The Death of English: How AI Can Dismantle Education’s Gatekeeping Tongue
For decades, English has functioned as the global education gatekeeper. Students interested in studying in the U.S., U.K., Australia, and similar Anglophone destinations had to clear a single hurdle above all else: English proficiency. This meant taking one of now many standardized tests (TOEFL, IELTS, PTE-A, Duolingo English Test, etc.), each packaged and sold by corporations sitting in the center of the student mobility ecosystem.
Consider Maria, a brilliant engineering student from Colombia, fluent in Spanish and Portuguese, with strong technical skills demonstrated through her portfolio of projects. She's denied admission to her dream program because her TOEFL score falls three points short. Or Ahmed from Egypt, who speaks Arabic, French, and has functional English, but can't afford the $200 test fee and prep courses. They represent millions of students globally.
The Gatekeeping System
IDP Education, which co-manages IELTS, hauled in $410.7 million AU in English Language Testing revenues for FY2025, though that's a 15% drop year-on-year, with testing volumes down 18%. Despite the dip, language-testing remains a high-margin engine in their business.
Duolingo’s foray into this space is a strategic business expansion and one that helped fuel its multibillion-dollar IPO. For these entities, English dominance is not just a matter of language but a revenue stream.
Governments reinforce the system. The UK Home Office, U.S. Department of State, and Australian Department of Home Affairs all require proof of English for student visas. Universities, even those with robust international student support systems, may outsource admissions decisions to minimum test scores rather than holistic evaluations, especially those without ESL programs to support students with lower English proficiency. English test scores serve a similar function as the SAT or ACT – a biased test that is easy to use as a line in the sand for admissions. Rankings agencies, such as QS and Times Higher Education, elevate English-medium universities to the top tiers of their lists, reinforcing the illusion that English equals quality.
The result is a cycle: universities want international students for tuition revenue, governments want them vetted, and testing companies sit in the middle collecting fees. The student pays in money, time, and lost opportunities.
The dominance of English is not neutral. It is a direct inheritance of colonial rule. For students from South Asia, Africa, or the Caribbean, English proficiency requirements often feel like a continuation of historic subjugation. Why should a Nigerian or Indian student, already fluent in two or three languages, be denied access to higher education because their TOEFL score falls a few points short?
This argument resonates especially as global education attempts to decolonize itself. Diversity and inclusion efforts ring hollow when students are still being filtered by colonial-era language hierarchies. Students from the Global South, where English education quality varies dramatically by economic class, face a double burden: overcoming both historical disadvantage and contemporary financial barriers.
Enter AI: The Great Equalizer
The problem for the English-language empire is that technology, geopolitics, and student agency are colliding to challenge its supremacy.
AI translation technology has advanced dramatically in recent years. The latest AI models are approaching, and in some cases matching, traditional machine translation services. Studies comparing GPT-4, Google Translate, and DeepL across multiple language pairs demonstrate competitive performance, with AI systems showing particular strength in handling context and nuance. While this advancement drastically lowers the barrier for comprehension and class participation, critics correctly point out that high-stakes original academic writing, especially at the graduate level, still requires a degree of linguistic precision that AI translation tools have yet to consistently master. This tension is where the current system digs in its heels. This linguistic precision is compounded by the need to master disciplinary conventions. Academic writing demands specific rhetorical practices (hedging, citation voice, structural flow) unique to each field, and simply translating a student's thoughts may fail to train them in the professional discourse required for their field's publishing ecosystem.
For students, imagine attending a Mandarin lecture and listening through earbuds in Spanish or Arabic. Or typing in French, submitting in English, with the translation handled by AI. Several U.S. school districts have provided AI translation devices to multilingual and refugee students, with teachers reporting improved lesson comprehension and more active class participation.
While most universities have yet to implement such systems at scale, the technology is ready. The question is not whether this is technically feasible but whether institutions have the will to adopt it broadly.
Platforms like Preply harness AI to adapt learning experiences, boosting learner outcomes through a hybrid of algorithmic personalization and real tutor guidance. One out of three students advance a full CEFR level in 12 weeks. Google Translate now offers live "practice" modes with personalized conversational exercises across languages. This shifts language acquisition from a barrier to an embedded, AI-supported experience.
And I don’t know about you, but I’ve had my eye on any number of AI-powered glasses for real-time translation when traveling. My own excitement is about the sheer joy of instant, unmediated connection with someone whose mind I’m eager to know, but whose language I don’t yet speak. Why wouldn't we want that same level of open access for education?
Devil’s Advocate
Some arguments we might hear:
“English is the global lingua franca.”
True, English is still the most widely studied second language in the world. Scientific publishing, international business, and diplomatic forums are overwhelmingly conducted in English. For now, English retains enormous practical utility.
