Zuki ReadMy child reads well but doesn't understand what they read: how to detect it and improve reading comprehension at home
Team at Zuki ReadWritten with AI assistance20 min readIn summary
- Reading aloud fluently and understanding what is read are two distinct skills: the classic formula of the Simple View of Reading (Gough and Tunmer, 1986) separates them into decoding and language comprehension. A child can have the first at full capacity and the second weak.
- There is a 10-minute home test to determine which of the two is the problem: close the book and ask four questions of four different types (literal, vocabulary, inferential, and opinion). The one that fails tells you where to work.
- In Spain, PIRLS 2021 (IEA, published in May 2023) placed 4th-grade students at 521 points in reading comprehension, 7 below the 528 of 2016. This is not a problem of an isolated child.
- What moves the needle at home is not reading more pages: it is remembering without the text in front (active recall), discussing what has been read, and rereading with a specific question in mind.
- Zuki Read, the reading comprehension app for children aged 6 to 12 from AIsToHire, automates just that part: it writes a story tailored to the child's age about what they want and asks them 10 questions of those same four types, corrected on the server, with retries and hints.
If you have arrived here by searching Google for something like "my child reads well but doesn’t understand what they read", the first thing: you are not describing a rare child. You are describing a pattern that has had a name in educational research for forty years and a fairly concrete way of addressing it at home. This article explains how to distinguish the real problem in ten minutes, what the five most common causes are, and seven exercises you can do with any book. In the end, you will see how Zuki Read, the mini app for children's reading comprehension from AIsToHire, turns that routine into something the child wants to do — but the seven exercises work just as well with a library book and a notebook.
A warning before we begin, because it matters: none of what follows is a diagnosis. If you suspect dyslexia, a language disorder, or an attention problem, that should be evaluated by a school counselor, a speech therapist, or an educational psychologist. Neither this article nor Zuki Read diagnoses anything.
Reading aloud and understanding what is read are two distinct skills
They are two separate things and can be had separately. The Simple View of Reading (Gough and Tunmer, 1986) formulates it as a product: reading comprehension = decoding × language comprehension. Since it is a multiplication, it is enough for one of the two factors to be low for the result to sink, even if the other is perfect. A child who decodes wonderfully and understands little of the language reads beautifully and does not grasp it.
That is why the "reads well" you hear at home is misleading: what you are evaluating when you hear them read is the decoding —that they convert letters into sounds with speed and accuracy— and that says nothing about whether they are constructing meaning.
| Decoding | Language Comprehension | What you see at home |
|---|---|---|
| Good | Good | Reads fluently and tells you the whole story. Everything in order. |
| Good | Weak | Reads beautifully but doesn’t know what it was about. This is the case of this article. |
| Weak | Good | Stumbles, syllabicates, gets tired — but if you read it to them, they understand everything. |
| Weak | Weak | General difficulty: the first step is to rule out causes with a professional. |
The third row is important: if your child understands everything when YOU read the same text aloud to them, the bottleneck is not comprehension, it is fluency. They are two distinct plans of action and it is advisable not to confuse them.
How many children read well and do not understand what they read
There is no exact figure for "readers who decode well and comprehend poorly" in Spain, but there are two serious references on the general state of reading comprehension, and an academic estimate of the specific profile. Studies on poor comprehenders place their prevalence at approximately between 3% and 10% of schoolchildren, with such a wide range precisely because it depends a lot on where the cutoff point is set (review by Spencer, Quinn, and Wagner, 2014).
| Source | Publication Date | Data |
|---|---|---|
| PIRLS 2021 (IEA), 4th grade | May 2023 | Spain: 521 points in reading comprehension, compared to 528 in the 2016 edition |
| PISA 2022 (OECD), 15 years | December 2023 | Spain: 474 points in reading; OECD average: 476 |
| Review on poor comprehenders (Spencer, Quinn, and Wagner) | 2014 | Estimated prevalence of the profile "decodes well, comprehends poorly": ~3-10% depending on the criteria |
Translated: in a classroom of 25 children, it is reasonable to have one or two with this exact profile, and quite a few more with comprehension below what their reading aloud suggests.

