AI & Tech Intelligence · Education
AI in Education: How It's Used, What Works, and the Risks
AI in education by the numbers: 4 in 5 students now use AI for schoolwork (Stanford HAI 2026), 60% of teachers use it, and fewer than 10% of schools have policies. Sourced.
TL;DR — Roughly four out of five US high-school and college students now use AI for schoolwork, yet over 80% say their teachers never taught them how to use it (Stanford HAI AI Index, 2026).
Stanford HAI · Gallup / Walton Family Foundation · UNESCO · Stanford | 3,157 words · 17 sections | data: CSV + JSON
Executive summary
AI in education has moved from experiment to default faster than schools can write rules for it. The Stanford HAI AI Index 2026 finds that roughly four out of five US high-school and college students now use generative AI for schoolwork — most often for research, editing and brainstorming — while more than 80% report their teachers never taught them how to use it, and only half of middle and high schools have any AI policy at all (Stanford HAI, 2026). Teachers are adopting too: a 2025 Gallup survey of 2,232 US public-school teachers found 60% used AI during 2024-25, with weekly users saving an estimated 5.9 hours a week, about six weeks over a school year (Gallup, 2025). Purpose-built tools have scaled quickly — Khan Academy's Khanmigo tutor grew from about 40,000 to 700,000 K-12 students in a single year, and Duolingo's GPT-4-powered Max drove daily active users up 51% to over 40 million (Khan Academy; Duolingo, 2025). But governance lags badly: UNESCO found fewer than 10% of institutions have formal generative-AI guidance (UNESCO, 2023). Three tensions define the field. Academic integrity: AI detectors are unreliable and biased against non-native and neurodivergent writers (Stanford, 2023). Equity: only 40% of primary schools worldwide even have internet (UNESCO GEM). And policy: the EU AI Act classifies education AI as high-risk, with core obligations from August 2026. The evidence supports cautious, teacher-led adoption — not prohibition, and not hype.
“It is my hope that this Guidance will help us redefine new horizons for education and inform our collective thinking and collaborative actions that can lead to human-centred digital learning futures for all.”
Key findings
Four in five students use AI — but few were taught how
The Stanford HAI AI Index 2026 education chapter reports that about four out of five US high-school and college students now use generative AI for school-related tasks — most commonly research, essay editing and brainstorming. The same data show a preparedness gap: more than 80% of students say their teachers never explicitly taught them how to use AI, only half of middle and high schools have an AI policy, just 6% of teachers describe those policies as clear, and only 34% say their school has any policy on AI and academic integrity. Adoption is running years ahead of instruction and governance.
Source: Stanford HAI AI Index 2026 · 2026 · confidence: High
Teachers save about six weeks a year with AI
A Gallup survey conducted for the Walton Family Foundation from 18 March to 11 April 2025, covering 2,232 US public K-12 teachers, found that 60% used AI tools for work during the 2024-25 school year and about 30% used them at least weekly. Weekly users reported saving an estimated 5.9 hours a week — roughly six weeks across a 37.4-week school year — most often on lesson preparation, which was the single most common weekly use. Between 57% and 74% of AI-using teachers said the tools improved the quality of their work, with 16% or fewer reporting any decline.
Source: Gallup / Walton Family Foundation · 2025 · confidence: High
Governance is missing almost everywhere
A UNESCO global survey of more than 450 schools and universities found that fewer than 10% had developed institutional policies or formal guidance on the use of generative AI applications. The same work noted that in high-income countries more than two-thirds of secondary pupils were already using generative AI to produce schoolwork, and that of the hundreds of institutions surveyed, only two reported a policy amounting to an outright ban. UNESCO's first global guidance on the technology urges age limits for independent AI use and mandatory data-privacy protection.