Yes, but technology increasingly erodes this necessity. Multilingual AI systems are already better than most humans at instant translation. The reliance on English as a "neutral" meeting point will diminish as digital tools handle the mediation. Moreover, Mandarin, Spanish, and Arabic are rising as academic and economic powerhouses. China's Belt and Road Initiative is tied to expanding Mandarin-language education abroad. Latin America's demographic and economic growth make Spanish a natural contender for regional dominance. Within a generation, English may be only one of several "global academic languages," not the unquestioned standard.
“Students need English to succeed academically.”
Students without strong English skills may struggle in Anglophone universities, both in class and socially. Requiring English proficiency upfront is meant to protect both the student and the institution. This is a function of the system, not the students, though. Universities deliver everything in English because that is how the system has been designed. If institutions invested in multilingual teaching models or AI-enabled learning supports, the barrier could be lowered without sacrificing academic quality. The European Higher Education Area already demonstrates this as many universities successfully operate in multiple languages with strong academic outcomes.
“Governments need English benchmarks to process visas fairly.”
This is the most substantial concern. How do we assess student readiness without standardized tests? How do we ensure academic integrity when AI mediates communication? How do we verify that AI translation preserves academic nuance and disciplinary conventions?
The answer isn't to abandon all assessment but to evolve it. Universities could develop AI-mediated portfolio assessments where students demonstrate competence in their field using their strongest language, with AI providing translation for evaluators. Institutions could implement probationary periods where students' actual performance, rather than test scores, determines continuation. Academic integrity concerns exist with or without AI. We already navigate plagiarism detection and contract cheating – sometimes it feels like that’s all we are navigating. The solution is adapting our pedagogy and assessment methods, not clinging to outdated gatekeeping mechanisms.
The United States, famously, does not have an official language. Why does the U.S. Department of State insist that Fs and Js demonstrate English proficiency when a student can speak 5 languages already and can easily use AI tools? The requirement is likely less about maintaining power structures and more about risk mitigation for the visa category. The U.S. doesn't have an official language, but the majority of day-to-day life, especially post-secondary education, is conducted in English. The DoS wants to ensure the student can survive socially, legally (e.g., reading a lease), and academically without total dependency on an AI device for basic functioning. While the intent may be practical survival, the effect is still to reinforce a colonial-era hierarchy. The question is, where is the line for “survival” drawn? At being able to order coffee, or at being able to write a doctoral dissertation in flawless English? While AI is ready for the classroom, we must be honest that it still has vulnerabilities in unstructured social and emergency settings (e.g., poor connectivity, device failure). The goal isn't zero English proficiency, but shifting the threshold from the current, high bar for academic production to a lower, functional threshold for societal navigation, allowing AI to handle the cognitive load of advanced study.
Of course, the defenders of English dominance have their points, and they deserve to be taken seriously. I myself am a language nerd. I love etymology, studying languages, and speak a few with varying proficiency. I truly believe one must learn a language to develop the critical skills we need to succeed in a global economy and workforce, like cultural awareness, negotiation, empathy, humor; however, I don’t believe we need them to be successful in an academic setting. Not anymore.
How do we get from there to here?
Knowing that universities are not known for moving quickly, but with how quickly technology is advancing there is no time to wait, a happy medium would be to set up a transition in phases:
Phase 1 (Years 1 - 2): Pilot Programs
Universities pilot AI-mediated instruction in specific programs
Maintain English requirements but offer AI-supported alternatives
Collect data on learning outcomes, student success, and implementation costs
Develop best practices for AI-assisted academic assessment
Phase 2 (Years 3 - 4): Expand and Refine
Scale successful models across more institutions and disciplines
Governments pilot alternative visa assessment methods
Establish industry standards for academic AI translation
Create support systems for students transitioning between models
Phase 3 (Years 5+): New Normal
English proficiency becomes one option among many for demonstrating readiness
AI infrastructure is standard in international education
Assessment focuses on subject mastery, not language gatekeeping
New metrics for student success that value multilingual competence
Can we also make a shift in our daily teaching and administrative lives? Celebrate multilingual competence. Value students’ linguistic diversity as strength. AI-mediated learning highlights adaptability, not deficiency.
Where ESL Programs Can Fit, If They Pivot
In the U.S., ESL and Intensive English Programs (IEPs) have long served dual roles: gateways for students and revenue generators for universities. Since 2015, enrollment in IEPs has dropped, forcing closures. These programs face a choice: evolve or become obsolete. AI may threaten IEPs, but it also opens a path forward for nimble administrators to evolve their programs from gatekeepers to bridges.
IEPs’ value lies not in teaching students against foreign thresholds but empowering them with AI literacy, communication skills, and intercultural competence. Forward-thinking programs are already adapting. Brent Warner at Irvine Valley College uses AI tools to help ESL students take control of their learning. (Check out his website!) Research shows AI applications enhance motivation, engagement, and reduce anxiety in language learning contexts, though AI must complement, not replace, human teaching.
