Day 4: AI-Powered Assessment for English Teaching and Learning
Promise
After this lesson you can choose the right type of assessment for your class and use one free AI tool to check speaking, writing, listening or grammar, while you, the teacher, make the final decision.
Audience: English teachers and trainee teachers, including adults in Réunion (French first language); plain English A2–B1 in the new parts.
Time to read: 25 minutes · Format: web page
Before you start
You will be able to: name four types of assessment, match an AI tool to a skill, and turn an AI result into your next lesson step.
Three words you need first:
- assessment — a way to find out what learners can do.
- formative — small, regular checks during learning, not for a grade.
- rubric — a list of criteria for a score, for example fluency and grammar.
You need: about 25 minutes. A class you teach in mind helps.
📌 Day 4: AI-Powered Assessment
📌 Introduction: The AI Assessment Revolution
Assessment in ELT has evolved from **pen-and-paper tests** (1980s) to **computer-adaptive testing** (2000s) to **AI-powered, personalized feedback** (2026). Today’s tools don’t just measure learning—they predict gaps, adapt in real-time, and scaffolding next steps.
🔍 Why AI?
AI solves three core problems in assessment:
- Bias: Human graders favor fluency over accuracy; AI evaluates predefined criteria (e.g., CEFR descriptors).
- Scalability: AI can assess 100 speaking samples in the time a human grades 5.
- Actionable Data: AI doesn’t just score—it diagnoses (e.g., “Your learners struggle with past perfect form, not use“).
Example: Cambridge’s Linguaskill uses AI to score writing/speaking against CEFR bands instantly, with feedback like:
Segment 1 of 3
📚 Assessment Types in 2026
1. Formative Assessment
Definition: Ongoing, low-stakes feedback to improve learning (not grade it).
🔍 Key Features
- Frequency: Daily or weekly (e.g., exit tickets, peer feedback).
- Purpose: Identify gaps during learning (e.g., “80% of students confused ‘affect’ vs. ‘effect'”).
- AI Tools:
- Pearson Versant: Scores speaking in real-time.
- EduApps: Auto-grades short writing tasks.
Example: After a lesson on conditionals, AI analyzes students’ chatbot conversations and flags:
🛠 Practical Activity
AI-Powered Exit Ticket:
- Students record a 30-second audio response to: “Explain a 2nd conditional rule with an example.”
- Upload to Speechace (free tier).
- AI generates:
“Accuracy: 75% · Fluency: 80% · Error: ‘If I will go’ → Teach: ‘If + subject + past simple’.”
2. Summative Assessment
Definition: High-stakes evaluation of learning after instruction (e.g., final exams).
🔍 Key Features
- Frequency: End of unit/term (e.g., Cambridge B2 First).
- Purpose: Certify proficiency (e.g., “This student is CEFR B2”).
- AI Tools:
Example: Oxford’s Oxford Test of English uses AI to:
🛠 Practical Activity
AI Mock Exam:
- Students take a free EF SET practice test (AI-scored).
- AI generates:
“Listening: C1 · Reading: B2 · Focus: Improve skimming for gist in reading tasks.”
- Teacher uses data to plan review sessions.
3. Diagnostic Assessment
Definition: Identifies strengths/weaknesses before instruction to personalize learning.
🔍 Key Features
- Frequency: Start of course/unit (e.g., placement tests).
- Purpose: Answer: “What does this student already know?”
- AI Tools:
- Oxford Placement Test: AI adapts difficulty in real-time.
- Cambridge English Test: 25-question adaptive quiz.
Example: British Council’s Listening Diagnostic:
🛠 Practical Activity
AI Gap Analysis:
- Students complete a 15-minute EF SET Quick Check.
- AI reports:
“Grammar: 80% · Vocabulary: 60% · Weakness: Phrasal verbs. Assign: British Council Phrasal Verbs Unit.”
4. Ipsative Assessment
Definition: Compares students to their past performance (not peers/norms).
🔍 Key Features
- Frequency: Ongoing (e.g., portfolio reviews).
- Purpose: Answer: “How much has this student improved?”
- AI Tools:
Example: A student’s writing samples over 3 months:
December: “Last week, I went to the park with my friends.” (A2) → +1 CEFR sub-level.
🛠 Practical Activity
AI Progress Portfolio:
- Students record a 1-minute speaking sample monthly (e.g., “Describe your weekend”).
- Upload to VoiceThread; AI compares fluency/accuracy to prior samples.
- Teacher + student review:
“↑ Fluency: 15% faster · ↓ Errors: ‘goed’ → ‘went’ (fixed!). Next goal: Use 2 past perfect sentences.”
Visual 1 — Timeline
What it shows: when each type of assessment happens in a course.
How to draw it: Draw one line from the start to the end of a course and mark the four types on it.
