Readability

Readability: Method And Evidence

Readability: A Tutor That Listens To Reading

The description is direct: real-time reading help through an AI tutor that listens, corrects and encourages, with gains in fluency, accuracy and words correct per minute stated in under 60 days.

Listening is the entire proposition, and it addresses a genuinely hard logistical problem rather than an invented one.

Reading aloud to an attentive adult is one of the most reliably effective things a child can do, and every teacher knows it. It is also close to impossible to supply at the volume required. A class of thirty gives each child a few minutes a week, and at home the adult is usually cooking dinner.

The result is that the child who most needs to practise reading aloud — the one who is slow, hesitant and self-conscious about it — gets the least of it, because the practice is uncomfortable for everyone in the room.

An app that listens patiently, corrects without a sigh and never conveys impatience removes the social cost from the exercise. The child psychologist quoted on the site names this specifically: learning at their own pace without the pressure and expectations that produce self-doubt.

Reading aloud to an attentive adult is one of the most effective things a child can do. It is also almost impossible to supply at the volume required.

Levelled Books And Moving Up A Grade

The Smart Reading Program is described as automatically selecting the right reading material for each child, to minimise struggle and maximise improvement.

Level matching is quietly the most important feature in any reading programme, and getting it wrong ruins everything downstream.

There is a well-established range where practice works. Too easy and nothing is learned; too hard and the child spends all their attention decoding individual words with none left for meaning — and, worse, learns that reading is unpleasant. That association, once formed, is far harder to undo than any specific skill gap.

Automatic levelling that responds to performance is exactly right in principle, because a child's level moves and a parent has no reliable way to judge it.

The caveat is that a child's interests matter as much as their level. A dinosaur-obsessed seven-year-old will attempt a book two levels above them if it has dinosaurs in it, and motivation carries a reader further than difficulty matching does. A system that optimises purely on accuracy can select correctly and lose the child anyway. Broader tutoring tools such as Khanmigo address the school-subject side rather than this specific mechanic.

How Does An App Hear A Child Read?

Recognising a child reading aloud is meaningfully harder than ordinary speech recognition, and it is worth understanding why before judging any product in this category.

Children's voices sit outside the acoustic range most speech systems were built around, with higher pitch and different formant structure. Vocabularies are smaller, and pronunciation is inconsistent in ways adult speech is not.

Then there is the specific difficulty of the task. A struggling reader produces exactly the input that breaks recognition: long pauses mid-word, restarts, self-corrections, invented pronunciations of unfamiliar words, and a substituted word that fits the sentence perfectly.

That last case is the hard one. A child reading house for home has made an error worth catching. A system tuned too tightly interrupts constantly and destroys the child's confidence; tuned too loosely it misses the errors it exists to catch.

The app has one advantage no general recogniser has: it knows exactly what the text says. Constraining recognition to the expected words makes the problem tractable, and it is why reading tutors can work where open-ended child speech recognition struggles. What the site does not describe on these pages is what happens to the audio itself, which is the question we would want answered before a child uses it daily.

• Check the privacy policy for how children's voice audio is handled.
• Sit with your child for the first sessions to see how correction feels.
• Let interest override level when the two conflict.
• Treat the dashboard as a trend line, not a weekly verdict.
• Keep reading together — the app replaces the listener, not the relationship.

Words Correct Per Minute And What It Measures

Words correct per minute appears throughout the site, and it is a real measure rather than a proprietary score — which is worth knowing, because it means the number is comparable to something.

It comes from curriculum-based measurement, a decades-old approach where a standard passage is timed and errors subtracted, and it is used in schools precisely because it is quick, repeatable and sensitive to change.

Its value is as a screening and progress signal. A child well below the expected range for their grade is worth looking at more closely, and a child whose rate climbs over eight weeks is improving in a way that is hard to fake.

Its limits are equally well documented. Fluency is necessary for comprehension but nowhere near sufficient — some children read a passage quickly and accurately and cannot say what happened in it, and a rate measure alone will miss that entirely.

This is exactly why the comprehension component matters as much as the fluency tracking, and why the two being in one app is the sensible design. The claim we cannot assess is the evidence framing: the site refers to using Readability in a blind study without naming it or linking a result, and a stated 60-day improvement window carries no source. Neither is unusual in consumer education, and neither should be read as a published finding. Families educating at home will need a curriculum around it, which is what LittleLit AI supplies.

Comprehension Questions Asked Aloud

IVQA — interactive voice-based questions and answers — is presented as an industry first, prompting, correcting, assessing and encouraging young readers to ensure they understand the material.

The design decision worth noticing is that the questions are spoken and answered aloud rather than tapped.

That distinction matters more than it appears. A multiple-choice comprehension question can be answered correctly by elimination, by recognising a familiar phrase, or by guessing — which is why comprehension quizzes flatter weak readers so consistently.

Asking a child to say what happened requires them to reconstruct and articulate it, which is a much better test and a much better exercise. Retelling is one of the standard classroom techniques for exactly this reason.

It also keeps the mode consistent. A child who has been reading aloud stays in speaking mode rather than switching to tapping, which keeps the session continuous. The vocabulary mechanic follows the same logic: long-press a word to hear it pronounced and see the meaning, with the word saved to a personal list. Encountering a word in context and then reviewing it is how vocabulary is actually acquired — considerably better than word lists divorced from any sentence. Adaptive practice platforms like LearnFast AI apply the same spacing principle across other subjects.

Building A Reading Habit At Home

The app can supply the listener. It cannot supply the routine, and the routine is what produces the result.

This sequence gets a habit that survives past week three.

1. Read the privacy policy for how your child's voice recordings are handled.
2. Set up each child separately so their levels and dashboards stay their own.
3. Sit in for the first few sessions and watch how corrections land.
4. Fix a short daily slot rather than a long weekly one.
5. Let your child pick books on interest when it conflicts with level.
6. Check the trend monthly, and talk to their teacher if it stays flat.

Readability: What Parents Get

• A listener that never sighs — the child reads aloud and is corrected in the moment.
• Spoken comprehension questions rather than multiple choice, so answers must be reconstructed.
• Words correct per minute, a real school measure, tracked alongside accuracy and comprehension.
• Automatic level matching plus vocabulary saved from words the child looked up.
• Three readers per account, each with a separate dashboard, across phones, tablets and laptops.