The future ESL program won't teach to arbitrary proficiency thresholds but will focus on what AI cannot provide: cultural nuance, academic discourse communities, professional communication skills, and the human elements of language that remain essential even with AI translation.
The Institutional Counterforce: Academic Capital
Critics worry about creating AI dependency. Won't students miss crucial cognitive development that comes from language learning? This concern deserves serious consideration.
Language learning does develop cognitive flexibility, cultural awareness, and interpersonal skills. But we must question whether forcing this development through high-stakes testing in a single dominant language is the only or best path. Students can still choose to develop English proficiency. Many will, given its current advantages. The difference is choice versus coercion. Navigating AI-mediated multilingual environments requires its own sophisticated skill set: prompt engineering, cross-cultural communication through technology, and critical evaluation of AI output. These are not a lack of cognitive skills, but a shift in the required cognitive skills. These are the competencies of the future workplace. Language learning is a crucial scaffold for developing complex, abstract thought. Relying on AI might weaken the development of critical thinking, memory, and cognitive flexibility that comes from the struggle of linguistic encoding.
The biggest systemic friction to this shift comes not from testing companies but from the Anglophone academic publishing ecosystem. Research visibility and career advancement are inextricably linked to publishing in high-impact journals, the vast majority of which are published in English.
Faculty promotion and tenure decisions at global universities are often based on the impact factor of journals where they publish. The highest-ranked journals in science, technology, medicine, and social sciences are overwhelmingly published in English, creating a vicious cycle of prestige. A scientist in Brazil or Korea who publishes a breakthrough paper in their native language risks zero international visibility and limited career impact compared to one who struggles to translate it into a second-rate English journal. For faculty, resistance to de-centering English is not ideological for them either; it is professional survival. They will argue: "How can we admit students who cannot read the foundational literature or publish their own work in the journals that define our field?"
Faculty Labor and Linguistic Load
The shift to AI-mediated assessment and instruction places a new, unpaid burden on faculty. Even with AI assistance, reading a graduate thesis translated from Mandarin or Arabic will be slower and more demanding for a professor whose primary language is English. How do they verify that the nuance of the argument has survived the translation process? Faculty will need training in prompt engineering for academic use and in how to evaluate AI-mediated student work for both content mastery and academic integrity. There is also unacknowledged labor. Universities must recognize and compensate this increased linguistic labor. Without institutional support (reduced teaching loads, specialized training), faculty will resist the change by clinging to the simplest gatekeeper: mandatory English proficiency. This resistance is exacerbated by the liability and assessment integrity challenge. If a faculty member cannot read the student's original language, they are, in effect, grading the AI's translation. If the translation is subtly flawed, and the faculty awards a high grade, they risk devaluing the degree and their program's reputation. This is why assessment must pivot: it must focus on demonstrating subject mastery through non-linguistic methods (e.g., code, mathematical models, practical demonstrations) alongside the AI-mediated written work, providing the faculty a secondary, non-linguistic verification track.
A Path to Multilingual Publishing
The solution is not just AI in the classroom, but AI in the publishing infrastructure. Publishers must invest in AI tools to efficiently translate, edit, and cross-reference submissions. Journals could publish a core English abstract alongside the full paper in the author’s original language, all searchable by AI. Institutions must update promotion and tenure guidelines to value and reward publications in high-impact, non-English journals, thus breaking the monolingual publishing monopoly. This is the deeper fight: English dominance is woven into the very fabric of academic capital. Until we decolonize the metrics of faculty success, resistance will remain a powerful force.
Moving Forward
Universities and visa regimes that ignore the AI revolution risk obsolescence. While some institutions boast "future readiness," their loyalty to English-only standards undermines that claim. Aspiring students will migrate to more inclusive, AI-forward institutions that empower multilingual access. There are concrete paths forward:
For Universities:
Begin piloting AI translation in low-stakes courses
Develop alternative assessment methods that value subject mastery over language gatekeeping
Invest in AI infrastructure as core educational technology
Partner with sending countries to develop multilingual pathways
For Governments:
Pilot alternative visa assessment methods based on institutional accountability rather than individual testing
Recognize AI-mediated communication as valid for student success
Align visa policy with domestic language diversity values
For Testing Companies:
Pivot from gatekeeping to genuine support services
Develop AI-enhanced assessment tools that evaluate readiness, not just language
Consider subscription models for ongoing support rather than high-stakes, one-time tests
For Students:
Develop AI literacy alongside language skills
Build portfolios that demonstrate competence beyond test scores
Advocate for inclusive assessment methods at target institutions
Welcome to the era where multilingualism, supported and normalized by AI, reshapes access and equity. The death of English dominance is an inflection point. Will universities and governments adapt, or be left behind? I am not advocating for the death of English or abandoning language learning, but for separating its intellectual value from its role as an academic choke point. Students should be encouraged to learn English (or Mandarin, or Spanish) for the joy and skills it provides, not because a $200 test dictates their entire future.