Key points:
For many schools, the path to AI literacy remains unclear. The challenge is not simply AI’s rapid evolution, but its ability to generate answers almost instantly, forcing educators to reconsider how learning is guided and measured. The common response has been to teach students how to use AI tools, but that approach begins in the middle of the problem. Schools must first identify the foundational principles that should guide AI education.
Without that foundation, schools and instructors are left to navigate structural questions about curriculum, learning pathways, and the role of AI in existing and future courses. Many educators favor strict limits on its use, and their concerns deserve serious consideration. Yet a broader shift toward deliberate integration is already underway. The task is no longer simply to restrict AI, but to build curricula that address its risks while recognizing its growing presence in education and daily life.
AI is no longer optional
AI is not a fad, and education must confront the realities it presents. Outright bans will be futile and short-lived. Restrictions may delay students’ engagement with AI, but they cannot prevent students from encountering it in search engines, software, and everyday digital tools. Education must therefore teach responsible use rather than pretend AI can be excluded. AI is now woven into digital technology, much as threads form a fabric. Even simple online search increasingly begins with an AI-generated response. Although AI can produce incorrect answers, users are clicking through from search results to traditional web pages less often. This shift suggests that AI is becoming an accepted part of daily life, whether users consciously recognize the transition or not.
This shift demands stronger digital literacy education from primary school through higher education. As AI applications multiply and their capabilities expand, curriculum must evolve with the rapidly changing digital environment students already inhabit.
AI literacy within digital literacy
Digital literacy has evolved for decades. Before large language models, or LLMs, and widely available AI, it focused primarily on hardware, software applications, the internet, and the use and evaluation of digital content. Those skills remain essential to academic success and modern life, but AI literacy is now an integral part of digital literacy. As AI assumes more routine tasks, some older technical skills may receive less curricular attention. That shift can be constructive if education balances necessary procedural instruction with stronger judgment, verification, critical thinking, and learner agency.
Some software applications still present a learning curve steeper than Mount Everest. AI can level the playing field by giving complex applications a natural-language interface. In 2025, Microsoft CEO Satya Nadella predicted that apps would collapse because they are essentially elaborate interfaces layered over databases, while AI agents would become the primary connection between people and computers. He did not mean that applications would disappear. Rather, AI interfaces would absorb much of their functionality. Instead of clicking buttons, navigating ribbons, writing syntax in programs such as Excel, or memorizing a maze of menus, users will increasingly ask AI to complete tasks. Software agents have served users for years, most visibly through mobile apps that act on a user’s behalf and often eliminate the need to open a browser for a specific task. AI agents extend that model by interpreting requests, selecting actions, and working toward a goal.
AI offers the promise of making powerful applications accessible to far more people. Opening an unfamiliar program can be daunting when its interface is crowded with buttons, icons, specialized functions, and cryptic terminology. I look forward to the day when users can complete tasks that now require considerable software expertise with little or no learning curve. This shift does not diminish capability. It expands access to it.
Learner agency is the cornerstone
As authentic student work becomes more difficult to distinguish, learner agency must become the cornerstone of AI literacy. Teaching students about learner agency and engaging them in meaningful discussion about it provide the most logical starting point for AI literacy education. This principle should shape every level of education, especially higher education. Learner agency means taking ownership of one’s education and recognizing that each student is responsible for personal growth and development. Instructors and institutions can guide students toward their goals, but they cannot complete the journey for them.
It may be possible to navigate courses and earn credit without truly learning, but the consequences will eventually surface, perhaps sooner than a student expects. Courses are often scaffolded, and the knowledge and skills developed in one course become prerequisites for success in the next. Students must also understand that the world of work is highly skill-based. Success depends not only on content knowledge in a given field, but also on transferable skills such as effective writing, communication, critical thinking, and problem solving. A college degree may provide entry, but continued success depends on the ability to perform.