The four question test: ten minutes, a book, and nothing more
Take a text that your child has just read alone —two or three pages are enough—, close it and ask them four questions, one of each type. The type that fails tells you what needs to be trained. Do it with the book closed: with the text in front, many questions are answered by copying, and then you are not measuring comprehension, you are measuring aim.
| Type of question | What it measures | Example (about a pirate story) | If this fails |
|---|---|---|---|
| Literal | If they have retained what is written exactly | What was the name of the ship? | They are not paying attention to the text or reading too quickly |
| Vocabulary | If they deduce the meaning of a new word from the context | In the story, it says that the crew was "exhausted." What does that mean? | The vocabulary of the text is too advanced for them |
| Inferential | If they deduce what is not written but the text implies | Why do you think the captain did not want to open the chest in front of everyone? | Here is the most typical failure of the "reads and does not understand" profile |
| Opinion | If they take a stance and justify it with the text | Do you think what the cabin boy did was right? Why? | They understand but do not connect with the story, or they lack confidence |
One detail that changes the outcome: leave five seconds of silence after asking. Most adults wait less than a second before rephrasing or giving the answer, and with that, we rob the child of just the moment when they were retrieving the information.

The five most frequent causes of reading well but not understanding
In the vast majority of cases of children without diagnosed difficulties, the problem falls into one of these five boxes — and often in two at once. Identifying which one saves months of "reading more" without results.
| Cause | Telltale sign | What to do first |
|---|---|---|
| 1. Insufficient vocabulary for that text | Fails vocabulary questions; skips words without asking | Lower the level of the text and work on 3-5 new words per session, in context |
| 2. Fluency not yet automatic | Reads well but shows visible effort; gets tired on the second page | Shorter texts and re-reading the same passage, not new texts |
| 3. Does not infer | Nails the literal questions but blanks on the whys | Explicit inferential questions and think aloud in front of them |
| 4. Lacks prior knowledge of the topic | Understands stories but gets lost in science or history texts | Two minutes of conversation about the topic before reading |
| 5. The text is above their level | Fails everything, including literal, and gets demoralized | Lower the level without drama: frustration stifles reading |
About cause 1, there is a useful threshold: research on lexical coverage (Hu and Nation, 2000; Nation, 2006) estimates that one needs to know about 95-98% of the words in a text to understand it without help. These studies were done with second language readers, so the number is indicative, but the idea holds: if there are five words on a page they do not know, the text is not "a little difficult", it is inaccessible.
And about cause 2, a reference for reading speed, with two warnings up front: these are U.S. norms in English (Hasbrouck and Tindal, 2017) and are indicative, not a diagnosis. They serve to position oneself, not to label anyone.
| Grade (approximate equivalence) | Words per minute, 50th percentile, end of year |
|---|---|
| 1st grade | ~60 |
| 2nd grade | ~100 |
| 3rd grade | ~112 |
| 4th grade | ~133 |
| 5th-6th grade | ~146 |
If your child is well below the band for their grade, the priority work is fluency: comprehension does not have surplus resources while decoding continues to require conscious effort.
Seven reading comprehension exercises to do at home
These seven work with any book and with sessions of 15 minutes. They are ordered by what yields the most results in the least time, and the first three are, by far, the ones that move the needle the most.
- Close the book and tell it (active recall). Two minutes. The child summarizes without the text in front of them. It is the most powerful exercise on the list: Roediger and Karpicke (Psychological Science, 2006) demonstrated that retrieving what was read from memory produces significantly better retention after a week than re-reading the same text. Re-reading feels productive but is not as effective; recalling is hard but works.
- Ask for the "why", not the "what". Three minutes. For every literal question, ask two inferential ones: why did they do that?, how do you think they felt?, what will happen now?. This is the muscle that this type of reader has without training.