Source: UNESCO · 2023 · confidence: High
2026: adoption is universal, readiness is not
The 2026 evidence points in one direction — students and teachers have adopted AI en masse, while schools, regulators and detection tools are still catching up. The Stanford HAI AI Index 2026 documents near-ubiquitous student use alongside a governance vacuum, and reports that over 90% of countries now offer computer science to primary or secondary students, with China and the United Arab Emirates mandating AI education from the 2025-26 school year. On the regulatory side, the EU AI Act's AI-literacy duty took effect in February 2025 and its full high-risk obligations for education systems apply from August 2026. Meanwhile the reliability of AI-detection tools continues to be challenged in peer-reviewed work, pushing institutions from Vanderbilt to Yale to switch detection off rather than risk false accusations. The through-line: the near-term policy fight is not whether students use AI — they already do — but how to teach with it fairly.
What is AI in education, and why now?
AI in education is the use of artificial-intelligence systems — increasingly generative AI such as large language models — to support teaching, learning, assessment and school administration. In practice that spans adaptive tutoring systems, chatbots that explain concepts, tools that draft lesson plans and rubrics, automated feedback on writing, and analytics that flag struggling students. The defining shift since late 2022 is generative AI: tools like ChatGPT, Google Gemini and Anthropic's Claude that produce human-like text and can tutor, draft or grade on demand.
The 'why now' is speed of adoption. The Stanford HAI AI Index 2026 finds roughly four out of five US high-school and college students already use AI for schoolwork, and a 2025 Gallup survey found 60% of US K-12 teachers used it in a single year (Stanford HAI, 2026; Gallup, 2025). UNESCO reports that in high-income countries more than two-thirds of secondary pupils were using generative AI for schoolwork within a year of ChatGPT's launch (UNESCO, 2023). No previous education technology — not the calculator, not the internet, not the interactive whiteboard — reached classroom ubiquity this fast.
That speed is exactly why the topic is contested. Adoption has outrun the evidence base, teacher training and regulation simultaneously, producing a field where genuine learning gains, real cheating risks and deep equity concerns all coexist. The rest of this report separates what the data supports from what it does not.
How is AI used for personalized learning and tutoring?
AI's most-cited educational promise is personalized tutoring at scale — a one-to-one 'tutor for every student' that adapts to each learner's pace. The flagship example is Khan Academy's Khanmigo, a GPT-4-based tutor and teaching assistant that, according to Khan Academy, grew from roughly 40,000 to 700,000 K-12 students in the 2024-25 school year, with the organisation projecting more than one million in 2025-26 (Khan Academy, 2025). Khanmigo is designed as a Socratic guide that refuses to simply give answers, and Khan Academy made its teacher tools free to all US teachers with Microsoft's support.
Language learning is the other proven arena. Duolingo's premium tier, Duolingo Max, launched in March 2023 powered by OpenAI's GPT-4 and adds AI 'Roleplay' conversation practice, 'Explain My Answer' and a 'Video Call' feature with an AI character. Duolingo reported daily active users up 51% to over 40 million and monthly active users around 130 million in 2025, crediting AI-driven course creation for launching scores of new courses at once (Duolingo, 2025). Adaptive systems tailor difficulty in real time, targeting an optimal challenge level for each learner.
The honest caveat: rigorous, independent evidence that AI tutors durably raise learning outcomes at scale is still thin and mostly vendor-reported. The mechanism — immediate, patient, individualized feedback — is well grounded in learning science, but the field is young, and the strongest claims come from the companies selling the tools. Personalization is AI's clearest opportunity in education and its least independently proven.
How do teachers use AI for planning, grading and admin?
Teachers use AI mainly to reclaim time on the work that surrounds teaching — lesson planning, materials, feedback and administration. In the 2025 Gallup survey of 2,232 US public-school teachers, 60% used AI during 2024-25 and preparing to teach (lesson planning) was the single most common weekly use, cited by about 20% (Gallup, 2025). Weekly users saved an estimated 5.9 hours a week — roughly six weeks per school year — and 57% to 74% reported that AI improved the quality of their work.