The point it makes obvious: Diagnostic is at the start, formative is all the way through, summative is at the end, ipsative compares you with yourself.
START ------------------------------------------ END [Diagnostic] [formative] [formative] [formative] [Summative] "what do you "how is it going?" "what level know now?" are you now?" Ipsative: you in March vs you in June (no other students)
EXAMPLE (class in Réunion): An A2 adult evening class: a short placement test in week 1 (diagnostic), an exit ticket each Thursday (formative), a mock exam in the last week (summative), and a monthly 1-minute voice recording that each learner compares with the last one (ipsative).
Check 1: A learner compares this month’s recording with last month’s recording. Which type of assessment is this?
Show the answer
Segment 2 of 3
🤖 AI Tools for Assessment: A Practical Guide
1. AI for Speaking Assessment
| Tool | Features | Example Use Case | Link |
|---|---|---|---|
| Speechace |
|
Student records: “My favorite holiday was…” → AI flags: “‘was go’ → Teach: ‘went’.”
|
Try Speechace |
| ELSA Speak |
|
Student says: “I leef in Paris.” → AI: “/iː/ → /ɪ/. Practice: ‘ship’ vs. ‘sheep’.”
|
Try ELSA |
| Pearson Versant |
|
Job applicant takes test → AI report: “Fluency: B2 · Grammar: B1 · Recommend: Practice complex sentences.”
|
Try Versant |
💡 Pro Tip: Combine tools! Use ELSA for pronunciation + Speechace for grammar/fluency.
2. AI for Writing Assessment
| Tool | Features | Example Use Case | Link |
|---|---|---|---|
| Grammarly |
|
Student writes: “I goed to the store.” → AI: “Did you mean ‘went’?”
|
Try Grammarly |
| Quill.org |
|
Activity: “Combine these sentences using ‘although’.” → AI scores responses and suggests scaffolding.
|
Try Quill |
| Write & Improve |
|
Student submits: “My weekend was bore.” → AI: “B1 · Error: ‘bore’ → ‘boring’. Tip: Use ‘-ing’ for adjectives.”
|
Try Write & Improve |
💡 Pro Tip: Use Write & Improve for CEFR alignment + Quill for targeted practice.
3. AI for Listening Assessment
| Tool | Features | Example Use Case | Link |
|---|---|---|---|
| ELSA Listen |
|
Student listens: “The meeting is at 3 PM.” → Types: “The meet is at 3.” → AI: “Error: ‘meet’ → ‘meeting’.”
|
Try ELSA Listen |
| British Council Listen & Watch |
|
Student watches a B1 video → Takes quiz → AI: “70% on detail questions. Practice: B1 Listening for Gist.”
|
Try BC Listen |
| Sanako Connect |
|
Teacher assigns a lecture → AI reports: “Class average: 65% on inference questions. Focus: Teach signal words (‘however’, ‘therefore’).”
|
Try Sanako |
4. AI for Grammar/Vocabulary
| Tool | Features | Example Use Case | Link |
|---|---|---|---|
| NoRedInk |
|
Student practices: “She go to school.” → AI: “Error: ‘go’ → ‘goes’. Next: 3rd person -s rules.”
|
Try NoRedInk |
| Kahoot! |
|
Teacher inputs: “Past simple irregular verbs.” → AI generates a 10-question quiz with explanations.
|
Try Kahoot! |
| Quizlet |
|
Student studies “phrasal verbs” → AI quiz flags: “‘look up’ confused with ‘look for’.”
|
Try Quizlet |
💡 Pro Tip: Use NoRedInk for grammar and Quizlet for vocab retention.
Visual 2 — Decision table
What it shows: which tool fits which skill.
How to draw it: Draw four columns, one for each skill, and write one tool from this page under each.
The point it makes obvious: Pick the tool by the skill you want to check, then check its result yourself.
SPEAKING WRITING LISTENING GRAMMAR/VOCAB
Speechace Write & Improve BC Listen&Watch Quizlet
ELSA Speak Quill (quiz scores) Kahoot!
\ | | /
+------> teacher reads the result and decides <------+
EXAMPLE (class in Réunion): A B1 learner writes 100 words about a weekend in Saint-Pierre, puts it into Write and Improve, and brings the feedback to class. The teacher picks one error type for the whole group.
Check 2: You want feedback on pronunciation sounds. Which kind of tool from the table should you choose?
Show the answer
🛠 AI Assessment Designer
Use this tool to generate a **custom AI-powered assessment plan** for your class. Select your focus and level:
Segment 3 of 3
🎯 Practical Activities: AI in Your Classroom
Activity 1: AI Speaking Rubric
Use Speechace to assess a student’s 1-minute response to: “Describe a time you felt proud.”
