TeflToday is a teacher-built AI platform offering 10 tools (8 premium + 2 free) for TEFL, TESOL, and ESL educators. It generates CEFR-aligned lesson plans, grades student writing with diagnostic feedback, and produces Cambridge exam practice materials. Created by Ian L. Evans, a British TEFL teacher with 20+ years of classroom experience.

How accurate is the AI Writing CEFR Grader compared with a human teacher?

The grader assesses writing against four rubric dimensions used by Cambridge examiners: Content, Organisation, Language control, and Task Achievement. It provides a CEFR level estimate with detailed error categorisation. However, it is a teaching support tool — not an official Cambridge or IELTS score. Teachers should use it alongside their own professional judgement, especially for borderline cases or high-stakes assessment.

What rubric categories does the CEFR Writing Grader use?

The CEFR Writing Grader evaluates four dimensions: (1) Content — how fully the writing addresses the task; (2) Organisation — paragraph structure, coherence, and linking devices; (3) Language — grammar range and accuracy, vocabulary precision, and collocation usage; (4) Task Achievement — appropriacy and register for the task type (essay, email, report, review). Each dimension receives specific feedback with examples from the student's text.

Do you store student writing? Is it used to train AI models?

Student writing submissions are processed in real time for grading and feedback only. They are not stored permanently, not used to train AI models, and not shared with third parties. TeflToday serves EU-based teachers and handles data in accordance with GDPR principles.

Which Cambridge exams and parts are supported?

TeflToday supports PET (B1 Preliminary), FCE (B2 First), CAE (C1 Advanced), and CPE (C2 Proficiency). Covered paper types include Use of English (Parts 1–4: multiple choice cloze, open cloze, word formation, key word transformations), Speaking prompts with structured assessment criteria, and Writing tasks with CEFR-graded feedback.

How much does TeflToday cost?

€6 for 30 days of full access to all 8 premium tools. No subscription, no auto-renewal, no credit card required to sign up. The English Level Test and C.I.P.R.E.S. C1 Grammar Tool are always free with no account needed.

What are the limitations of AI CEFR writing graders?

AI graders work best for identifying grammar errors, vocabulary range, and structural issues. They are less reliable for assessing creativity, cultural nuance, humour, and highly idiomatic expression. The CEFR level returned is an estimate based on rubric criteria, not an official certification. Teachers should treat AI feedback as a first pass and apply their own judgement for final grading decisions.

Can I export lesson plans and worksheets?

Lesson plans and grading feedback are generated as formatted text that can be copied, printed, or pasted into documents (Google Docs, Word, PDF). The platform is web-based and works on any device with a browser — no software installation required.

Generative AI Policy for Language Schools: Responsible Use Guidelines

Generative AI in Education: Responsible Use Policies & Institutional Guidelines

Quick answer

A responsible generative-AI policy for language education keeps a human in the loop for decisions that affect learners, protects student data privacy, is transparent about when and how AI is used, and avoids relying on AI-detection tools, which research shows are not reliable enough to judge academic honesty.

Scope of this guide

This page summarises the principles that reputable bodies — including the U.S. Department of Education’s Office of Educational Technology and the TeachAI initiative — recommend for responsible generative-AI use in education, and turns them into practical policy points for language schools and independent teachers. It is general guidance, not legal advice.

Rules and available tools change quickly. Treat any AI policy as a living document, review it regularly, and check the requirements that apply in your own country and institution.

Human in the loop

The U.S. Department of Education’s guidance emphasises keeping a “human in the loop”: AI can assist, but a qualified person should remain responsible for decisions that affect learners — grading, feedback, and anything with consequences. AI output should be reviewed and can be overridden, never applied automatically to a student.

In practice, human-in-the-loop means:

A teacher reviews AI-generated feedback or grades before they reach a learner.
AI suggestions are treated as drafts, not final decisions.
Learners can question an AI-assisted decision and reach a human.

Student data privacy

Pasting student work or personal information into a commercial AI tool may share that data with a third party. In the United States, FERPA protects the privacy of student education records; comparable regimes (such as the GDPR in Europe) apply elsewhere. A safe default is to avoid entering personally identifying student information into general-purpose AI tools unless your institution has a data-processing agreement that permits it.

Anonymise before you paste: remove names and identifying details from student work before using an AI tool to help with feedback, unless a vetted, agreement-backed tool is in place.

AI detectors are not reliable evidence

It is tempting to run student work through an “AI detector”, but research and the vendors’ own disclaimers indicate these tools produce false positives and false negatives and can be biased against non-native English writers. A detector score should not be treated as proof of dishonesty. Address academic integrity through task design, process evidence (drafts, notes) and conversation with the learner instead.

What to do instead of trusting a detector score

Concern	Weak approach	Stronger approach
Suspected AI-written work	Rely on a detector percentage	Ask for drafts/process; discuss the work with the learner
Discouraging misuse	Ban and police	Design tasks that require personal/process evidence
Transparency	No policy	Clear, shared rules on when AI use is allowed and how to cite it

Building an institutional policy

A workable policy names what is allowed, what is not, and how AI use should be disclosed. It should cover permitted uses, data-privacy rules, the human-in-the-loop requirement for assessment, and a simple way for learners to cite or acknowledge AI assistance.

Frequently asked questions

Why does the U.S. Department of Education advocate a “human-in-the-loop” approach to AI?

The Department’s guidance stresses that AI should support educators rather than replace their judgement, because decisions that affect learners — grading, feedback and interventions — carry consequences that require professional and ethical responsibility. Keeping a qualified human reviewing and able to override AI output guards against automated errors and bias, preserves accountability, and keeps a person answerable to learners for decisions made about them.

Are AI detection tools reliable for identifying student academic dishonesty?

No. Available evidence and the vendors’ own caveats show AI-detection tools generate both false positives and false negatives and can be biased against non-native English writers. A detection score is not proof that work was produced by AI, so it should not be the basis of an academic-integrity decision. Institutions are better served by task design, requiring drafts and process evidence, and discussing the work directly with the learner.

What student data privacy laws (e.g., FERPA) apply when using commercial generative AI tools?

In the United States, FERPA protects the privacy of student education records, and entering identifiable student information into a commercial AI tool may constitute sharing that data with a third party. Other jurisdictions have comparable laws, such as the GDPR in Europe. Unless the institution has a data-processing agreement permitting it, the safe default is to anonymise student work and avoid entering personally identifying information into general-purpose AI tools.