- Think out loud in front of them. Two minutes. You read a paragraph and verbalize what you are doing inside: "wait, it doesn’t make sense that the captain is angry... oh, right, because they said that before...". You are teaching an invisible process by making it visible. It is one of the recommended practices in the Institute of Education Sciences guide from the U.S. on reading comprehension in the early grades (IES, 2010).
- Teach them to ask questions themselves. Three minutes. Let the child ask you three questions about what they read. To formulate a good question, one must have understood, so the exercise self-corrects. It is one of the four steps of Palincsar and Brown's reciprocal teaching (1984): summarize, question, clarify, and predict.
- Hunt for words in context. Two minutes. Three new words per session, not one more. First, let them guess from the context, then you check. Looking up fifteen words in the dictionary kills the reading; three enrich it.
- Re-read with a mission. Three minutes. Not "read it again," which is the most boring, but "read it again and tell me exactly where it says they were scared". Re-reading with a specific goal is active searching, not passive repetition.
- Let them choose the topic. As long as it takes. A child who reads about dinosaurs, soccer, or whatever obsesses them has plenty of prior knowledge (cause 4 disappears) and can handle texts that are more difficult than their theoretical level. Interest is the cheapest lever there is.

The 15-minute routine, spread throughout the week
Fifteen minutes four days a week yield more than an hour on Sunday, because comprehension is a skill and skills are consolidated with distribution, not bingeing. This is a distribution that works without turning reading into homework.
| Day | Minutes | What is done | Exercise |
|---|---|---|---|
| Monday | 15 | Read a short text they choose and recount it with the book closed | 1 and 7 |
| Tuesday | 15 | Same text: three why questions and re-reading with a mission | 2 and 6 |
| Wednesday | — | Break | — |
| Thursday | 15 | New text; this time they ask you the questions | 4 and 5 |
| Friday | 15 | You read a paragraph thinking out loud and they follow | 3 |
| Weekend | free | Reading for pleasure, without any questions | — |
The "weekend" column is not filler. If every reading session ends in an exam, reading becomes a chore and the medium-term effect is the opposite of what you want.
What Zuki Read does exactly, and what it does not do
Zuki Read is a mini web app from AIsToHire that does the following: the child writes an idea —"a dragon who wants to be a chef"— or presses Surprise me!, and the AI writes a new story tailored to their age and level, with 10 comprehension questions of the same four types from the test above. They can read it or listen to it, respond with retries and hints, and earn points and medals based on how many they get right on the first try. It is available in Spanish, English, French, Italian, Portuguese, and German.
What it adjusts for age is not a label: they are concrete parameters of the generator.
| Age band | Words in the story (easy / medium / difficult) | Distribution of the 10 questions |
|---|---|---|
| 6-7 years | 260 / 340 / 430 | 7 literal (1 vocabulary) + 3 simple inferential |
| 8-9 years | 420 / 560 / 700 | 4 literal (1 vocabulary) + 4 inferential + 2 opinion |
| 10-12 years | 620 / 800 / 1000 | 4 literal (1-2 vocabulary) + 4 inferential + 2 opinion |
Note that the distribution changes with age: at 6 years almost everything is literal, and from 8 years half of the questions already require inferring or giving an opinion. Additionally, the manual level only moves within the age band: a 6-year-old in "difficult" never receives a text from the 10-12 band.
And this is what it does not do, stated clearly:
| What it does do | What it does not do |
|---|---|
| Generates stories and questions tailored to age and level | Does not diagnose dyslexia, ADHD, or any disorder |
| Corrects on the server and saves correct answers on the first try | Does not replace school, speech therapy, or guidance counselors |
| Shows real comprehension metrics in the adult zone | Does not send reports by email: they must be checked in |
| Allows re-reading of the last stories offline | The test cannot be played offline (the correct answer lives only on the server) |
| Records speed and accuracy if the child reads aloud | Does not save any audio of the child, only the two numbers |
| Allows up to 8 profiles with independent statistics | Is not a classroom tool nor generates reports for the teacher |
How Zuki Read fits with the seven exercises
Five of the seven exercises are within the app; two remain yours, and it's good to know which ones so you don't delegate what cannot be delegated.
