Is AI proficiency the same as being good at writing prompts?

No. Writing prompts is only one part of it. AI proficiency also includes understanding what AI can and cannot do, checking results for accuracy and bias, considering privacy, copyright, and safety, and ensuring that people retain final judgment and responsibility.

Can children use generative AI on their own?

Rather than deciding uniformly by age alone, consider the service's age requirements, a guardian's judgment, and the educational setting together. For young children, it is more appropriate for a parent, guardian, or teacher to review outputs with them for factual accuracy, upsetting content, and exposure of personal information than for them to use it independently. UNESCO has proposed setting a minimum age of 13 for the use of generative AI tools in schools.

How should I verify an AI response?

First, identify factual claims, dates, figures, and quotations in the response separately. Then compare them with primary materials or trustworthy official sources, and confirm important decisions with an expert or responsible staff member. After correcting errors, record which AI tool was used and where so that the verification process is documented.

What is the safest starting point when introducing AI at work for the first time?

Choose one repetitive, low-risk task. Examples include drafting from materials that can be made public, comparing summaries, organizing ideas, and converting table formats. At the same time, establish which tools are permitted, what information must not be entered, who is responsible for review, and how sources and copyright will be handled.

What should be covered first in AI education for older adults?

It is appropriate to practice in small steps, starting with tasks the person needs, such as understanding health information, accessing public services, or communicating with family. Key habits are not treating AI-generated health, financial, or administrative information as definitive; not entering sensitive information such as authentication codes or account details; and checking suspicious requests with family or an official institution.

How Can We Build AI Proficiency for Different Age Groups?

By jihyukboinPublished September 20, 2026

Building AI proficiency for different age groups is not about teaching everyone to use the same tools in the same way. It is about developing the ability to understand, ask questions, verify, and make safe decisions in ways that fit each person's foundational digital skills and goals in daily life, learning, and work. Young children may benefit most from noticing AI in everyday life and reviewing outputs with a guardian; teenagers from critical inquiry and responsible creation; adults from making practical tasks better in ways that can be reviewed; and older adults from needs-based accessibility and scam prevention. At every age, the principle is the same: people must not treat an AI answer as the final judgment; they must check the evidence and take responsibility. www.oecd.org www.unesco.org

What exactly is AI proficiency?

AI proficiency is a broader concept than being able to quickly press buttons in a particular chatbot or image-generation tool. Here, AI literacy means the knowledge, skills, and attitudes needed to understand how artificial intelligence is used, critically evaluate its outputs, and use it appropriately while considering its social impact and ethics. OECD and UNESCO discussions of education likewise address human-centered perspectives, critical evaluation, and ethical use, rather than tool proficiency alone. www.oecd.org www.unesco.org

In practice, it involves the following connected elements.

Element	Meaning	Everyday question
Understanding how it works	Knowing that AI produces results based on data and rules or learned patterns	Why did I get this recommendation?
Designing questions	Clearly communicating the purpose, conditions, audience, and format	Do I need a summary or a comparison?
Verifying outputs	Comparing facts, dates, figures, quotations, and logic against other evidence	Where can I verify this claim?
Safety and rights	Considering privacy, trade secrets, copyright, discrimination, and bias	Is it okay to enter this information?
Responsible judgment	Using AI suggestions as reference while people remain responsible for decisions and explanations	If it is wrong, who will explain the evidence and correct it?

Generative AI refers to AI that creates new outputs such as text, images, code, and summaries. However, a plausible sentence is not necessarily true. AI cannot be assumed to automatically verify sources when producing an answer or to responsibly carry out contextual and value judgments as a human would. In areas that can significantly affect individuals—such as health, hiring, assessment, finance, and law—AI outputs should not be used as the sole basis for decisions. They should be separately reviewed against official materials and by the responsible person or an expert. www.who.int

Why should learning levels not be determined by age alone?

