AI tools for English teachers are software that uses artificial intelligence to plan lessons, give students feedback, or give students someone to practice with. Only some of them add speaking practice. The rest save the teacher time on other work, which is useful but solves a different problem.
Choosing one is like hiring a teaching assistant. Before you read any job applications, decide which job you need done.
Illustrative scenario: Ms. Ploy leads the English department at a secondary school in Thailand. Her teachers already use a chatbot to draft worksheets. The school's new goal is more speaking practice for every student, and the principal has asked her to recommend one tool by the end of term. She appears throughout this article.
How many English teachers already use AI tools?
Most teachers have tried AI tools, and use is growing fast, but much of it is for planning rather than student practice. In a British Council survey of 1,348 English language teachers in 118 countries (Edmett et al., 2023), 24% said they used no AI tools at all. Most of the rest did.
Use has grown since. The EdWeek Research Center (2026) found that 61% of US teachers used AI in their work in 2025, up from 34% in 2023. Both figures cover all kinds of use, so they tell you teachers are ready to try AI. They don't tell you students are speaking more.
What types of AI tools do English teachers use?
AI tools for English teachers can be grouped into four types, and only two of them put students in front of a microphone. The table compares them on how they work, not on which is better.
| Type | What it does | Adds speaking minutes for students? | Who gets feedback | What the teacher sees |
|---|---|---|---|---|
| General chatbot for planning | Drafts lesson plans, worksheets and questions | No | The teacher | The draft it produces |
| Writing feedback tool | Comments on student essays and paragraphs | No | Students, on writing | Comments and scores on writing |
| Speech scoring and feedback | Listens to a student read or speak and scores features such as pronunciation and fluency | Yes | Each student, on speaking | Scores by student and by task |
| Conversation practice | Lets students hold a spoken conversation with an AI character | Yes | Each student, after the conversation | Depends on the tool: scores, transcripts or nothing |
For Ms. Ploy's goal, the first two types are useful but off-target. The decision is between the last two, or a platform that combines them.
How does AI speaking practice work in a lesson?
The teacher sets a task, every student speaks to the tool at the same time, and the tool returns feedback to each student. In practice, a lesson usually runs in five steps:
- The teacher picks a task linked to the unit: words, a reading passage, or a conversation scenario.
- Students speak to the tool, in class or at home, using a headset or a quiet corner.
- The tool gives each student feedback on what it can measure, such as sounds, pace and pauses.
- Students try again, using one piece of feedback.
- The teacher checks the class results and chooses who to hear live in the next lesson.
The value is in steps 2 and 4. Every student speaks, and every student gets a second try, without the rest of the class waiting.
What does the research say about AI conversation practice?
Studies show a moderate positive effect on speaking, but most are small, so treat the evidence as promising, not settled. A meta-analysis by Hou and Min (2025) pooled 16 studies of dialogue-based systems and found a medium overall effect on second-language speaking.
Two details matter for schools. Only five of the 16 studies involved K-12 students, and 15 of the 16 had fewer than 60 learners. The British Council report makes a similar point: AI tools can help students practice outside class and may reduce their fear of speaking, but more research is needed to know whether those benefits last.
How should a school choose an AI tool for speaking practice?
Start with the speaking problem you have, then test one tool with real classes before you buy. Ms. Ploy's checklist looks like this:
- Name the goal. For example: "Every Grade 8 student speaks for at least five minutes a week, with feedback."
- Check fit with your textbook. Can teachers turn their own lesson material into speaking tasks, or must they use the tool's content?
- Check the feedback. Which speaking features does it cover, and is the feedback simple enough for a 13-year-old to act on?
- Check what the teacher sees. Look for a class view that shows who hasn't practiced and who needs help.
- Check safety and data. Age suitability, where recordings are stored, and who can see students' conversations.
- Run a short trial. Two weeks with two classes, with two or three goals agreed before day one.
The UNESCO guidance on generative AI in education (2023) asks schools to check tools for ethical and teaching suitability, and recommends data protection and an age limit of 13 for independent conversations with general AI platforms. Tools built for schools should be able to explain how they handle both.
How do schools use ELSA Schools for speaking practice?
ELSA Schools lets teachers turn their own textbook material into speaking practice, and gives every student instant feedback. Teachers can create Study Sets for key words and sentences, Read Aloud tasks for full passages, and AI Role-play conversations with ready-made scenarios for each English level. A teacher dashboard shows each student's progress over time.
Two school systems in Japan show what this can look like. In Kyotango City, Kyoto, the share of students who felt confident speaking English rose from 3.9% in the first year of junior high to 35.9% by the third. In Shibuya City, Tokyo, the share of students at all eight public junior high schools who said they were good at talking in English with classmates rose from 47% to 72% between April and October. Neither program had a comparison group, so these results show change over time, not proof of cause.
Where do AI speaking tools fall short?
AI tools can measure features of speech, but they can't yet judge everything a teacher hears. IELTS (2026) notes that automated scoring handles pauses and speech rate well and finds nuance, argument and implied meaning harder. Strong accents can also make speech harder for software to process.
There are practical limits too. Thirty students speaking at once is loud, so tasks need headsets or rotation. Students without a device at home need in-school time. And a conversation with an AI character is not the same as a conversation with a classmate. Keep live pair and group work in the lesson.
What should teachers ask an AI tool vendor?
Ask about speaking minutes, feedback, teacher visibility and data, in that order.
- How many minutes of speaking does a typical student do each week on your platform?
- Which speaking features does your feedback cover, and how do you describe its limits?
- Can teachers build tasks from their own textbook?
- What does the teacher see, and what can't the teacher see?
- Is the content suitable for our students' ages, and where is student data stored?
- What does a trial include, and how do we judge it?
Related reading
- How can AI speech analysis help students who struggle to speak in English?
- Blended English learning: how it builds speaking confidence
- How can schools give students enough IELTS speaking practice?
Recap
- Most AI tools for teachers save planning time. Only speech feedback and conversation tools add speaking practice.
- Research on AI conversation practice is promising but based on small studies.
- Choose by goal, textbook fit, feedback, teacher view and data safety, then run a short trial.
- Keep live speaking in the lesson for interaction a tool can't judge.
- Book a demo to see ELSA Schools with your own students.