A study of 67 children shows the answer: AI builds understanding, parents build hearts
Introduction: One Picture Book, but Who Reads It?
The idea that ChatGPT is changing the world is almost cliché by now. In an age where AI writes code, drafts reports, and even creates artwork, can it also take over reading picture books to children?
This isn’t idle curiosity. In English as a Foreign Language (EFL) environments, parents often lack confidence in their own English proficiency. They want to read English picture books to their children, but worry about pronunciation, and feel at a loss for how to ask questions in English. The thought “Wouldn’t AI do a better job?” is only natural.
There’s a method called Dialogic Reading. Rather than simply reading a book aloud, the reader asks the child questions, responds to their answers, and co-constructs the story together. Decades of research have validated its effectiveness, but the catch is that it requires the reader to ask the right questions and adapt flexibly to the child’s responses.
Could a Large Language Model (LLM)-based AI fill this role? A research team from Penn State and China designed an experiment to find out.
The Core Research Question
The team asked:
“How do LLM-based AI-guided dialogic reading and parent-led traditional reading compare in terms of children’s learning outcomes and engagement?”
They measured three specific dimensions:
- Reading Comprehension — How well did children grasp the story content?
- Vocabulary Acquisition and Story Retelling — Did they learn new words and reconstruct the narrative?
- Engagement — How deeply were children involved in the reading process?
The Experiment: An E-Book with an AI Reading Partner
What Is Dialogic Reading?
Dialogic reading rests on two core strategies:
- PEER: Prompt → Evaluate → Expand → Repeat. A sequence designed to transform the child from passive listener into active story participant.
- CROWD: Completion prompts, Recall questions, Open-ended questions, Wh-questions, and Distancing prompts. Five question types designed to elicit language production and deeper thinking.
How Did the AI Implement These Strategies?
The team developed an e-book app featuring a bilingual (Chinese-English) conversational agent (CA) powered by an LLM. This AI:
- Asked CROWD-type questions at each scene in the story
- Provided personalized feedback and scaffolding (hints and support calibrated to the child’s level) based on responses
- Switched between Chinese and English to communicate in whatever language the child could understand
The parent group used the same e-book, but without the AI agent — parents themselves led the conversation during reading.
Study Design and Participants
- Participants: 67 children aged 5–8 in China (learning English as a foreign language)
- Experimental group (AI group): Read the e-book with the LLM-based conversational agent
- Control group (Parent group): Read the same e-book with a parent
- Random assignment: Children were randomly placed into groups
- Measures: Reading comprehension, vocabulary acquisition, story retelling (immediate and delayed tests)
- Engagement analysis: Video-recorded sessions coded for behavioral, affective, and cognitive engagement
Results: AI Wins on Comprehension, Parents Win on Emotion
Learning Outcomes
- The AI group scored significantly higher on reading comprehension than the parent group. This gap held not only on the immediate test but also on the delayed test administered later.
- For vocabulary acquisition and story retelling, there was no significant difference between groups. Both showed similar levels of improvement.
Engagement: This Is Where It Gets Interesting
- Children in the AI group showed higher behavioral engagement. They maintained longer eye contact with the screen and actively responded to the AI’s questions.
- Children in the parent group showed higher affective engagement. They laughed more, showed more surprise, expressed wonder, and produced more narrative-relevant vocalizations — spontaneous remarks connected to the story.
Put simply, children reading with AI “focused and solved problems,” while children reading with parents “enjoyed the story together.”
Why Did AI Outperform on Comprehension?
The researchers point to two mechanisms.
First, the AI conversational agent systematically prompted self-explanation, active recall, and meaning negotiation. Questions like “What happened in this scene?” and “Why do you think this character was sad?” were delivered consistently, with step-by-step hints based on the child’s responses. This structured interaction led to deeper story understanding.
Second, parents in EFL settings tended to focus heavily on teaching English vocabulary during reading. Interactions like “What’s this called in English?” and “Repeat this word” were frequent, while questions about the story itself were relatively scarce. This is a natural instinct — the whole point of reading an English picture book is to learn English. But it inadvertently undermined comprehension of the story’s meaning and flow.
The AI was optimized for “helping children understand the story.” Parents were focused on “teaching English words.”
But There’s Something Only Parents Can Give
The most noteworthy finding in this study may be what AI couldn’t do.
Children who read with parents showed richer emotional responses to the story. Laughter, surprise, exclamations of “Wow!” and spontaneous story-related remarks — “I feel bad for this bear,” “That happened to me too” — were all more frequent in the parent group.
This isn’t just about “having a nice time.” When a child responds emotionally to a story and connects it to their own experience, that represents a deeper layer of narrative understanding. This kind of emotional resonance also lays the foundation for reading habits and positive attitudes toward books.
AI asks good questions and delivers precise feedback, but it can’t create the moment when a child looks up, makes eye contact, and says “Mom, look at this!” with a smile. At least not yet.
Practical Takeaways
Use AI as a Tool, Not a Replacement
The message of this study isn’t “AI is better than parents.” AI showed strengths in systematic questioning and scaffolding, while parents showed strengths in emotional connection and story immersion. They don’t compete — they complement each other. AI can be a tool that boosts reading comprehension, but it cannot replace the experience of reading together with a parent.
For Parents in EFL Settings: Focus on the Story, Not Just Vocabulary
When reading English picture books, try asking “Why do you think this character did that?” instead of “What’s this word?” That simple shift can meaningfully improve your child’s reading comprehension. Your English doesn’t need to be perfect. You can ask story questions in your native language. What matters is thinking together about the flow and meaning of the story.
What to Look for in AI Reading Apps
Not all AI reading apps are the same. The AI that proved effective in this study didn’t simply read text aloud — it asked questions grounded in dialogic reading strategies (PEER/CROWD) and provided tailored feedback based on the child’s responses. If you’re considering an AI reading app for your child, look for whether it’s a passive narrator or an active conversational partner.
Limitations
- With only 67 participants, the sample size is small, limiting generalizability.
- The study focused on Chinese EFL children, so results may not directly transfer to Korean or other language contexts.
- Technical issues — speech recognition failures, content loading delays — caused difficulties for some children using the AI. These barriers may have affected both engagement and learning outcomes.
- Parents in the control group received no training in dialogic reading strategies. The AI was optimized for dialogic reading; the parents were not — making this an arguably uneven comparison.
- This was a short-term experiment; long-term effects remain unknown.
Final Thoughts
The message of this research comes down to this:
AI can develop a child’s mind, but developing their heart is still a parent’s job.
It’s encouraging that AI can boost comprehension through systematic questioning — especially in foreign language settings where parents struggle to manage alone. But the moments when a child laughs at a story, gasps in surprise, and says “That happened to me too” — those are created not by an AI on a screen, but by the person sitting beside them, turning the pages together.
The best reading experience is probably one where AI and parents each play to their strengths.
Source: Xiao, F., Zou, E. W., Lin, J., Li, Z., & Yang, D. (2025). Parent-led vs. AI-guided dialogic reading: Evidence from a randomized controlled trial in children’s e-book context. British Journal of Educational Technology, 56(5), 1784–1813. https://doi.org/10.1111/bjet.13615