| Exercise | In Zuki Read |
|---|---|
| 1. Active recall | The 10 questions come up with the story already read; the child can return to the text, but the first answer is given from memory (and only correct answers on the first try score the maximum) |
| 2. Asking why | The generator itself reserves 4 of the 10 questions for inferences from the age of 8 |
| 3. Thinking out loud | It remains yours. No app can replace an adult verbalizing how they reason |
| 4. Let him ask | It remains yours, although the app has an optional open question at the end where he shares in his words what he thought |
| 5. Words in context | There are 1-2 vocabulary questions per story, always about a word from the text and always in its context |
| 6. Reread with a mission | After the third mistake on a question, the app suggests going back to the story without losing what has already been answered correctly |
| 7. Let him choose the topic | It is the starting point: the story is written based on the idea he writes |
The details of the correction deserve a line: when he makes a mistake, the three options are reshuffled and a hint appears (which can be turned off in settings); the explanation of why that was the correct answer is not shown until the end, in the story review. There is no timer or penalty for taking time — the question only advances when he answers correctly.
The points reflect that same criterion:
| Correct on the attempt | Points |
|---|---|
| 1st | 10 |
| 2nd | 7 |
| 3rd | 4 |
| 4th or later | 2 |
A story with 10 correct answers on the first try also adds 20 bonus points. The story medal is also calculated only with correct answers on the first attempt: 9 or more give gold, 7 or 8 silver, and 5 or 6 bronze. Repeating an already completed story adds half the points and does not give bonuses or affect streaks, so that repeating is not a shortcut.
What to look at in the adult zone after a few stories
The adult zone is behind a four-digit PIN and shows metrics derived from actual stories, not estimates. Three of them truly answer the question you started with.
- First try accuracy (%): the percentage of correct answers on the first attempt out of the total. It is the honest figure; the percentage "with retries" always looks nice and doesn't mean much.
- Breakdown by type of question: how many have been asked and how many he has answered correctly on the first try in literal, inferential, vocabulary, and opinion questions, separately. If the literal questions are at 90% and the inferential ones at 40%, you've just pinpointed the exact problem this article is about.
- Read vs. listened: how many stories he has read and how many he has listened to with narration. A child who gets many correct answers but listens to almost everything is training oral comprehension, which is valuable, but it's not the same.
You will also see medals, average time per story, the evolution of the last ten, and, if he has used reading aloud, an average of words per minute and accuracy. About that last one: the app calculates it with the browser's voice recognition, it is indicative and does not save audio; it serves to see a trend over the weeks, not to compare it with any table.
To start with the level set correctly, the app offers a short calibration story of 5 questions: depending on how many he gets right on the first try, it sets the initial level of the profile. From there, if he chains three stories with a gold medal, it suggests leveling up, and if two of the last three remain at bronze or without a medal, it suggests leveling down. It always suggests: it never changes it on its own, and it can be turned off.
When this stops being a home matter
There are signs that call for assessing the situation with a professional instead of continuing with home exercises. If you see several of these at once, talk to the school tutor or the counselor, and do not use an app as a substitute:
- Marked and sustained refusal to read, or crying and distress associated with reading.
- Reading clearly below that of his classmates after an entire course of work.
- Persistent confusion of letters, omissions, or reversals beyond the age of 7-8.
- Difficulty also when you read the text to him: if he does not understand what he hears, there may be something in oral language behind it.
- Attention difficulties that appear in all tasks, not just when reading.
Zuki Read does not evaluate any of this. It is a tool for practice and motivation; the assessment is for a professional.
Frequently Asked Questions
Is it normal for my child to read perfectly and then not be able to tell me anything?
