For parents
How ReadHatch actually works
A short guide to what's under the hood — how questions are generated, what makes a comprehension quiz good, and how to read the results as a parent.
Written by Mark Clayton, founder of ReadHatch.
The problem this app solves
Your kid closes a book. You ask "how was it?" They say "good" and disappear. You have no idea whether they read it, skimmed it, or invented a plot on the walk from the couch to the kitchen.
School reading programs solved this with quizzes. Accelerated Reader has been around since the 1980s for exactly this reason. But if your kid isn't in a school that runs AR — homeschool, summer, private school without the license, or a family that reads outside the school list — you're back to "how was it?" and "good."
ReadHatch is what fills that gap. A short, real quiz on the specific book, generated in seconds, calibrated to your kid's age. Pass it and you have proof they read it. Fail it and you have a specific signal about what happened.
The four principles
What we designed for, in order.
Comprehension, not recall
A quiz that only asks 'what color was the dog?' proves the kid skimmed the first chapter. Good comprehension questions ask about inference (why did the character do that?), theme (what is this book really about?), vocabulary in context, and specific plot details that only appear if you finished the book. ReadHatch mixes all four.
Age-scaled, not one-size-fits-all
The same book asked of a 7-year-old and a 12-year-old should not produce the same quiz. Sentence structure, vocabulary, and abstraction all shift with age. ReadHatch quizzes are generated with the reader's age as an input, so the same book yields age-appropriate questions.
70% is a passing score for a reason
Comprehension isn't binary. A kid who gets 7 out of 10 read the book and understood most of it. A kid who gets 4 out of 10 didn't. 70% draws the line where 'passed' actually means something without demanding perfection on a set of questions they've never seen.
No account for the child
Kids under 13 are covered by COPPA, and the safest path is not to collect their data at all. In ReadHatch, the parent signs in. The child has a reader profile with a first name and an age — nothing else. No login, no email, no third-party marketing.
How the quizzes are made
Each ReadHatch quiz is generated by AI on demand. The inputs are:
- The book.Title, author, and any additional metadata pulled from Open Library's ~30 million-title catalog.
- The reader's age. This controls vocabulary, sentence structure, and the level of inference expected.
- A question-type mix. Ten questions across plot details, inference, theme, vocabulary in context, and a closer.
The model is instructed to produce questions where the wrong answers are plausible enough that a kid who didn't read the book would guess wrong, but obvious to a kid who did.
Questions parents ask
What does a good ReadHatch question look like?
A mix of question types across the 10 questions on any given book. Roughly: three questions on specific plot details, two on inference (character motivation, cause and effect), two on theme or main idea, two on vocabulary in context, and one on the ending. Questions are multiple choice with one correct answer and three distractors that are plausible but wrong for someone who read the book.
How are quizzes generated?
ReadHatch identifies the book (title, author, and available metadata via Open Library), then uses AI to generate the 10-question quiz on demand. Difficulty is scaled to the reader's age. Each quiz is fresh — retakes generate a new set of questions.
What if my kid fails a quiz?
Retakes are always allowed, and a new set of questions is generated for each attempt so it isn't a memorization game. Points are only awarded on the first passing attempt, so there's a real motivation to read carefully the first time. A failed quiz is a signal — the kid either skimmed, read it a while ago and forgot, or the book was above their reading level. All three are useful information.
How should I use the quiz results as a parent?
Treat the quiz score like a temperature check, not a grade. A 7 or 8 means 'they read it.' A 5 or below on a book at their level usually means they skimmed. The dashboard tracks scores across books so you can see whether the kid is consistently passing at a given reading level — that's a better read on progress than any single quiz.
Why 10 questions, not 5 or 20?
Five is too few to distinguish 'passed' from 'lucky.' Twenty is too long to hold a kid's attention right after they finished a book — comprehension questions right after reading are far more accurate than the same questions an hour later, so the quiz has to be short enough that the kid takes it now. Ten hits the balance and lets 70% be a meaningful threshold.
Is AI reliable enough to generate these quizzes?
For well-known books, yes — the model has seen thousands of pages of context about the book. For obscure books, the metadata via Open Library still gives it enough to produce a fair-difficulty quiz on the story shape (protagonist, setting, arc). If you ever see a question that seems wrong for a book you know, email us — that's exactly the feedback that improves the system.
Try it with one book
Pick a book your kid just finished, take one quiz for free, see whether the questions feel fair. That's the whole first-visit experience.