Grading and feedback are a more cautious frontier. AI can draft feedback, generate rubrics and give first-pass scores, but educators and regulators consistently warn against handing final grading decisions to a machine. Student opinion tracks this line: Gallup found 51% of Gen Z K-12 students think teachers should be allowed to use AI for lesson planning, but only 39% support its use for grading (Gallup, 2025). The emerging norm is AI as a drafting assistant with a human making the judgement.
The binding constraint is training, not willingness. The Stanford HAI AI Index 2026 found that more than 80% of students were never taught how to use AI, only 6% of teachers call their school's AI policy clear, and only 34% report any academic-integrity policy on AI (Stanford HAI, 2026). Teachers are adopting AI faster than their institutions can support them — the productivity gains are real, but they are happening in a governance vacuum.
How are students actually using generative AI?
Student use of generative AI is now the norm rather than the exception, and it is overwhelmingly for legitimate study tasks — with a cheating minority that gets most of the attention. The Stanford HAI AI Index 2026 reports that about four out of five US high-school and college students use AI for schoolwork, most commonly for research, essay editing and brainstorming (Stanford HAI, 2026). UNESCO similarly found that more than two-thirds of secondary pupils in high-income countries were using generative AI for schoolwork within a year of ChatGPT's release (UNESCO, 2023).
The reality is messier than 'students cheat with ChatGPT'. Much use is a study aid — asking for explanations, summarising readings, generating practice questions or checking one's own work — the kind of support a private tutor would give. The problem is that the same tool that explains a concept can also write the essay, and the line between assistance and outsourcing is drawn differently by every teacher and institution, often without a written policy to point to.
This is the core governance failure the data exposes: near-universal student adoption meets near-total absence of clear rules. When over 80% of students use AI and fewer than 10% of institutions have formal guidance, students are left to invent their own norms — which is neither fair to them nor good for learning integrity (Stanford HAI, 2026; UNESCO, 2023).
What do the AI-in-education adoption numbers say?
Adoption statistics from 2023-2026 tell a consistent story: usage is high and rising among both students and teachers, while institutional policy trails far behind. The table below collects the most-cited figures with their sources so each can be checked and quoted independently.
Read together, these numbers frame the central mismatch of the field — demand-side adoption near saturation, supply-side governance in single digits. The strongest, most independent figures come from Stanford HAI, Gallup and UNESCO; tool-specific user counts are self-reported by vendors and should be read as directional.
| Metric | Figure | Source (year) |
|---|---|---|
| US students using AI for schoolwork | ~4 in 5 | Stanford HAI (2026) |
| Students never taught to use AI | 80%+ | Stanford HAI (2026) |
| US K-12 teachers using AI in 2024-25 | 60% | Gallup (2025) |
| Teachers using AI at least weekly | ~30% | Gallup (2025) |
| Weekly time saved per teacher | 5.9 hrs/week | Gallup (2025) |
| Secondary pupils using GenAI (high-income countries) | >2/3 | UNESCO (2023) |
| Schools/universities with formal GenAI guidance | <10% | UNESCO (2023) |
| Khanmigo K-12 students (growth in one year) | ~40k → 700k | Khan Academy (2025) |
| Duolingo daily active users | +51% to 40M+ | Duolingo (2025) |
What can AI actually do in a classroom? (use cases)
AI's classroom use cases fall into a handful of well-defined categories, each with a real tool already deployed at scale. The table maps each use case to what it does and a named, real-world example, so readers can see the difference between marketing categories and shipping products.
Two patterns hold across the table. First, the strongest use cases augment a human rather than replace one — a tutor that guides, a planner that drafts, a feedback tool that suggests. Second, the most consequential uses (grading, admissions, proctoring) are exactly the ones regulators treat as high-risk, which is where the policy section below focuses.
| Use case | What it does | Real example |
|---|---|---|
| Personalized tutoring | One-to-one adaptive help, Socratic prompts, hints | Khanmigo (Khan Academy) |
| Language practice | AI conversation, instant feedback, roleplay | Duolingo Max |
| Lesson planning | Drafts plans, rubrics, differentiated materials | ChatGPT, Google Gemini, MagicSchool |
| Writing feedback | First-pass comments and revision suggestions | ChatGPT, Grammarly |
| Administrative support | Emails, reports, parent communications | Microsoft Copilot, Google Gemini |
| Accessibility & special needs | Read-aloud, simplification, executive-function support | Assistive AI tools (e.g. SPSM-guided use, Sweden) |
| AI literacy instruction | Teaching how AI works and its limits | National CS/AI curricula (China, UAE, Sweden) |
Does AI cause cheating, and can detectors catch it?