It is common, and there is an explanation: reading aloud measures decoding, and recounting what has been read measures comprehension. These are distinct skills (Gough and Tunmer, 1986) and can progress at different rates. What is not a good idea is to let it go assuming that "it will come to him eventually," because the gap tends to widen: those who comprehend better read more, and those who read more comprehend better.
At what age should I be concerned?
Up to 7 or 8 years old, it is expected that a good part of the child's effort goes into decoding, and comprehension may be lacking because of that. Starting from 3rd grade, when reading should be reasonably automatic, if he still cannot summarize what he has just read, it is worth working on it explicitly and discussing it with the tutor.
How much time should I dedicate each day to see some progress?
Fifteen minutes, four days a week, is a realistic and sustainable dose. What matters is not the total number of minutes but that the session includes recalling from memory what has been read: Roediger and Karpicke (2006) found that retrieval clearly surpasses rereading in retention after a week. Half an hour of rereading yields less than ten minutes of recounting with the book closed.
Does reading more books solve the problem by itself?
Usually not, if the child continues to read in the same way. Adding pages without changing anything else reinforces the superficial reading they already had. What changes the outcome is what happens after reading: summarizing without the text, answering inferential questions, and rereading with a specific goal.
Does Zuki Read work if my child has dyslexia?
It can serve as practice, but it is important to be precise: the app does not diagnose or treat dyslexia, and any plan should be conducted by a professional. What it does offer are accessibility supports that usually help: three font sizes (16, 20, and 24), high contrast mode, an easy-to-read font with the spacing between letters and words and line spacing recommended by the British Dyslexia Association, and narration of the story aloud at three speeds with phrase highlighting.
Are the stories written by AI safe for a 6-year-old?
The generator operates with child safety rules above any idea the child writes: no violence, weapons, death, or real danger, no terrifying antagonists, no romantic content, no political or religious references, and no brands. If the idea proposed by the child is not suitable, it is not rejected: it is redirected while preserving what excites him (a "dragon war" becomes a "great dragon contest"). Additionally, the generated text undergoes a subsequent validation before reaching the child.
Can my two children use the same account?
Yes: it allows up to 8 profiles per account, each with its own age, level, points, medals, and separate statistics. There is a family podium and a shared library, with one rule: a child only sees stories from age bands equal to or lower than theirs, so they do not encounter texts that are too advanced for them.
How do I know if he is truly understanding or just guessing through trial and error?
By looking at the first attempt accuracy in the adult area, which is the metric that counts only the correct answers on the first try. That is the figure that cannot be inflated by trying options. And if you compare it by question type, you will see if the problem is specifically in the inferential questions, which is common in this profile.
Do I need to be present while he is using it?
It is not essential for exercises 1, 2, 5, 6, and 7, which are within the app. It is essential for exercises 3 and 4 —thinking aloud in front of him and having him formulate questions— which cannot be replaced by any tool. The reasonable combination is: the app four days, and a five-minute conversation about a story he liked, with you, once or twice a week.
Zuki Read, in one sentence
Zuki Read is the mini app from AIsToHire that writes personalized stories with AI for children aged 6 to 12 and accompanies them with 10 comprehension questions —literal, vocabulary, inferential, and opinion— with retries, hints, medals, and a protected adult area with a PIN that shows the real first attempt accuracy, broken down by question type. It is available in six languages on the Zuki Read page, the reading comprehension app for children from AIsToHire.
It does not diagnose or replace a professional: what it does is make the seven exercises from this article something your child wants to do, on whatever topic he likes, four days a week and in fifteen minutes.
Try Zuki Read for free: let your child choose what their next story is about and discover, with the 10 comprehension questions, what they are really understanding.
Try Zuki Read for free: let your child choose what their next story is about and discover, with the 10 comprehension questions, what they are really understanding.Keep reading
View all articles →
My child reads well but doesn't understand what they read: how to detect it and train reading comprehension in primary school

Discussing via WhatsApp or talking in person: what science says about couple conflicts through messages

Mind map, concept map, and outline: differences and when to use each one
