Skip to the main content
PeabodyMake a free story

AI story generators for kids, compared by name: guardrails, chips and coaching

Eleven AI story makers for kids checked in October 2026, on input style, coaching, reading level and safety, plus what research says about kids making stories with AI.

We checked eleven AI story makers aimed at families on 27 October 2026. Nearly all of them are one-shot generators: a grown-up types an idea, and a finished story comes back. Only a couple coach the child, only one ties the story to the phonics a child is learning, and almost every safety claim is the company's own word. Research on children making stories with AI says the most useful AI asks questions instead of writing everything. That is also the lesson of our own worst test this week. Peabody (peabody.kids) is free for everyone through the end of 2026.

We make Peabody, so weigh our view with that in mind. An earlier post compared kinds of generator without naming products. This one names them, says where each claim comes from, and starts with our own mistake.

The test that went badly

Our founder typed a long, specific idea into Peabody's story maker: a dragon who is the principal of a school, who is embarrassed because the kids make fun of his purple jacket. He set the story to the first reading level, for a child who knows almost no words yet.

What came back was a pattern book about a cat and a hat. The dragon, the school, the jacket and the teasing were all gone.

Nothing crashed, and every reading-level rule was kept. The problem was in the instructions we gave the writer. For our earliest level, the prompt said to keep the story simple and use only two or three of the child's ideas. It came after a long list of sight-word, phonics and sentence rules, and the ideas came last. The cheapest model we could use put all its effort into the rules and dropped the idea.

The idea fits the level fine. Here is the same story inside the same limits (short sentences, almost all known words):

This is Mr. Fang. Mr. Fang is a dragon. He is the boss of the school. He has a purple jacket. The kids see the jacket. The kids laugh and laugh. Mr. Fang is sad. … A little kid is cold. Mr. Fang gives her the jacket. Now she is warm. Mr. Fang can smile. The end.

So the hard part isn't the reading level. It's making the child's idea the thing that can't be dropped, and checking that it wasn't. We're rebuilding the story maker around that (more at the end), and the comparison below is how we looked at what others do first.

How we compared them

For each product we looked at the website, the app store listing, the privacy policy and any independent reviews, and asked:

  • Input: a free-text box, guided chips (pick a character, a place, a theme), voice, or photos?
  • Coaching: does it ask the child anything, or does it just generate?
  • Reading level: none, an age picker, or tied to what the child is learning?
  • Safety: what does it say it does, and is anyone outside the company checking?
  • Who it's for: does a child use it, or a grown-up?

Products change quickly, and some details (prices especially) we couldn't confirm. Where something is only the company's claim, we say so.

The comparison

Checked 27 October 2026. Storybird and Khanmigo aren't AI story generators; they're in the table because they show two things the generators lack (human moderation, and a coach that won't do the work).
ProductWho typesInputAsks the child questions?Reading levelSafety, as stated
Gemini Storybook (Google)Grown-upFree text, photos, a child's drawingNo; you refine it in chatNoneGoogle's filters. Common Sense Media rated Gemini for under-13s high risk (Sept 2025)
Oscar StoriesChild or grown-upPick a character, job and interestsNoNone (says ages 6–8)"Reviewed by educators" (company claim)
Bedtimestory.aiGrown-upFree text, child as heroNoNone stated"Secure" (company claim)
ElloChildPreset settings, characters and plots (2024 version)Coaches reading aloudMatched to the week's phonicsReal-time filters, grown-up alerted (company claim)
Stories with Alexa (Amazon)Child, by voiceSpoken plot, characters, themeLets kids choose what happens nextNone statedKid profile inside a paid kids' plan; parents can review voice history
StoryBeeGrown-up or teacherPickers (age, genre) and free textNoAge picker"No inappropriate content" (no method given)
ReadKidzGrown-up creatorFree text and templatesNoThree age bands"Automated + editorial" review (company claim)
Childbook.aiGrown-upFree text, photosNoNoneNone found
StorybirdChildWrites over professional artNot AINot AIHuman moderators review every public story
Khanmigo Writing CoachStudentTheir own writingYes, and won't write it for themSchool levelSchool product, teacher view
ChatGPT / Claude used directlyGrown-upOpen chatOnly if askedNoneChatGPT is 13+ with parental controls; Claude's consumer terms are 18+

Checked 27 October 2026. Storybird and Khanmigo aren't AI story generators; they're in the table because they show two things the generators lack (human moderation, and a coach that won't do the work).