AI has made outsourcing schoolwork trivially easy, but the tools meant to catch it are unreliable enough that leading universities are switching them off. A widely cited Stanford study (Liang et al., 2023) tested seven GPT detectors and found they wrongly flagged 61% — 61.3% — of essays by non-native English speakers as AI-generated, while classifying native-speaker writing near-perfectly. Independent 2024-2025 testing puts false-positive rates for non-native, heavily edited or technical writing in the range of 5-12%, high enough to wrongly accuse real students.
The bias is not random. Peer-reviewed and conference work has found that non-native English learners and neurodivergent writers are disproportionately flagged, because their writing patterns can superficially resemble machine text. Because a false accusation of cheating can end a student's academic standing, several institutions — including Vanderbilt, Northwestern and others — have disabled AI detection in Turnitin rather than rely on it. The honest position is that current detectors cannot be trusted as sole evidence of misconduct.
The constructive response most experts favour is assessment redesign over surveillance: in-class writing, oral defences, process portfolios, and assignments that ask students to use and critique AI transparently. The cheating risk is real, but the detection arms race is one schools are losing — the durable fix is changing what and how we assess, not buying a better detector.
Does AI widen the digital divide in education?
AI risks widening educational inequality because the divide starts with electricity and connectivity, long before any question of AI access. UNESCO's Global Education Monitoring data show that in 2023 only about 40% of primary schools, 50% of lower-secondary and 65% of upper-secondary schools worldwide were connected to the internet, and roughly one in four primary schools lacked electricity entirely (UNESCO GEM). A tutor-for-every-student is meaningless where there is no reliable power or bandwidth.
Even where infrastructure exists, access to capable AI is uneven. Premium tools, faster models and paid tiers concentrate among wealthier students and better-funded schools, raising the prospect of a two-tier system in which advantaged learners get sophisticated AI tutoring while others get nothing or low-quality substitutes. The OECD stresses that jurisdictions must ensure equitable devices, connectivity and professional learning so that all students and teachers can benefit from generative AI, rather than only those already ahead (OECD).
There is a counter-current worth weighting fairly. Free tools like Khanmigo-for-teachers and the base tiers of major chatbots can, in principle, democratise access to tutoring that was once the preserve of families who could afford private tutors. Research from Sweden also finds generative AI can be especially valuable for students with executive-function challenges. Whether AI narrows or widens gaps is not determined by the technology — it depends on public investment in infrastructure, free access and teacher training.
What are the risks, and what do UNESCO and the EU require?
The main risks of AI in education — cheating, bias, privacy loss, over-reliance and inequity — each have a recognised mitigation, and regulators are beginning to require them. The EU AI Act classifies many education AI systems as high-risk under Annex III: tools that decide admissions, evaluate learning outcomes or grade, and systems that monitor or proctor exams. Its AI-literacy obligation applied from February 2025, and full high-risk obligations for these systems apply from August 2026 (EU AI Act, Annex III).
UNESCO's global guidance is the reference framework on the pedagogical side: it calls for age limits on independent AI use, mandatory data-privacy protection, teacher training, and above all a human-centred approach in which AI never replaces a teacher's judgement. The OECD complements this with an AI-literacy framework for primary and secondary education and is building AI literacy into PISA — the 2025 cycle adds a 'Learning in the Digital World' assessment and PISA 2029 will measure media and AI literacy directly (OECD; UNESCO).