Age can help identify a starting point for education, but it is not a standard for determining ability. People of the same age may differ in literacy, experience with smartphones and computers, internet access, available assistive devices, language, disability, and reasons for learning. Rather than asking, “What should someone this age learn?” it is better to first ask, “What can this person do independently now, and what support and safeguards do they need?” UNESCO views AI education competency levels and teaching methods as needing to be adapted to local conditions, including learners' existing competencies, teacher competencies, class time, infrastructure, and costs. unesdoc.unesco.org

For example, an adult unfamiliar with smartphones may first need to learn how to read a screen, check copied text, and distinguish official websites from advertisements, rather than how to automate complex tasks with AI. Conversely, a teenager who is comfortable with digital tools may still be inexperienced in evaluating sources and protecting personal information.

Educational goals also differ by person. For students, the inquiry process and expressing their own thinking matter; for employees, work quality and security matter; and for older adults, access to needed services, scam prevention, and ways to request help may be more urgent. Age groups are best viewed not as a fixed ranking of ability, but as one clue for asking about a person's learning context.

What principles are needed across all age groups?

First, distinguish AI answers from facts. Claims, dates, figures, legal or policy information, and quotations in an answer should be checked separately. When checking, do not rely only on short snippets in search results; compare primary materials, trustworthy sources such as public agencies, educational institutions, and academic institutions, and expert explanations when necessary.

Second, assess how sensitive information is before entering it. In addition to personally identifiable information such as names, contact details, and national identification numbers, people should not casually enter account credentials, banking information, patient information, student information, or a company's nonpublic documents and business information. In organizations, it is necessary to clearly define permitted tools and the categories of information that must not be entered, rather than relying only on individuals to be careful. AI data governance and accountability are treated as important risk factors in high-impact areas. www.who.int

Third, keep AI-generated outputs separate from your own judgment process. For an assignment, record where AI was used and what was revised. At work, distinguish the source material, AI draft, review comments, and final version. This makes it possible to correct errors when they arise and explain the outcome in your own words.

Fourth, consider potential discomfort or harm. Do not use outputs as they are if they portray a particular group in sweeping terms, evaluate an individual without evidence, or offer advice that could harm someone. “The AI said so” cannot serve as a basis for responsibility. UNESCO presents human-centered, ethical use as a core competency and addresses AI as something that should complement human judgment and learning. www.unesco.org www.unesco.org

What should preschoolers and younger elementary school students learn first?

At this stage, experiences that help children notice AI in everyday life are more appropriate than increasing the amount of independent tool use. You can talk together about features children already encounter, such as video recommendations, voice assistants, photo filters, and autocomplete: “Why might this feature show something like this?”, “Could information and rules made by people be involved?”, and “Is it always right?”

The key is not to regard AI as magic. A starting point is developing, through play and observation, a basic sense that AI is influenced by data and design created by people and can make mistakes. For example, when a feature that classifies animal photos makes an error, do not end with a competition to get the right answer. Discuss why it may have been confused based on the photo's lighting, angle, or background.

If generative AI needs to be used, a teacher or guardian-mediated approach is safer. An adult enters the question and reads the result with the child, checking whether it is factual, whether it contains frightening content, and whether personal information such as a name or photo has been included. UNESCO has noted that insufficiently validated generative AI tools may not be appropriate for primary education and has proposed restricting the age for using generative AI tools in schools to 13. This is an international organization's policy proposal; it should not be regarded as identical to the laws of each country or the age requirements of individual services. www.unesco.org www.unesco.org

At this stage, adults should repeat short checking questions rather than immediately provide the right answer. Questions such as “Who made this statement?”, “Should we ask another book or an adult too?”, and “Is it okay to write my name here?” can lead to later habits of verification.

How can older elementary students and teenagers learn “critical use”?

During adolescence, it is important to avoid two extremes: banning AI or using it as an answer machine. Students should try using AI while working on assignments and interests, but practice treating its output as a draft to review rather than the end point of a submission. This means making questions more specific, cross-checking with other materials, finding and correcting errors, stereotypes, and unclear sources, and explaining their own use process.

For instance, when researching a local environmental issue, students can ask AI for a draft research outline or interview questions. They can then verify the reference year and sources of statistics, compare them with materials from local public agencies or original reports, and supplement perspectives missing from the AI response. In the final work, they can briefly state which parts of the AI draft they consulted and what they revised. What matters in this process is not getting a longer answer, but evaluating evidence and developing one's own argument.