What we found

Almost every generator expects a grown-up at the keyboard. That's reasonable, since a free-text box connected to a large model is risky to hand a six-year-old.

Most are tools for parents, not for children

Almost every generator expects a grown-up at the keyboard. That's reasonable, since a free-text box connected to a large model is risky to hand a six-year-old. But it means the child is the audience, not the author. The few products a child drives alone (Ello, Alexa's stories, Oscar's pickers) keep the child's choices narrow on purpose.

Free text or chips is really a safety choice

Every product makes the same trade:

  • Free text lets a child ask for anything, including a dragon principal in a purple jacket. It also lets them ask for things no children's book should contain, so the output needs screening.
  • Chips and pickers (pick a hero, pick a place) can't go anywhere bad, but every story ends up built from the same few parts.

Ello's 2024 story feature was all presets, and its site now says it writes "original stories", so it may have moved toward free text. Character.AI went the other way in late 2025: after lawsuits, it moved teenagers off open chat onto guided "Stories".

The products that get this right let the child type or say what they want, then screen the input and the output. Chips are offered as help, not as a fence.

Almost nobody coaches

This was the biggest gap. Most tools take the idea and generate. Only two coach:

  • Khanmigo asks students about their writing and declines to write it for them, but it's a school writing tutor, not a story maker.
  • Ello coaches the child's reading aloud, not their storytelling.

We found no AI story maker that looks at a child's idea and says: great hero, but what goes wrong?

Reading level is usually an age picker

Where there's any control, it's usually an age menu that changes sentence length. Only Ello ties a story to the sounds a child is learning that week. An age doesn't tell you what a child can read, so a "5–7" story can be far too hard for one five-year-old and far too easy for another.

Safety is mostly a promise

The most concrete measures we found:

  • Storybird: human moderators read every public story, though it isn't an AI generator.
  • Ello: real-time filters with a grown-up alerted.

Everywhere else, safety is a sentence on the homepage. We found no independent tests of kids' story apps, only of general chatbots and AI toys (Common Sense Media rated three AI toys unacceptable in January 2026, with over a quarter of their replies inappropriate). We also found no documented case of a dedicated kids' story app producing an inappropriate story. That isn't the same as proof they're safe.

The rules are tightening:

  • An updated US children's privacy rule took full effect in April 2026, counting voice recordings as personal data and requiring a published data-retention policy.
  • California's companion-chatbot law has applied since January 2026.
  • The FTC opened a study of AI chatbots and children in September 2025.

What the research says about kids making stories with AI

The studies are small, and many are preprints or conference papers, so read these as directions, not verdicts.

The studies are small, and many are preprints or conference papers, so read these as directions, not verdicts.

An AI that asks questions can work like a good reading partner. In a randomized study of 117 children aged 3 to 6, a conversational agent that asked questions during a story did as well as a human partner on story talk and comprehension (Xu and colleagues, Child Development, 2022). A small 2024 study of ChatGPT reading with 5- and 6-year-olds found the children liked its questions and praise, but parents often had to rephrase when it misunderstood the child.

An AI that does the work can leave the learner worse off. In a field trial with about 1,000 high-school students, plain GPT-4 raised practice scores, then left students scoring 17% lower when the AI was taken away. A version built to give hints instead of answers didn't have that drop (Bastani and colleagues, PNAS, 2025). That was math with teenagers, not stories with young children, but it's the best evidence we have that an AI which asks beats an AI which tells.

AI help lifts weaker storytellers and can flatten stronger ones. With adult writers, AI ideas made stories more novel, most of all for the less creative writers. The AI-assisted stories were also more alike, and readers credited their authors with less ownership (Doshi and Hauser, Science Advances, 2024). A 2026 review of 22 child studies found possible benefits for creativity and confidence, and named the matching risks: children accepting the first thing the AI offers, falling into templates, and losing their own voice.

Choice motivates, in small doses. A meta-analysis of 41 studies found that having a choice raises motivation, more for children than adults, and works best with a few choices a session and three to five options each (Patall and colleagues, Psychological Bulletin, 2008). That's about two to four questions, each with three or four answers to tap.