The table pairs each major risk with the mitigation the evidence and guidance support. The consistent principle across UNESCO, the OECD and the EU is human oversight: AI as an assistant to a professional educator, subject to transparency, privacy protection and a person accountable for consequential decisions.
| Risk | Why it matters | Mitigation |
|---|---|---|
| Academic dishonesty | Easy outsourcing of graded work | Assessment redesign; transparent AI use; oral/in-class work |
| Unreliable detection | False accusations, biased against non-native writers | Do not use detectors as sole evidence; human review |
| Data privacy | Student data fed to third-party models | Data-protection rules; keep student texts out (EU AI Act; Skolverket) |
| Bias & inequity | Uneven access; biased model outputs | Free access, connectivity investment, equity audits (OECD) |
| Over-reliance | Erosion of independent thinking | AI literacy; teach critique, not just use (UNESCO, OECD) |
| High-stakes automation | Grading/admissions decided by AI | High-risk classification; human accountability (EU AI Act) |
How are Europe, the Nordics and Sweden handling AI in schools?
Europe is regulating AI in education harder and earlier than any other region, led by the EU AI Act's high-risk classification of school AI systems and its phased obligations through August 2026 (EU AI Act, Annex III). This makes the EU the global reference point for how far formal regulation of classroom AI can go, in contrast to the largely policy-free environment UNESCO documented worldwide.
Sweden illustrates the balancing act. The national agency Skolverket has issued recommendations on generative AI and cheating, and its conditions are strict: any use must comply with the EU AI Act, student personal data and texts must be kept out of third-party tools, and no AI may replace a teacher's professional judgement or a grading decision. Sweden also introduced a dedicated AI school subject — first for STEM programmes and then expanding — framing AI as a sociotechnical system students should understand critically, not just operate.
The Nordic model leans on institutions rather than prohibition. Sweden's special-needs agency SPSM published guidance in November 2025 on AI from a special-education perspective, seeing scope to individualise support for neurodivergent students while warning against a dependence-creating 'care culture'. Universities such as Gothenburg have embraced generative AI with structured policies, and national efforts like WASP-ED and AI Sweden aim to scale AI education. The Nordic bet is that strong public institutions, teacher trust and equity funding can turn AI into a leveller rather than a divider — a bet that depends on execution, not technology.
What's the outlook for AI in education?
The near-term outlook is convergence on 'teach with it, don't ban it' — because banning has already failed in practice. With four in five students using AI and only two of 450+ institutions worldwide opting for outright bans, the realistic policy question is how to integrate AI fairly, not whether to allow it (Stanford HAI, 2026; UNESCO, 2023). Expect rapid growth in AI-literacy curricula: over 90% of countries now offer computer science to school students, and China and the UAE have mandated AI education from 2025-26 (Stanford HAI, 2026).
Assessment will change more than instruction. As detection stays unreliable, schools will shift toward AI-resistant and AI-inclusive assessment — process portfolios, oral defences, in-class work and assignments that require transparent AI use. Personalized tutoring will keep scaling commercially, but the field badly needs independent, peer-reviewed evidence of durable learning gains before the strongest claims can be trusted.
The two variables that will decide outcomes are equity and evidence. If public investment closes the connectivity and access gaps, AI could democratise high-quality tutoring; if it does not, AI will widen existing divides. And if independent research confirms real learning gains, adoption accelerates on solid ground; if it does not, today's enthusiasm will look like the last decade's overhyped ed-tech cycles. The technology is not destiny — policy, funding and rigorous evaluation are.