Activity 2: AI Writing Feedback
A student submits this to Write & Improve:
What feedback will the AI most likely provide?
Activity 3: AI Listening Diagnostic
Your B1 students take the EF SET Quick Check. The AI reports:
What’s the best next step?
Visual 3 — Loop
What it shows: how AI results feed the next lesson.
How to draw it: Draw four boxes in a circle with arrows.
The point it makes obvious: The AI score is only useful when it changes what you teach next.
[ Learners do a task ] --> [ AI gives a score ]
^ |
| v
[ Teach the gap ] <------ [ Teacher checks the score ]
EXAMPLE (class in Réunion): The quiz shows that most of the group miss inference questions. Next lesson, the teacher teaches signal words like “however” with a short B1 listening, then gives a new quiz.
Check 3: In Activity 1, which thing will a speaking tool not judge well?
Show the answer
📚 References & Further Reading
Academic sources:
- Bachman, L. F., & Palmer, A. S. (2010). Language Assessment in Practice. Oxford University Press. [corrected 28 Oct 2026: not a 2nd edition]
- Fulcher, G., & Davidson, F. (2007). Language Testing and Assessment: An Advanced Resource Book. Routledge. [corrected 28 Oct 2026]
- Cambridge English. (2020). AI in Language Assessment (title and year UNSOURCED). Read more.
- Oxford University Press. (2021). Assessment for Learning (title and year UNSOURCED). Explore.
Free tools:
- Speechace (speaking)
- Write & Improve (writing)
- EF SET Quick Check (diagnostic)
- Quill (grammar)
Worked example — one writing task, from AI score to next lesson
The task: A B1 learner writes: “Last weekend I go to the beach with friends. We eated picnic.” Use an AI tool and plan the next step.
| Step | What happens | Why |
|---|---|---|
| 1 | Put the text into a writing tool such as Write and Improve. | You get fast feedback on accuracy. |
| 2 | Read the result yourself: past simple errors (“go”, “eated”). | The teacher checks that the AI is right. |
| 3 | Look at the whole class: do many learners make the same error? | One error pattern is worth a lesson. |
| 4 | Plan 10 minutes on irregular past verbs, then ask learners to rewrite. | The score changes what you teach. |
| 5 | Give feedback on ideas with your own rubric. | AI tools are weak on ideas and content. |
Self-explanation prompt: Why is step 2 needed? Because an AI score can be wrong, and only the teacher knows the class and the task.
Your turn (guided): Do the same with one speaking recording. Step 1: choose a tool from the speaking table. (Check: does your next lesson come from the result?)
Common mistakes
| Mistake | Why it happens | Fix |
|---|---|---|
| Trusting the AI score without checking | It looks precise. | Read a few results yourself before you act. |
| Using only summative tests | Exams feel official. | Add short formative checks every week. |
| Grading ideas with a speaking tool | The tool gives one score. | Use a human rubric for ideas and content. |
| Uploading learner data to any app | Free tools are easy to try. | Check the tool’s privacy rules and ask learners first. |
Glossary
| Term | Plain definition | Where first used |
|---|---|---|
| formative assessment | small checks during learning | Segment 1 |
| summative assessment | a test at the end, often for a grade or level | Segment 1 |
| diagnostic assessment | a check at the start to see what learners know | Segment 1 |
| ipsative assessment | comparing a learner with their own past work | Segment 1 |
| CEFR | Common European Framework of Reference: levels A1 to C2 | Segment 1 |
| auto-marker | software that scores speaking or writing | Segment 2 |
| inference | understanding what is meant but not said | Segment 3 |
Quick review
- Which type of assessment happens at the start of a course?
- What does ipsative assessment compare?
- Name one AI tool for writing on this page.
- Who marks Speaking in the Oxford Test of English?
- What should you do after an AI tool gives a score?
Show all five answers
2) A learner with their own past work.
3) For example Write and Improve, Quill or Grammarly.
4) Trained human assessors (the computer marks Listening and Reading).
5) Check it yourself, then plan the next lesson from the gap it shows.
Fact check, 28 October 2026
Checked against the publishers’ own pages: Cambridge Linguaskill does use AI auto-marking for Speaking and Writing, with human examiners when the auto-marker is not sure. I corrected the Oxford Test of English example (Speaking is marked by human assessors) and two book references (Bachman and Palmer 2010; Fulcher and Davidson 2007). These statements are UNSOURCED and still need a source: the history line (1980s, 2000s, 2026); “AI can assess 100 speaking samples in the time a human grades 5”; the claim that human graders favour fluency over accuracy; the free-tier details and features listed for each tool; “EduApps” and “ELSA Listen”; and the two 2020/2021 reading titles.
Status: rewritten in lecture-notes style on 28 October 2026. Not tested with readers.