org) and a network of international researchers and educators.

The framework organizes AI literacy into four domains: Engage with AI, Create with AI, Manage AI, and Shape AI. Each domain contains competences that span knowledge, skills, and attitudes, with Learner Expectations describing what students should be able to do at basic, intermediate, and advanced levels. A draft released in May 2025 drew feedback from more than 2,000 stakeholders across more than 100 countries on every continent. The finalized version places explicit emphasis on metacognition, self-regulation, and the judgment to recognize when not to use AI — treating refusal as a legitimate and responsible choice.

The framework connects directly to the PISA 2029 Media and AI Literacy assessment, meaning these competences will shape international benchmarking within three years. The European Commission cites data showing 68 percent of teenagers already use AI, while education systems have lacked the frameworks to integrate it thoughtfully. The AILit Framework is designed to fill that gap.

What the coverage gets wrong

Reporting on the AILit Framework has framed it as a tech-readiness story: schools must prepare students for the AI era. The better frame is about what the framework assumes is already in place. Its first domain asks students to critically evaluate AI outputs, identify bias, and assess reliability. Those require advanced reading comprehension and working memory skills. The 2024 NAEP data shows 40 percent of U.S. fourth graders cannot clear the Basic reading benchmark — the worst result in two decades. Reading science (Geary, Dehaene, Wolf; IDA 2025 multi-system definition) is clear that those foundational processing systems must be explicitly built; no technology layer delivers them automatically. An AI literacy curriculum placed on top of unaddressed processing gaps does not produce critical evaluators; it produces students who generate and accept AI outputs without the cognitive tools to question them, which is precisely what the AILit Framework was designed to prevent.

The assumption the framework cannot make for you

The framework is genuinely useful. Its emphasis on agency, judgment, and the question of when not to use AI is rare in technology-in-education policy documents, and its classroom examples ground the competences in real practice across subject areas. But every competence in the Engage with AI domain — critically evaluating AI outputs, identifying bias, assessing reliability, choosing the right tool — requires something the framework takes for granted: that students have the foundational cognitive processing skills to do that evaluative work. They do not all have them.

Reading science has made this point clearly for decades. David Geary’s framework of biologically primary versus secondary knowledge establishes that spoken language is something children absorb naturally; reading is a roughly 5,000-year-old cultural invention the brain never evolved for, and it has to be explicitly and systematically taught. Stanislas Dehaene’s neuroimaging work shows reading works by repurposing circuits built for other tasks. Maryanne Wolf put it plainly: “We were never born to read.” None of that changes when a child sits in front of an AI tool instead of a workbook. The foundational processing systems — phonological awareness, working memory, auditory processing, processing speed — have to be built first. The International Dyslexia Association’s 2025 definition explicitly names this multi-system reality: reading difficulties involve combinations of cognitive, neurobiological, and environmental factors, none of which an AI tool addresses by default.

The same cognitive micro-skills that stall a child’s reading development — working memory, processing speed, auditory discrimination — are the precise skills that “Engage with AI” requires. A child who uses an AI tool to read text they cannot yet decode, or to summarize content they cannot yet evaluate, is not building AI literacy. They are outsourcing the cognitive work the foundation was supposed to build. The system that adds AI competence frameworks on top of unaddressed foundational gaps does not produce AI-ready students; it produces students who are dependent on the tool for the thinking the tool was meant to support.

Key Takeaways:

The framework is strong; the assumption is the problem: The AILit Framework’s core competences — critical evaluation, metacognition, when not to use AI — all require foundational cognitive processing skills that 69 percent of U.S. fourth graders have not yet mastered.

Reading is not optional prep for AI literacy: Reading science (Geary, Dehaene, Wolf; IDA 2025 multi-system definition) shows foundational processing skills must be explicitly built; the same working memory and processing speed gaps that stall reading development stall the evaluative work AI literacy demands.

Ask your school three questions: Is AI being used to build evaluative skills or bypass them? Does the school know your child’s foundational processing gaps? When AI compensates for a missing skill, what is the plan to build it?

What to ask before your child’s school goes AI-first

If your child struggles with reading, working memory, or processing speed, AI tools rolling out in their classroom are not automatically useful — and they carry a specific risk: masking the gap instead of closing it. A child who writes prompts and accepts outputs without the evaluative foundation to question them is not more capable. They are more dependent.

Three questions worth asking your school before AI literacy becomes the headline initiative: Is the AI being used to build evaluative skills, or to bypass the need to develop them? Does the school know whether my child has foundational processing gaps that affect how they engage with and read text? And when an AI tool compensates for a missing skill, what is the plan to actually build that skill?

The AILit Framework’s final domain, Shape AI, asks learners to influence how AI systems are designed and governed. That requires sustained critical argument, text comprehension, and civic reasoning. Those are exactly the skills that fall apart when foundational processing gaps go unaddressed for years. A child who arrives at ninth grade with unresolved phonological, working memory, or auditory processing challenges does not become a shaper of AI by learning to prompt it. They become a user who accepts what AI tells them, which is what every technologist who built the framework would say they were trying to prevent.

Every parent deserves to know whether their child’s school is building the foundation AI literacy requires, or piling AI tools on top of gaps no one has named. The obstacle is not a shortage of technology. It is a system that keeps adding layers to a foundation that has not been assessed or repaired. You are the one who notices that your child avoids the reading, not the tool. Learning Success All Access starts with a multi-system look at which processing skills need work, then gives you the program to build them — so the AI literacy curriculum landing in your child’s classroom actually has something solid to land on.