How AI Evaluates Your Child's Reading

When your kid submits a reading summary on Wise Kid, an AI reads it and sends back a score within seconds. That probably raises a few questions.

What's it looking for? Is it fair? Will it crush my 7-year-old's spirit?

Good questions. Here's how it works.

What the AI actually measures

Every submission gets scored on 4 dimensions, each on a 1 to 5 scale:

• Specificity: Did they mention actual details from the book? Character names, plot points, specific events? Or was it "the book was good and stuff happened"?
• Structure: Is the summary organized? Does it have a beginning, middle, and end? Or is it a stream-of-consciousness run-on sentence?
• Insight: Did they go beyond just retelling? Any opinions, connections, or "I noticed that..." moments?
• Vocabulary: Are they using age-appropriate (or better) language? Not SAT words; just writing that shows growth.

These four combine into an overall score from 1 to 5.

It knows your kid's age

This matters a lot. A 6-year-old writing "the dog was funny and he ate the shoe" is doing great work. A 12-year-old writing the same thing probably phoned it in.

The AI adjusts its expectations based on the child's birth year and reading level. A 4 out of 5 for a 3rd grader means something different than a 4 out of 5 for a 7th grader.

Feedback that builds confidence

Every evaluation includes:

• Encouraging feedback: specific praise for what the kid did well. Not generic "great job!" but "You described the setting of the cave really vividly."
• Reading tips: 1 to 2 concrete suggestions for next time. "Try explaining why the character made that choice" or "What did this book remind you of?"
• Book recommendations: 2 to 3 books similar to what they just read, so they always have a "what's next."

We specifically tuned the AI to be warm and constructive. No child gets told their summary is bad. The worst score comes with the most helpful tips.

The Authenticity Shield

One thing parents ask: "What if my kid pastes in something from ChatGPT?"

We built a dual-AI detection system for that. The primary model checks 8 signals (vocabulary above grade level, perfect grammar, formulaic structure, lack of personal voice, etc.). If it flags something, a second AI from a completely different provider runs an independent check.

Both models have to agree before a submission gets flagged. One disagreement and the flag drops. We'd rather miss an AI-generated summary than wrongly accuse a kid who just happens to write well.

Vocabulary extraction

The AI also pulls out interesting words from each summary and adds them to your child's Word Bank. Over time, you can see their vocabulary expanding, which is one of the clearest indicators of reading growth.

Keep reading: Reading Comprehension: What Parents Get Wrong explains why summarization is so effective. And here's What a Reading Tracker Actually Does for Your Kid.

---

The whole system runs on Claude (by Anthropic), tuned with a custom prompt that's been refined across thousands of evaluations. It's not perfect, but it's consistent, fast, and way more encouraging than a tired parent trying to read a book report at 9 PM.