Additional measures of competence, including examinations, licensure requirements, and professional credentials, will likely become more important across the employment landscape. Although employers may seek an AI-literate workforce, AI proficiency cannot replace the subject knowledge, disciplinary expertise, and broad intellectual skills traditionally associated with a degree or credential. The ultimate test remains job performance. Learner agency connects knowledge to skill, making its development central to meaningful learning and future success.
Teach the literacy, not the tool
The objective, then, is not simply to teach students how to use a collection of AI tools. Mass-market AI systems are built around natural language processing, or NLP, and their interfaces are designed to be intuitive. Typing a prompt requires little technical training, but using AI well demands judgment. Students must learn to provide relevant context, define constraints, refine requests, and evaluate the results. These practices belong to the broader discipline of AI literacy, not to any single product.
The essential concepts remain consistent across platforms and products. That is the proper domain of AI literacy. A specific tool can demonstrate a principle, but no single tool should become the focus of AI literacy education. Products will change, interfaces will be redesigned, and market leaders will come and go. Students need knowledge and habits that remain useful through those changes.
Understanding uncertainty
Two foundational concepts students often misunderstand are non-determinism and probability, both of which are central to LLMs. Students must understand these ideas to use AI effectively. LLMs are trained on large collections of data and generate responses by predicting and ranking probable sequences of language. They do not retrieve a single, fixed answer from a perfectly ordered storehouse of facts.
Non-determinism means that the same prompt can produce slightly different results each time. Different platforms can also produce different answers because they may rely on different training data, models, system instructions, or retrieval methods. AI is not static, deterministic, or invariably correct. In that sense, AI literacy extends the long-standing challenge of using online information wisely. Teaching students to corroborate information has been a mainstay of digital literacy education for decades. With AI, verification is more essential than ever.
Teaching students to practice caveat emptor
Many widely used AI platforms are generative systems. They do not merely retrieve information from a source and display it. They generate new language and other content from patterns learned from existing data. That distinction matters because generated output can include what is commonly called an AI hallucination: information that sounds plausible but is inaccurate, unsupported, or invented.
AI-generated information is often accurate, but even occasional errors can create substantial risk. A system that is usually right can be especially convincing when it is wrong. Students should therefore follow the principle of caveat emptor: let the buyer beware. Confidence, fluency, polished language, and apparent supporting evidence do not guarantee accuracy. AI can cite sources incorrectly, misrepresent their content, or invent sources entirely. Students need not become intellectual cynics, but they must investigate claims, verify that cited sources exist, confirm that those sources support the claims attributed to them, and seek independent corroboration. Healthy skepticism is essential to discovery and should be embedded in learning activities that use AI-generated content.
Design learning around critique and growth
When AI is integrated into learning, students can explore content more deeply, but that exploration must remain guided by instruction. Educators must define acceptable use clearly and design assignments that preserve student responsibility. One increasingly common approach is to reverse the usual assignment: students generate content with AI and then evaluate it against an instructor-provided rubric. The student becomes the evaluator, turning critique into a practical exercise in critical thinking and learner agency.
The process itself becomes central. Iterations, revisions, and versions give instructors evidence to examine and verify. AI did not invent this educational principle. Many instructors have used process-based techniques for years. For every paper I assign, I require a draft, a critique from the college writing center, and a final version. Each assignment asks students to demonstrate academic growth in their writing, formulation of ideas, presentation, and critical thinking about the content and subject matter. AI makes that visible process more important, not less.
The carpenter’s rule for AI
Because AI generates content rather than simply retrieving it, it is best treated as a starting point for research. Like Wikipedia, it can provide a quick overview, explain the main aspects of a subject, and point toward sources for further investigation. Yet it should never be mistaken for a fully reliable authority or a complete body of knowledge. AI synthesizes patterns from data, and that synthesis can be incomplete, misleading, or wrong. The enduring principle of caveat emptor therefore matters more than ever. Students should follow the carpenter’s rule: measure twice but cut only once. Verification must come before acceptance and action. That principle has always been fundamental to digital literacy, and AI makes it more important than ever.
Discussion about this post