Scoreboard (machine-readable data)
Every headline indicator with its value, period, source and confidence. Free to reuse under CC BY 4.0.
| Indicator | Value | Period | Source | Conf. |
|---|---|---|---|---|
| US students using AI for schoolwork | 80 pct | 2026 | Stanford HAI AI Index 2026 | High |
| US K-12 teachers who used AI in 2024-25 | 60 pct | 2025 | Gallup / Walton Family Foundation | High |
| Weekly time saved by AI-using teachers | 5.9 hours/week | 2025 | Gallup / Walton Family Foundation | High |
| Schools/universities with formal GenAI guidance | 10 pct (fewer than) | 2023 | UNESCO | High |
| Non-native essays wrongly flagged as AI | 61 pct | 2023 | Stanford (Liang et al.) | High |
| Primary schools connected to internet | 40 pct | 2023 | UNESCO GEM | Medium-High |
Methodology & verification
This report synthesises primary and peer-reviewed sources on AI in education and attributes every statistic inline to its named source, year and scope. Headline adoption figures come from the Stanford HAI AI Index 2026 (education chapter), the 2025 Gallup / Walton Family Foundation teacher survey (n=2,232 US public K-12 teachers, fielded 18 March-11 April 2025), and UNESCO's global survey of 450+ institutions and its Guidance for Generative AI in Education and Research. Detection-reliability figures draw on Stanford's 2023 study of GPT detectors (Liang et al.) and independent 2024-2025 testing. Connectivity and equity figures come from UNESCO Global Education Monitoring data and the OECD. Regulatory statements reflect the EU AI Act (Annex III) and national guidance from Sweden's Skolverket. Vendor-reported user counts (Khanmigo, Duolingo Max) are labelled as such and treated as directional rather than independently audited. Where sources measure different things or disagree, ranges are reported rather than a single point estimate. Figures were checked against source publications as of 30 July 2026.
Data dictionary
| Field | Type | Description |
|---|---|---|
| students_using_ai_pct | percentage | Share of US high-school and college students who use generative AI for school-related tasks, per Stanford HAI AI Index 2026 (reported as roughly four in five / ~80%). |
| teachers_using_ai_pct | percentage | Share of US public K-12 teachers who used AI tools for work during the 2024-25 school year, per the 2025 Gallup / Walton Family Foundation survey. |
| schools_with_genai_policy_pct | percentage | Share of surveyed schools and universities with formal institutional guidance or policy on generative AI use, per UNESCO's global survey of 450+ institutions (reported as fewer than 10%). |
Frequently asked questions
How is AI used in schools?
AI in schools is used mainly for personalized tutoring, lesson planning, writing feedback, administrative tasks and accessibility support. In 2024-25, 60% of US K-12 teachers used AI — most often for lesson planning — and weekly users saved about 5.9 hours a week (Gallup, 2025). Roughly four in five students use AI for research, editing and brainstorming (Stanford HAI, 2026).
What is AI tutoring and does it work?
AI tutoring uses generative AI to give one-to-one adaptive help, hints and explanations at scale — Khan Academy's Khanmigo grew from about 40,000 to 700,000 K-12 students in a year (Khan Academy, 2025). The mechanism (instant, patient, individualized feedback) is grounded in learning science, but independent peer-reviewed evidence of durable learning gains at scale is still limited, and most strong claims are vendor-reported.
How many students use ChatGPT and AI for schoolwork?
About four out of five US high-school and college students use AI for schoolwork, according to the Stanford HAI AI Index 2026, most commonly for research, essay editing and brainstorming. In high-income countries, UNESCO found more than two-thirds of secondary pupils were using generative AI for schoolwork within a year of ChatGPT's launch (UNESCO, 2023).
Can AI detectors reliably catch AI-written essays?
No. A 2023 Stanford study (Liang et al.) found seven AI detectors wrongly flagged 61% of essays by non-native English speakers as AI-generated, and independent 2024-25 testing puts false-positive rates for non-native or edited writing at 5-12%. Because of this bias, universities including Vanderbilt and Northwestern have disabled Turnitin's AI detection.
Is AI in education good or bad for equity?
It depends on investment. The divide starts with infrastructure: only 40% of primary schools worldwide have internet and one in four lacks electricity (UNESCO GEM). Free tools can democratise tutoring, but premium AI concentrates among wealthier students. The OECD stresses that equitable devices, connectivity and teacher training determine whether AI narrows or widens gaps.
How do teachers use AI, and is it allowed?