Images and videos also need to be read critically. Generated material may include scenes that never occurred, stereotypes about particular genders, occupations, or cultures, and sources that are difficult to verify. “Does it look real?” and “Is it true?” are different questions. Before sharing it with friends, students should find the original source or trustworthy reporting and consider whether they have improperly used someone else's face, voice, or work.

Students can learn coding, but coding alone is not the goal of AI education. Activities that investigate and discuss how data, recommendations, automated classification, and generated outputs affect different people are also important. UNESCO's AI Competency Framework for Students presents four areas—human-centered mindset, AI ethics, AI techniques and applications, and AI system design—across levels of understand, apply, and create. www.unesco.org

What should university students and early-career workers practice differently?

At this stage, what matters more than “how many tools can I use?” is whether AI helps make actual outputs in one's field or job more accurate and responsible. AI can assist with drafting questions for research, checking omissions in meeting-note summaries, structuring document drafts, organizing test items for code drafts, and explaining data analysis results. But creating a draft and taking final responsibility are different things.

For example, if meeting content is summarized with AI, it should be compared with the original recording or notes to ensure that decisions, responsible people, and deadlines have not changed. If you receive a code suggestion, you should review not only whether it runs, but also whether it meets requirements, handles errors, is secure, and passes tests. If analysis results are summarized, you should be able to explain the data's scope and limitations and the assumptions behind the interpretation.

It is difficult to assess good performance solely by the number of prompts or speed of output. The following questions can help gauge meaningful competence.

Did you define the purpose and constraints before assigning the problem to AI?
Did you verify the key claims and citations in the response?
Did you remove sensitive information from the input material?
If the result is wrong, can you find the evidence, correct it, and explain it?
Did you document the distinction between what AI did and what you decided?

Particular caution is needed in areas that can significantly affect people's opportunities and rights, such as hiring decisions, performance evaluations, medical advice, credit and financial decisions, and legal judgments. Because of issues involving AI bias, opacity, data management, and accountability, AI outputs should not be used as the sole criterion; review by responsible staff and experts, as well as official procedures, should remain separate. www.who.int

How should mid-career employees and career-transition learners get started?

Requiring people who are learning AI at work for the first time to absorb extensive theory or start with complex automation can increase the learning burden. A practical starting point is to choose one task from their own work that is repetitive but low risk. Examples may include drafting documents from materials that can be disclosed, comparing differences between summaries, organizing meeting agendas, converting table formats, and categorizing ideas.

A unit of learning can be structured as try → review → revise → standardize. First, generate a draft using materials that resemble real work but are not sensitive. Next, check for factual errors, missing conditions, excessive certainty, and awkward wording. After revision, document which input methods and review criteria were useful as team work standards. This turns individual tips into a reproducible process.

The organization's role also matters. Simply telling members to use AI makes it difficult to reduce the risks of disclosing personal information or trade secrets, copyright issues, and incorrect automation. At a minimum, organizations should state the permitted tools, information that must not be entered, who is responsible for reviewing outputs, how sources and copyright should be credited, and standards for retaining records. Teachers and education leaders need to design AI not as a device that replaces learners or teachers, but as a tool that complements learning design, critical thinking, and problem-solving. UNESCO's AI Competency Framework for Teachers likewise addresses human-centered perspectives, ethics, AI foundations and applications, AI pedagogy, and professional development together. www.unesco.org

What approach is suitable for older adults?

It is not appropriate either to treat older adults uniformly as non-users of AI or, conversely, to pressure them to learn the latest functions quickly. What matters is a pace of slow repetition centered on tasks with a clear purpose, along with the ability to receive help from a person when needed. Topics connected to everyday life—such as understanding health information, finding guidance on public services, communicating with family, and checking suspicious messages—can be a starting point.

