The Research Behind Gently Told
How a dad who loves tech and worries about screen time ended up reading academic papers on attachment theory at 2am.
A letter from the founder
I've been playing with AI since GPT-3 dropped in 2022. Not as an investor or a founder with a pitch deck. Just as a geek who couldn't quite believe what he was seeing. I spent months feeding it prompts, pushing its limits, finding the edges where it fell apart. I was hooked.
I'm also a dad. And like most parents I know, I have complicated feelings about technology and kids. We limit screen time. We cringe at the YouTube rabbit holes. We know the research about dopamine and developing brains. But we also know our daughter lights up when she hears a story with her name in it, and that there's something genuinely magical about a bedtime tale that knows she had a tough day at kindy.
So when I looked at the AI story generators out there, I was pretty disappointed. Most of them are doing the bare minimum. Throw a name into a template, pick a theme from a dropdown, get 500 words of generic nothing. The stories sound the same. They read the same. They feel like they were assembled, not written. AI slop, basically.
I figured we could do better. But “better” needed to mean something specific, not just “more words” or “nicer pictures.” So I started reading. Properly reading. And that's how I ended up at 2am with tabs open for Shechtman's bibliotherapy research, Bowlby's attachment theory, and Cincinnati Children's Hospital fMRI studies, wondering how I got here.
What the research actually says
Three things stood out from the reading.
Kids have wildly different attention spans at different ages, and most children's content ignores this completely. A three-year-old can sustain focus for about 7 minutes on average. A seven-year-old can handle 17. That's a massive gap. A longitudinal study by McClelland et al. followed 430 children for 21 years and found that attention span-persistence at age 4 predicted a 49% higher likelihood of completing a degree by age 25. Getting the length right isn't cosmetic. It shapes whether the story actually lands.
Audio-only content is cognitively harder than books with pictures. Dr. John Hutton's fMRI work at Cincinnati Children's Hospital scanned 27 preschoolers listening to stories in three formats: audio-only, illustrated, and animated. Audio-only stories required the highest cognitive load, with children's language networks “straining” to process without visual scaffolding. That means for audio stories, we need to bemore conservative with length, not less.
Different kinds of stories serve different purposes, and the research behind each one is genuinely deep. Bibliotherapy (using stories therapeutically) has decades of clinical evidence. Attachment theory goes back to Bowlby in the 1960s. Inquiry-based learning, social-emotional frameworks, growth mindset. These aren't buzzwords. They're well-studied fields with real evidence behind them.
So we built a per-age duration model. Each age gets its own target, calculated from attention research. We aim for about 60% of the average sustained attention span for that age. Long enough to tell a proper story. Short enough to finish while they're still into it.
Per-Age Story Duration
| Age | Attention Span | Our Target (60%) | Story Duration |
|---|---|---|---|
| Age 3 | 7–15 min | 7 × 60% = 4 min | 650 words (~4 min) |
| Age 4 | 10–20 min | 10 × 60% = 6 min | 900 words (~6 min) |
| Age 5 | 12–25 min | 12 × 60% = 7 min | 1,100 words (~7 min) |
| Age 6 | 15–30 min | 15 × 60% = 9 min | 1,350 words (~9 min) |
| Age 7 | 17–35 min | 17 × 60% = 10 min | 1,500 words (~10 min) |
💜 Bonding stories are about 20% shorter than the standard target. Attachment research (Bowlby, Ainsworth) found that the ritual around a story matters more than the story itself. When you read to your kid at bedtime, the closeness, the voice, the conversation afterwards: that's where the bonding happens. A shorter story leaves more room for it.
Sources: Hutton et al. (2017, 2018); McClelland et al. (2013); Ruff & Capozzoli (2003); Mindell et al. (2015); Brain Balance / CNLD normative data
How we use the research
We sort our stories into 4 tiers based on what they're trying to do. Each tier follows a different research framework, and those frameworks shape how the AI writes the story. This isn't a dropdown label. The framework principles are embedded in every prompt, guiding structure, emotional arc, language choices, and pacing.
Story Tiers
Each tier follows a different research framework. The framework principles are embedded in the AI prompts that generate every story.
Challenge Tier
Bibliotherapy Framework
Therapeutic stories for life transitions like potty training, school anxiety, and new siblings.
Learning Tier
Pedagogical Frameworks
Educational adventures based on research-backed learning theories.
Bonding Tier
Attachment Theory
Stories that strengthen parent-child connection through shared moments. Bonding stories are shorter (~80% of the standard age target) because attachment research shows the ritual itself drives bonding — the conversation after the story is as important as the story itself.
Support Tier
Neurodiversity Frameworks
Evidence-based stories for neurodivergent children using social narrative approaches, narrative therapy, and sensory-aware storytelling.
What's under the hood
The stuff you don't see, but your kid hears.
Regional accents & spelling
Choose from American, British, Australian, or New Zealand narration. The narrator doesn't just change their voice — the AI writes the story in regional spelling too. “Colour” not “color.” “Mum” not “Mom.” Because stories should sound like home.
