ELT AI Observatory
Dated, source-traced observation of AI developments in English Language Teaching, with the state of the evidence stated.
Pedagogy First. AI Enhanced. Future Ready.
What this is, and is not. The ELT AI Observatory is a curated, periodic, human-reviewed record compiled by NIR Academy. It is not peer-reviewed, not exhaustive, not real-time, and not an official or regulatory authority. It does not evaluate tools or institutions, set requirements, give legal advice, or recommend action.
What is the ELT AI Observatory?
The ELT AI Observatory is NIR Academy’s research-oriented record of how AI is developing in English Language Teaching. It documents what has been observed, traces each item to its sources, and states how strong the available evidence is.
The problem it addresses. Information about AI in ELT is fragmented and fast-moving, and it mixes documented facts, published research, vendor claims, and unverified talk. Teachers, academic leaders, and researchers cannot easily see which statements are documented, which are researched, and which are only asserted. The Observatory provides a stable, evidence-labelled reference so that this distinction is visible.
Publication status. The Observatory is being established. This page defines what it will publish and how evidence is treated. Outputs appear below as they are completed, with the coverage window stated on each.
What the ELT AI Observatory monitors
Coverage is English-language ELT and the sources the Observatory can access. Language, regional, and access bias in coverage is stated in each output.
Emerging AI developments
New model capabilities, modalities, and releases with a stated ELT-relevant use, such as speech, writing feedback, or conversational practice.
AI-in-ELT research
New studies, reviews, and datasets on AI and language learning, teaching, or assessment.
Tools and capabilities
The documented existence and stated functions of AI tools used in ELT. Never suitability or comparison.
Pedagogical trends
Reported changes in how AI is used in teaching and learning practice, with the source of each report.
Assessment developments
Changes by exam providers and assessment bodies regarding AI use, and published validity discussions.
Teacher and learner implications
Implications as reported or observed in sources, attributed to those sources.
Institutional developments
Publicly stated institutional decisions on AI in ELT contexts.
Policy and regulatory developments
Publication, scope, and effective dates of laws, regulations, and official guidance. Not legal advice.
Risks and emerging issues
Risks and concerns as documented or reported in sources.
Observation is one function, not all of them
The Observatory observes and labels evidence. Interpretation, evaluation, governance, and implementation are separate functions performed elsewhere in NIR. Its only permitted interpretation is evidence-bounded synthesis: stating where sources converge, conflict, or are silent. It never states what an institution or teacher should do.
| Function | What it means | Who performs it |
|---|---|---|
| Observation | Recording that a development occurred or a source exists, with date and provenance | ELT AI Observatory |
| Evidence labelling | Stating what a source supports, and how strongly, for a specific claim | ELT AI Observatory |
| Interpretation | Analysis applied to decisions, institutions, or contexts | NIR ELT Intelligence™ |
| Evaluation | Judging the suitability or quality of a specific tool or use | AI Evaluation & ELT Expertise; Compare AI Tools |
| Governance | Deciding what an institution must establish and oversee | AI Governance for ELT™ |
| Implementation | Documenting and operationalising requirements as policy | AI Policy Center |
How the ELT AI Observatory labels evidence
Each claim carries one descriptive label. A label attaches to a specific claim in a specific context, never to a tool, a company, or a topic. These are descriptive labels used by NIR, not a formal grading system and not a claim of systematic-review rigour.
Documented
A primary source confirms the item exists or was published. Says nothing about effectiveness.
Research-supported
One or more published studies address the claim. Number, design, and context are stated.
Mixed or contested
Sources conflict, or findings differ by context.
Preliminary
Early, limited, or single-context evidence, including preprints.
Reported
Present in secondary reporting or professional reports, not traced to primary evidence.
Unverified signal
In circulation, with no verifiable source. Never treated as evidence.
What the Observatory publishes
Six output types, each with a defined purpose and evidence level. Cadence is stated on the site and reflects actual capacity.
Evidence Signal
One development or claim, with source, tier, evidence label, what the source does and does not show, and review date.
Evidence level: carries an explicit label; no interpretation.
ELT AI Developments
A dated register of verifiable developments: capabilities, releases, regulatory texts, assessment changes.
Evidence level: Documented only.
Observatory Brief
A concise digest of items added during the coverage window, each with its label.
Evidence level: observation and labels; a brief note on conflicts or gaps.
Emerging Issues
Short notes on issues that appear across several sources, framed as open questions.
Evidence level: Preliminary or Mixed; attributed synthesis.
Trend Report
Recurring patterns across a longer window, with an evidence map.
Evidence level: labelled synthesis; no causal claims beyond the sources; no prediction.
Annual Observatory Report
Coverage, methods, developments, state of evidence by theme, changes in labels, research gaps, limitations, and corrections.