Accessibility is as important as learning content. Large text, sufficient contrast, voice input, simple screens, repeated practice, printed guidance, and phone or in-person help channels can be combined to reduce barriers arising from differences in devices and physical ability. WHO discussions on age-friendly communication and information address approaches that consider diverse needs in access to information, and recommend investment in digital infrastructure and literacy for older people, participatory design, and protection of rights. extranet.who.int www.who.int

Minimum safety rules for AI education for older adults should be simple. Do not treat AI-generated health, financial, or administrative information as confirmed guidance; do not enter passwords, authentication codes, account information, or ID information; if urgent money transfers or disclosure of personal information are requested, check through another channel with family or an official institution; and do not make decisions alone when the result is difficult to understand.

AI may be able to assist with voice guidance or organizing information, but that possibility does not mean human support should be eliminated. Keeping offline routes and choices available so that people uncomfortable with digital methods can access the same services is inclusive design.

In what order should AI outputs be verified?

Verification is not something only specialists do; it is a habit whose depth should be adjusted according to importance. For low-impact tasks such as polishing a simple sentence, you can start by reading it yourself to check its naturalness and meaning. Content related to health, money, law, assessment, or safety, however, requires a stricter process.

1. Break the answer into small claims

Do not believe or discard a long response as a whole. Divide it into verifiable units. For example, mark statements about “when,” “who,” “how much,” and “what should be done.” In particular, treat numbers, dates, proper nouns, names of laws or policies, and quoted language as items to verify.

2. Find primary materials and independent evidence

If AI provides a source, check whether it actually exists and whether its content supports the claim. Do not repeatedly rely on that source alone; compare it with official documents or other trustworthy materials. If you cannot find the source, it is safer to treat the content as an “unverified draft.”

3. Check context and scope of application

Even if one fact is correct, it may not apply directly to your situation. Check which country, point in time, and audience it applies to; whether there are exceptions; and whether it is a general explanation or personalized advice. In high-impact decisions, the principle that a responsible person or expert must review the context is important. www.who.int

4. Explain the final version in your own words

If you cannot explain why you adopted a revised result, verification is probably not complete. Rather than copying an AI-generated sentence directly, rewrite it to fit your own purpose and evidence, and keep a record of the use and revision process.

What are some common misconceptions?

The first misconception is that “younger people are automatically good at using AI.” Experience using tools and the ability to evaluate them critically are different. Even experienced users can share plausible false information or enter sensitive content, while people who learn later can build sufficient competence through repeated learning tailored to their own goals.

The second is the belief that “one good prompt produces an accurate answer.” Making a question more specific can improve usefulness, but a prompt does not replace fact-checking. It is especially necessary to check original sources when AI provides quotations, figures, or examples.

The third is the belief that “using AI means I no longer have to think for myself.” In education and work, AI can provide drafts, alternatives, and questions, but identifying the important problem, deciding what evidence to adopt, and accepting the impact of the result remain human responsibilities. UNESCO presents human-centeredness and ethical considerations as core pillars of competencies for students and teachers. www.unesco.org www.unesco.org

The fourth is the belief that “individual caution alone is enough for safety.” Individual care is necessary, but schools and organizations must also establish procedures for tool selection, personal-data handling, review responsibility, accessibility, appeals, and support. Safe use becomes possible when individual capability and environmental design are both in place.

How can I decide on the next step that suits me?

At the beginning, rather than following age-specific recommendations exactly, check the following four points. First, identify the task you are trying to solve now. Second, assess whether the consequences of an error in that task would be small or large. Third, confirm whether the materials may be entered into AI. Fourth, confirm whether there is a person and evidence to review the result and take responsibility for it.

For low-impact tasks involving information that can be made public, you can start with a small experiment. For example, summarize the same content in two ways, compare the differences, and correct what is missing yourself. For high-impact tasks, limit AI to a reference draft or question-generation tool, and make official materials and human review mandatory steps. Add adult supervision and safety rules for young users, and accessible screens and help routes for older users.

Ultimately, the purpose of AI proficiency education for different age groups is not to make everyone the same kind of skilled technology user. It is to help each person understand both the benefits and limitations of AI in their own circumstances, verify needed information, protect other people's rights and their own information, and make final decisions responsibly. Tools will keep changing, but the habits of asking questions, verifying, and revising provide a lasting foundation.