ELSA Speak Review

Learners who have studied English for a decade can read a contract and still dread a phone call. That gap is not a knowledge problem, and no amount of grammar study closes it.

Speaking is the one skill you cannot practise alone effectively, because you cannot hear yourself accurately. Your brain supplies what you meant to say.

ELSA Speak exists to close that specific loop: you speak, it tells you which sounds were wrong, and you try again — the thing a teacher does and a textbook cannot.

ELSA Speak: Pronunciation Feedback Per Word

• Feedback on your own voice, instantly, on the specific sounds you got wrong rather than a general score.
• A bilingual tutor so beginners can start a conversation in their native language and move across gradually.
• AI role-play across real scenarios, plus personalisation by level, goal, preferred accent and industry.
• Publishes efficacy research — a dedicated page most language apps do not have.
• Individual, business, schools and API products, with dashboards for classroom and team progress.

Why Is Speaking Harder Than Grammar To Learn?

ELSA's positioning is narrow and correct: practise real-world English with instant, personalised feedback tuned to your voice.

The reason that narrowness works is a genuine asymmetry between the four language skills.

Reading and writing are self-correcting. You can check a sentence against a reference, notice you were wrong, and adjust — the feedback loop closes without another person in it. Listening is nearly as forgiving; you can replay something until it resolves.

Speaking is not. The mismatch between the sound you produced and the sound you intended is invisible from inside your own head, which is why learners repeat the same mispronunciation for years without noticing.

Worse, the errors that matter most are the ones your first language makes inaudible to you. If your native language does not distinguish two English sounds, you cannot hear the difference in your own speech — and only an external listener, human or machine, can point at it.

You can check your grammar against a book. You cannot hear your own accent, because your brain supplies what you meant to say.

Accent Choice And What It Signals

Among the personalisation options sits preferred accent, alongside level, goals, progress and industry focus.

Offering a choice of target accent is a small feature with a defensible position behind it.

The unhelpful framing of pronunciation is that there is one correct English and everything else is an error. That is both wrong and demoralising — the majority of English conversations now happen between two non-native speakers, and intelligibility rather than nativeness is what actually matters.

Letting a learner choose a target treats accent as a preference rather than a deficiency, which is the accurate framing. Someone working with an American team has a different target from someone sitting a British exam.

Industry focus does related work. The vocabulary and register a nurse needs differs from a software engineer's, and practising the words you will actually say beats practising a general corpus — the same argument that makes scenario-based practice in Praktika more useful than free conversation.

A Bilingual Tutor And Role-Play Scenarios

The bilingual tutor is described as letting beginners start conversations in their native language, making the transition to English less intimidating so mastery builds gradually.

This is the best idea on the page, and it addresses a genuine chicken-and-egg problem.

An English-only conversational app assumes you can already hold an English conversation. For a true beginner that is exactly the thing they cannot do, which means the whole category is quietly closed to the learners who need it most.

Starting in the learner's own language removes that barrier. The instruction, the explanation and the encouragement arrive in a language they understand, while the practice gradually shifts.

AI role-play across varied scenarios is the complement. Ordering food, taking a call, sitting an interview — these are performance situations, and rehearsing them without an audience is the point. The anxiety that stops people speaking is social, and an app removes the social risk while keeping the practice. Listening practice from real video is a different discipline again, and where Language Reactor is stronger.

• Practise aloud daily, briefly, rather than weekly at length.
• Choose a target accent deliberately rather than accepting the default.
• Set the industry focus so you rehearse the words you will actually use.
• Use role-play for the situations that make you anxious, not the easy ones.
• Read the efficacy page before believing any improvement claim.

Ninety Million Downloads And What They Mean

Two figures sit near the top of the page: 90 million-plus downloads, and 90 million-plus learners who have already joined.

Those are the same number describing two different things, and only one of them can be right.

A download is an install, and installs vastly exceed users in consumer apps — people download, open once, and never return. A learner implies use. Treating them as interchangeable inflates the more meaningful figure to match the less meaningful one.

This is worth flagging not as a criticism of the product but because it is the single most common pattern in consumer app marketing, and reading it correctly is a skill worth having.

The displayed 4.9-star rating carries the same caveat — no platform is named and no review count is given, so it cannot be compared against anything. What is more informative is the existence of a dedicated efficacy page, which is discussed below.

Publishing Efficacy Research

ELSA carries a page in its main navigation labelled Efficacy, and repeats it in the footer as Proven Learning Efficacy.

This is genuinely unusual and the most credible thing about the product.

Language learning apps almost universally market on engagement — streaks, minutes, lessons completed — because engagement is easy to measure and improvement is not. Whether anyone actually got better at English is a research question requiring instruments, control groups and time.

A company that builds a page around that question is inviting a kind of scrutiny most of its competitors avoid entirely, and the mere presence of the page is a signal even before you read what is on it.

The caveat is the one that applies to all vendor-published research: read who conducted it, whether it was peer-reviewed, what was measured, and against what comparison. A page called Proven Learning Efficacy is a claim; the study behind it is the evidence, and the two deserve separate assessment. None of this concerns producing speech, which is what a synthesis tool like Voicemaker does.

Practising English Until It Sticks

Speaking improves through frequency rather than duration, which is the opposite of how most people study.

This sequence uses the app for the thing it is uniquely good at.

1. Set your level, goal, target accent and industry before the first lesson.
2. Practise aloud for ten minutes daily rather than an hour weekly.
3. Repeat the specific sounds it flags rather than moving on for variety.
4. Use the bilingual tutor if English-only conversation is still intimidating.
5. Role-play the situation you dread, not the one you find comfortable.
6. Re-record the same passage monthly and listen back to hear the change.