Evidence level: labelled synthesis; no readiness scores or rankings.
Methodology and limits
Source hierarchy
The hierarchy depends on the claim being made. A vendor document is the best source that a feature exists, and is not evidence that it works.
| Tier | Source type | Can support |
|---|---|---|
| Tier 1 | Primary documents: legal and regulatory texts, official guidance, standards documents, exam-provider announcements, institutional policy documents, vendor technical documentation | That something was published, announced, released, or enacted, and what it states |
| Tier 2 | Research literature: peer-reviewed articles and proceedings, systematic reviews, dissertations, preprints (flagged as not peer-reviewed) | Claims about effects, practices, and outcomes, within the study’s design and context |
| Tier 3 | Professional and organisational reports: association and exam-board reports, white papers, conference presentations | Reported practices and perspectives; flagged as not peer-reviewed |
| Tier 4 | Secondary reporting: quality press, trade press | Pointers to Tier 1 to 3 sources; not evidence on their own |
| Tier 5 | Signals: practitioner posts, community discussion, vendor marketing, social media | Recording that a claim is in circulation; never evidence of an effect |
Four kinds of statement
| Kind | Definition | Treatment |
|---|---|---|
| Observation | A dated fact about existence: a release, publication, regulation, or announcement | Published, with source and date |
| Evidence | What a specific source shows about a specific claim, with its design, context, and limits | Published, with an evidence label |
| Interpretation | Evidence-bounded synthesis: where sources converge, conflict, or are silent | Published only in Trend Reports, Emerging Issues, and the Annual Report, labelled as Observatory synthesis |
| Prediction | A forecast about future developments or effects | Not published. Open questions may appear as “Questions to watch”, without probability or direction |
Inclusion, verification, and review
- Included: items concerning AI and ELT, with an identifiable author or issuing body, a date, a retrievable source, and a stated coverage window.
- Excluded: items that cannot be traced to a source, effectiveness claims without stated methods, general AI news without a clear ELT implication, duplicates, promotional content, and anything identifying private individuals.
- Verified: traced to the most authoritative available source, with authorship, date, provenance, context, and relevant corrections or retractions checked where applicable. Any relationship between NIR Academy and the subject of an entry is disclosed.
- Reviewed: each Evidence Signal is re-reviewed at least every six months, or sooner if new evidence appears. Errors are corrected openly in a dated corrections log.
- Uncertain claims: labelled Preliminary, Reported, or Unverified signal, never placed in headlines, worded with calibrated language, and accompanied by what evidence would change the label.
What the ELT AI Observatory does not do
The Observatory must not:
- Establish governance requirements. Regulatory developments are recorded as observations only.
- Create institutional policies. Policy text belongs to the AI Policy Center.
- Certify or accredit any person, tool, or institution.
- Evaluate institutions. It holds no institutional data and reviews no practice.
- Provide legal advice. Legal entries are not interpretations of law.
- Replace academic research. It points to research; it does not substitute for it.
- Claim causal effects without evidence. Causal language only where the source’s design supports it, and attributed to the source.
- Rank tools without an explicit evaluation methodology. No “best of” lists.
- Duplicate NIR ELT Intelligence™. It offers no analysis for decisions.
- Duplicate the Readiness Index. It publishes no readiness scores.
- Become a generic AI news feed. Every item must pass the inclusion criteria and carry a label.
It also does not forecast, and it does not claim peer review, empirical validation, official authority, or real-time monitoring.
Examples of primary sources
The Observatory traces claims to primary sources. The links below are examples of the kinds of source involved, listed for orientation. These bodies are not partners of NIR Academy, and inclusion here is not an endorsement. Consult current versions directly.
- UNESCO: artificial intelligence in education official guidance on AI in education.
- OECD.AI Policy Observatory general AI policy information, not specific to ELT.
- EU AI Act, official text on EUR-Lex an example of a primary legal text. Entries record such texts and do not interpret them.
- Common European Framework of Reference for Languages (Council of Europe) a shared reference for describing language proficiency.
Frequently asked questions
What is the ELT AI Observatory?
What does the ELT AI Observatory monitor?
What does the ELT AI Observatory not do?
How is evidence labelled?
Is the Observatory peer-reviewed or real-time?
How is the Observatory different from the ELT AI Readiness Index™?
Follow the evidence, not the headlines
Start with the methodology, or suggest a source for consideration. Suggestions do not guarantee inclusion.
Read the methodology Suggest a sourceThe ELT AI Observatory is compiled by NIR Academy and is currently a working draft (v0.1). Evidence labels are descriptive and are not a formal grading system. Entries record what sources state and do not evaluate tools or institutions, give legal advice, or recommend action. NIR Academy offers related products and services; relationships relevant to an entry are disclosed on that entry. Last reviewed: October 2026.