Young children answer concrete questions best. Who, what and where come before why and how. Toddlers asked an either/or question out loud tend to pick the last option they heard, so for the youngest, answers should be pictures they tap, shown in a shuffled order.

Voice helps the youngest, but it mishears them. Typing and spelling hold back what young children can say in writing, and dictating gives longer stories. But speech recognition gets roughly a third of words wrong at ages 4 to 6. So a voice button should always show the child what it heard, with tap choices as the fallback.

Children understand far more than they can read. Listening comprehension runs ahead of reading comprehension well into elementary school. A beginner who can only read "Mr. Fang is sad" can still follow, and enjoy, a much richer story read aloud. Decodable text helps children practise sounding out, but there's little evidence it helps them understand. That's a strong reason to pair the few words a child reads with a fuller version they hear.

Waiting feels shorter when it's honest and busy. People prefer progress they can see. Pauses near the end make a wait feel longer, and seeing the work being done makes people value the result more. A loading screen that says "almost ready" in its first second gets this backwards.

What we're changing in Peabody

This is what we're building. None of it is live yet.

This is what we're building. None of it is live yet.

  1. The child's idea is a contract. Peabody will read the idea first and list what must be in the book: the hero, what's special about them, the place, the problem. The first three pages must show the hero and their main trait, and a check in code will send back any draft that drops them. Reading-level rules will decide how a thing is said, never whether it's in the story.
  2. A smarter writer. We'll move the story writer from the cheapest model to a stronger one. It costs about twice as much a story, which is still a few cents.
  3. A few questions before the preview. With a short idea like "dragons", Peabody will praise something specific and ask one to three simple questions: What's the dragon's name? What goes wrong? How does it end? Each comes with three tap answers, a microphone and a "You pick!" button. A rich idea like the dragon principal gets one question or none. A child can always skip to the story.
  4. Then choose. After the first pages, the child can make the whole book, add more, or change something.
  5. The child's own words are allowed. A beginner's book can use a few of the child's own words (dragon, jacket, purple), shown with a picture and read aloud first, so the story stays theirs and the words get learned.
  6. An honest wait. The loading screen will show the real steps and a time estimate based on how long stories actually take.

Until then, two things help on today's Peabody. For the youngest readers, put the most important thing first and keep it short ("a dragon principal with a purple jacket"). And for a story your child will hear rather than read, pick a higher level and read it aloud.

How to judge any AI story maker in five minutes

kids with no account, or read from the free library without making anything. For story starters to try with any generator, see AI story prompts for kids.
  1. Type something specific and a bit odd, like a dragon principal in a purple jacket. Is it on the first page?
  2. Ask who it's for. Can your child read the first page alone, or is it written for you to read aloud?
  3. See whether it asks anything back. A good partner asks what the hero wants or what goes wrong.
  4. Find the safety page. Is there a method, or just a promise? Who besides the company checks?
  5. Look at what it keeps. What happens to your child's typed or spoken ideas, and for how long?

You can try Peabody at peabody.kids with no account, or read from the free library without making anything. For story starters to try with any generator, see AI story prompts for kids.

Questions parents ask

It depends on who's reading:

Which AI story generator is best for kids?

It depends on who's reading:

  • A grown-up reading aloud: a flexible generator like Gemini Storybook makes a lovely picture book, as long as a grown-up does the typing.
  • A child learning to read: look for a tool that ties stories to what your child is learning.
  • A child learning to make stories: look for one that asks them questions rather than writing everything.

Do AI story generators have guardrails?

Most say they do. Few say how, and we found no independent testing of kids' story apps. The clearest measures are guided inputs, screening what goes in and what comes out, no open chat, and grown-up alerts. Treat any claim of perfect safety with suspicion.

Is it good for children to make stories with AI?

It can be, if the AI asks and the child decides. Studies suggest question-asking AI helps young children talk about and understand stories, while AI that writes everything can leave them with less of their own voice. Keep the child as the author.

Should young children type or talk to a story maker?

Talk, with a check. Under about seven, spelling gets in the way of ideas, so voice helps. But speech recognition often mishears young children, so the app should show what it heard and offer tap choices too.

Checked against how Peabody works on 27 October 2026. Written with help from AI. Peabody is a reading app, not a reading assessment: if you're worried about your child's reading, talk to their teacher. Questions, or something wrong? Write to hello@peabody.kids.

All posts →