Teachers most commonly use AI for lesson planning, materials, feedback and admin; 60% of US K-12 teachers used it in 2024-25 (Gallup, 2025). Grading is more restricted — regulators and agencies like Sweden's Skolverket require that AI never replaces a teacher's grading judgement, and the EU AI Act treats automated grading as high-risk.
What does the EU AI Act require for AI in education?
The EU AI Act classifies many education AI systems — admissions, grading, exam proctoring — as high-risk under Annex III. An AI-literacy obligation applied from February 2025, and full high-risk obligations for education systems apply from August 2026, requiring transparency, human oversight and accountability.
What are the biggest risks of AI in education?
The main risks are academic dishonesty, unreliable and biased detection, student-data privacy, inequity of access, and over-reliance eroding independent thinking. UNESCO, the OECD and the EU AI Act converge on the same mitigation: human oversight, AI literacy, privacy protection, and never letting AI make consequential decisions like final grades alone.
Glossary
- Generative AI
- AI systems such as large language models (ChatGPT, Gemini, Claude) that produce human-like text, images or code, and can tutor, draft or assess on demand — the technology driving the post-2022 surge in education. ↗
- Adaptive / personalized learning
- Software that adjusts content, pace and difficulty to each learner in real time, aiming to keep every student at an optimal challenge level — the core promise of AI tutoring tools like Khanmigo and Duolingo Max. ↗
- AI detector
- A tool that estimates whether text was written by AI. Detectors are error-prone and biased against non-native and neurodivergent writers, and are widely considered unreliable as sole evidence of misconduct. ↗
- AI literacy
- The ability to understand how AI works, use it critically and responsibly, and recognise its limits. Required of school staff under the EU AI Act from February 2025 and built into OECD/PISA assessments. ↗
Embed & cite this report
Free to reuse under CC BY 4.0. Embed the live-updating widget on your site, or cite the report directly — always with attribution to Affärslivet.
Embed (HTML) — auto-updating
<iframe src="https://xn--affrslivet-s5a.com/en/embed/ai-in-education" width="100%" height="520" style="border:1px solid #e3e3e6" title="AI in Education: How It's Used, What Works, and the Risks — Affärslivet" loading="lazy"></iframe> <p style="font:12px sans-serif">Source: <a href="https://xn--affrslivet-s5a.com/en/reports/ai-in-education">Affärslivet</a></p>
APA
Affärslivet Research. (2026). AI in Education: How It's Used, What Works, and the Risks. Affärslivet. Version 1.0. https://xn--affrslivet-s5a.com/en/reports/ai-in-education
MLA
Affärslivet Research. "AI in Education: How It's Used, What Works, and the Risks." Affärslivet, 2026-07-30, https://xn--affrslivet-s5a.com/en/reports/ai-in-education.
BibTeX
@techreport{affarslivet_ai_in_education,
title = {AI in Education: How It's Used, What Works, and the Risks},
author = {{Affärslivet Research}},
year = {2026},
note = {Version 1.0},
url = {https://xn--affrslivet-s5a.com/en/reports/ai-in-education}
} License CC BY 4.0 — free to cite, embed and republish with attribution to Affärslivet. Data also as CSV / JSON.
Sources
- Stanford HAI — 2026 AI Index Report, Education chapter
- Gallup / Walton Family Foundation — Teaching for Tomorrow: Unlocking Six Weeks a Year With AI (2025)
- UNESCO — Guidance for Generative AI in Education and Research; survey of 450+ institutions (2023)
- Stanford (Liang et al.) — GPT Detectors Are Biased Against Non-Native English Writers (2023)
- EU Artificial Intelligence Act — Annex III (High-Risk AI Systems, incl. education)
- OECD — Artificial Intelligence and Education and Skills; Digital Education
Part of Affärslivet AI Intelligence
This report is one part of Affärslivet's source-cited AI knowledge layer. Start with the big picture:
- The State of AI 2026 — our free, source-cited annual report
- AI Statistics 2026 — 230+ source-cited AI facts & figures
- The AI Intelligence hub — all our indexes, rankings, model & lab profiles