How can AI speech analysis help students who struggle to speak in English?

AI speech analysis scores a student's recorded English on skills such as pronunciation and fluency, and shows which words to work on. In schools, it gives every student speaking feedback between lessons, when one teacher can't listen to everyone.

It works like a music practice room with a tuner. A student plays, checks the reading, adjusts and plays again, and the teacher still decides what to teach next.

Illustrative scenario. A Grade 8 student records one minute about her weekend. Fluency is her lowest skill, so she taps a word where she paused and reads the tip. She records the same topic again, this time focusing on linking those words. Her teacher hasn't heard either attempt yet. The next lesson starts from the report she handed in.

This guide is for schools where English is taught as a foreign language. It covers why students struggle to speak English at school, what AI speech analysis measures, how students and teachers use it, and where it falls short.

What is AI speech analysis?

AI speech analysis turns a recording into text, then scores how the student spoke on skills such as pronunciation, fluency, vocabulary and grammar.

"Speech analysis" also means other things, which is why searches for it return mixed results. The table separates the terms.

Why do students struggle to speak English at school?

Where school is taught in another language, students often get one or two English lessons a week, little speaking time in them, and little feedback.

English is a subject, not the language of school

In Vietnam, Indonesia or Brazil, students learn maths, science and history in their own language. English is one subject among many, so most of their English contact happens in a few lessons a week.

Speaking shows the gap most clearly. The EF English Proficiency Index 2025 scores adults on reading, listening, writing and speaking. In all three countries, the speaking score sits well below the reading score.

The index tests adults who chose to take EF's free online test, so it isn't a survey of school students. The global average score in the 2025 edition is 488.

Governments are adding English lessons and teachers

Each of these countries has a plan to raise English levels in schools.

• Vietnam. Decision 2371/QD-TTg (27 October 2025) sets out a programme to make English the second language in schools from 2025 to 2035. It requires all schools to teach English as a compulsory subject from grade 1. The Ministry of Education and Training estimates it needs about 22,000 more English teachers for preschool and primary, and plans to train at least 200,000 teachers to teach in English.
• Indonesia. English becomes a compulsory subject in elementary schools from the 2027–2028 school year. Minister Abdul Mu'ti said the focus is confidence in using English to communicate, and that teachers shouldn't dwell on students' grammar mistakes when they speak.
• Brazil. The national curriculum (BNCC) made English compulsory from grade 6 in all public and private schools. A 2023 Pearson and Inter-American Dialogue study of two states, Mato Grosso and Minas Gerais, found students get two English classes a week in lower secondary and one in upper secondary.

These plans add lessons and teachers. Classes stay large, though, so each student's speaking time inside a lesson stays short.

There isn't enough speaking time

Take a school with two 45-minute English lessons a week and 40 students per class. Suppose half of each lesson goes to explanation, instructions and whole-class work. The remaining 22 minutes, shared 40 ways, is about 33 seconds of speaking per student per lesson.

Over two lessons a week and a 36-week school year, that comes to about 40 minutes of speaking in a whole year. These are assumptions, not measurements. Pair work raises the total, but the teacher still can't hear most of it.

Feedback arrives late, or not at all

When a teacher can listen to only a few students per lesson, most spoken mistakes go unheard. A student can repeat the same mistake for weeks before anyone points it out.

Speaking in front of others is stressful

Language researchers have linked classroom speaking and anxiety for decades. Horwitz, Horwitz and Cope (1986) described foreign language classroom anxiety as three parts: worry about communicating, test anxiety and fear of being judged. Their questionnaire includes items such as feeling self-conscious speaking in front of other students.

The gap shows up in school data too. In one ELSA Schools rollout in Kyoto Prefecture, Japan, 3.9% of Grade 7 students said they felt confident speaking English. By Grade 9, the share was 35.9% (ELSA Data).

What does AI speech analysis measure?

It measures five skills where students often have trouble: pronunciation, intonation, fluency, vocabulary and grammar.

The tool scores each skill separately, so one overall number doesn't hide where the problem is. In ELSA Schools' Speech Analyzer, each skill also comes with a short tip.

The middle column shows common examples, not data. The tips are shortened from the app, and the exact tip depends on what the student says.

How does AI speech analysis help students speak?

It lets each student practise without an audience and see exactly what to fix before trying again.

It answers each of the three struggles above.

• More speaking turns. A student can record as often as they like, in class or at home, so practice no longer waits for the teacher's attention.
• Faster, more specific feedback. Results arrive once the student submits, and they show where each issue happened. A review of education research (Hattie and Timperley, 2007) found that feedback is among the major influences on learning, and that its effect depends on the type of feedback and how it's given.
• Less pressure. Recording alone removes the audience. ELSA Schools has no leaderboards or student rankings, so no one compares scores.

The tool also gives students something to aim for. ELSA Schools' Rephrase tab shows a more natural version of what the student said, with every change marked.

How does a student use AI speech analysis?

A student records a short talk, reads the scores and word-level feedback, then records again with one thing to fix.

1. Record. Speak for a minute or two on a set topic, in a quiet place.
2. Review. Check the overall result, then the weakest skill, then the words flagged in the transcript.
3. Try again. Pick one change and record the same topic.