Four-layer safety filtering
Every story passes through topic restrictions, prompt injection guards, AI safety settings, and a final content review before your child hears a single word. We're paranoid about this, and we think you should expect that.
Audiobook-quality production
Stories get a personalised title jingle, mood-matched background music from a curated library, and narration with SSML-tuned pacing and expression. It's not a robot reading text. It sounds like a bedtime story should sound.
Real personalisation
Not just swapping a name into a template. We ask about your child's age, interests, favourite animals, what's happening in their life right now. The AI uses all of it to write a story that feels like it was made for them — because it was.
Why bother with any of this?
Because kids deserve better than content that was assembled from a template and shipped without anyone thinking about whether it actually works.
I'm not anti-AI. Obviously. I built an AI product. But I do think there's a real difference between using AI because it's cheap and easy, and using AI because you can do something with it that wasn't possible before. A personalised story that knows your kid's name, their favourite animal, the fact that they're starting school next week and feeling nervous about it? That wasn't possible for some random dad in Auckland to make three years ago. Now it is. But only if you put in the work to make it good.
“Good” means the story actually lands. It means a three-year-old gets a four-minute story that finishes before they lose focus, written in language that sounds like a real picture book and not a chatbot. It means a therapeutic story for a kid dealing with a new sibling follows the same five-step arc that clinical bibliotherapy uses. It means a bonding story is short enough that there's still time for a cuddle and a chat about it afterwards. Mindell's study of 10,085 families across 14 countries found that what mattered most was the consistency of the bedtime ritual, not its length. Shorter and reliable beats long and sporadic.
It's just me, my wife, and our toddler (who serves as head of quality assurance). We're still figuring things out. But the research is the bit I'm not willing to cut corners on. Every story that comes out of Gently Told has been shaped by actual evidence about what works for kids at different ages and stages.
If you're the kind of parent who reads the labels on everything, you're in the right place.
📖 What we read
The papers and books we actually used. Not a bibliography padded out to look impressive. These are the ones that shaped decisions.
Attention & Development
- • Hutton, J. S., et al. (2018). Differences in functional brain network connectivity during stories presented in audio, illustrated, and animated format. Brain Imaging and Behavior, 14(1), 130–141.
- • Hutton, J. S., et al. (2017). Shared reading quality and brain activation during story listening in preschool-age children. The Journal of Pediatrics, 191, 204–211.
- • McClelland, M. M., et al. (2013). Relations between preschool attention span-persistence and age 25 educational outcomes. Early Childhood Research Quarterly, 28(2), 314–324.
- • Mindell, J. A., et al. (2015). Bedtime routines for young children: A dose-dependent association with sleep outcomes. Sleep, 38(5), 717–722.
- • Ruff, H. A., & Capozzoli, M. C. (2003). Development of attention and distractibility in the first 4 years of life. Developmental Psychology, 39(5), 877–890.
- • Gaertner, B. M., et al. (2008). Child and maternal temperament and sustained attention in preschool. Merrill-Palmer Quarterly, 54(4), 514–542.
- • Brain Balance Centers; CNLD Testing & Therapy. Normative attention span guidelines by age (collated from developmental literature).
Bibliotherapy Framework
- • Shechtman, Z. (2009). Treating Child and Adolescent Aggression Through Bibliotherapy. Springer.
- • Pardeck, J. T. (1994). Using literature to help adolescents cope with problems. Adolescence.
- • Heath, M. A., et al. (2005). Bibliotherapy: A resource to facilitate emotional healing. School Psychology International.
Pedagogical Frameworks
- • Dweck, C. S. (2006). Mindset: The New Psychology of Success. Random House.
- • Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review.
- • Durlak, J. A., et al. (2011). The impact of enhancing students' social and emotional learning: A meta-analysis of school-based universal interventions. Child Development, 82(1), 405–432.
- • CASEL. (2020). CASEL's SEL Framework.
- • Bybee, R. W. (2006). The 5E Instructional Model. NSTA.
Attachment Theory
- • Bowlby, J. (1988). A Secure Base: Parent-Child Attachment and Healthy Human Development. Basic Books.
- • Ainsworth, M. D. S. (1978). Patterns of Attachment. Erlbaum.
- • Mindell, J. A., et al. (2015). Bedtime routines for young children: A dose-dependent association with sleep outcomes. Sleep, 38(5), 717–722.
- • Sroufe, L. A. (2005). Attachment and development: A prospective, longitudinal study from birth to adulthood. Attachment & Human Development, 7(4), 349–367.
- • Harter, S. (2012). The Construction of the Self. Guilford Press.
Neurodiversity Frameworks
- • Gray, C. (2015). The New Social Story Book (15th Anniversary Edition). Future Horizons.
- • White, M. & Epston, D. (1990). Narrative Means to Therapeutic Ends. Norton.
- • Dunn, W. (2007). Living Sensationally: Understanding Your Senses. Jessica Kingsley Publishers.