In ELSA Schools: Speech Analyzer sits inside the ELSA Schools Student App, which runs in a web browser. Students keep their sign-in and language setting, and there's nothing to install. It became available there on 24 September 2026.

• Students tap Speech Analyzer in the bottom navigation, then New recording.
• The results open in seven tabs: Overall, one per skill, and Rephrase.
• The overall result is an indicative CEFR level, such as B2, shown with a percentage. The Council of Europe's CEFR describes progress on six levels, A1 to C2.
• Tapping a word in the transcript shows feedback on that word for the selected skill.
• Download Results gives a PDF report. It adds estimated speaking scores for IELTS, TOEFL, PTE and TOEIC, which are approximate and not official.
• Each recording stays in a personal library with its score and date.

The ELSA Schools version covers free-form recording. Exam practice such as IELTS and EIKEN stays on the classic Speech Analyzer site.

How is it different from other speaking practice?

AI speech analysis scores what a student chooses to say, while read-aloud tasks and conversation partners score prompted speech.

The formats answer different questions, so a school can use more than one. The table compares how each is built, not which is better.

How can teachers use AI speech analysis between lessons?

Teachers can set short recordings that they couldn't hear in class, then read the results to decide who needs a conversation.

Three routines fit a normal timetable. They're teaching suggestions, not product features.

• Warm-up. Students record a short answer about the day's topic while the teacher walks the room.
• Homework. Students record a talk after a unit, download the PDF and hand it in. The teacher reads the score breakdown and picks who to talk to.
• Then and now. Students record the same kind of task at the start and end of a unit and compare. Each recording in the library shows its score and date.

To see whether the practice is working, compare each student's recordings across a term. Look at which skill stays lowest, not only the overall result. A skill that stays low after several attempts is the one worth class time.

Where does it fall short, and what should schools avoid?

It scores the language a student uses, not the quality of their ideas, and its results guide practice rather than replace a grade.

Limits to know

• Teacher judgement. Tone, content and context still need a teacher.
• Estimates. The IELTS, TOEFL, PTE and TOEIC scores in the PDF are approximate guides, not official results.
• Transcription errors. Speech tools can mishear words, especially with background noise or a weak microphone, and that can affect the feedback.

Common mistakes

1. Treating the score as a grade. It's a practice guide. Grading a free-speech recording on its score alone puts more weight on it than it was built to carry.
2. Recording in a noisy room. Background noise causes misheard words and skews the feedback. A quiet corner and a headset help.
3. Reading only the overall result. The useful detail is in the skill tabs and the word-level transcript.
4. Using it once. A single recording shows a starting point. Progress shows up when a student records the same kind of task again.
5. Skipping the retry. Feedback helps most when the student acts on it straight away, with one change on the next attempt.

What should schools ask an English training vendor?

Ask what the tool scores, how results reach teachers, how it handles student data and how it can be tried.

1. What does it score, and on which scale? Look for named skills and a recognised scale such as CEFR.
2. Can a student see where each issue happened?
3. How do results reach the teacher?
4. How is student data handled? Ask who owns the data, how AI is used and what consent is needed. In ELSA Schools, no student personal information is sent to AI models. Ask for the data processing agreement in writing.
5. Can a few classes try it first? ELSA Schools offers a 14-day trial with 5 to 10 accounts after a demo. Speech Analyzer is included in the licence, which has a minimum of 50 student licences on a 12-month term.

Related reading

• ELSA Schools

Frequently asked questions

Can AI analyze a student's spoken English?

Yes. AI speech analysis turns a recording into text, then scores how the student spoke on skills such as pronunciation, intonation, fluency, vocabulary and grammar. It shows where each issue happened in the transcript, so the student knows which words to work on before recording again.

How accurately does AI recognise what a student says?

It depends on the recording. Speech tools can mishear words, especially with background noise or a weak microphone, and a misheard word can affect the feedback. A quiet room and a headset help. Teachers should treat results as a practice guide, not a final grade.

How can a student check whether their spoken English is good?

A student can record a short talk with an AI speech analyzer and read the result for each skill. In ELSA Schools, the overall result is an indicative CEFR level with a percentage. Comparing recordings across a unit shows which skill improves and which stays lowest.

Why is speaking often the weakest English skill?

Where school is taught in another language, students often get only one or two English lessons a week, and large classes leave little speaking time in each. In the EF English Proficiency Index 2025, speaking scores trailed reading scores by 44 to 61 points in Vietnam, Indonesia and Brazil. The index tests adults who chose to take EF's online test, not school students.

Can AI speech analysis replace a speaking teacher?

No. It scores the language a student uses and gives a first round of feedback between lessons. A teacher still judges tone, content and context, and decides who needs a conversation. In ELSA Schools, results reach the teacher as a PDF the student shares.

Next step

• Students struggle to speak because they get few turns, little feedback and an audience.
• AI speech analysis scores recorded speech on five skills and shows where each issue happened.
• Used between lessons, it gives every student a first round of feedback, and the teacher decides who needs a conversation.

To see Speech Analyzer with your own students, book a demo to unlock a free ELSA Schools trial.