Most children's audio guides are bad. Not in a subtle way, but in the way that a seven-year-old will tell you about directly: "This is boring, can we go?"
The standard industry solution has been to record a separate kids' track. Take the adult script, simplify the vocabulary, add a cheerful narrator, maybe throw in some sound effects. It costs tens of thousands of dollars, it sounds patronizing to anyone over eight, and it never gets updated because updates mean re-recording everything.
There's a better way to think about this. Instead of writing two scripts, you design a character and let AI deliver content through that character in a way that adapts to the child. The difference between these two approaches is the difference between a museum that families visit once and one they come back to.
Why the Separate Kids' Track Fails
The traditional model looks reasonable on paper. You hire a voice actor, write a simplified script, maybe create a mascot. The museum now has a "family-friendly" offering they can promote.
In practice, the problems stack up fast.
It's expensive to produce. A full children's audio guide (scripting, recording, editing, sound design) can cost as much as the adult version. For a museum with 40 stops, that's easily another $15,000-30,000. Most institutions can't justify that spend, which is why the majority of museums simply don't offer a children's version at all.
It's frozen in time. When the museum updates an exhibit, the adult script might get revised. The children's version almost never does. Within a year, the kids' guide is describing objects that have moved or referencing temporary installations that closed months ago.
It targets nobody well. A five-year-old and a twelve-year-old have radically different cognitive abilities, attention spans, and interests. A single "children's track" tries to serve both and satisfies neither. The five-year-old doesn't follow the narrative. The twelve-year-old feels talked down to.
It's monolingual. Producing a children's track in one language is already expensive. Producing it in six languages is nearly impossible for most budgets. So international families, often the visitors who need the most support, get nothing.
Character Design Over Script Writing
The shift that AI makes possible: instead of writing a script for children, you design a character that speaks to children.
This is a different kind of creative work. You're not choosing every word. You're defining a personality, a knowledge level, a tone, a set of rules about how to communicate. Then the AI generates appropriate content through that character for every stop, every question, every language.
Think of it as the difference between writing a screenplay and casting a role. The screenplay fixes every line. Casting defines who the person is, and they improvise within those boundaries.
A well-designed children's character might have rules like: use short sentences, max two new vocabulary words per stop, always connect abstract concepts to something physical the child can see, end each stop with a question. Those rules apply whether the child is looking at a Roman mosaic or a contemporary sculpture. The character stays consistent. The content stays relevant.
We've seen this work in practice. A character built for 6-9 year olds at a history museum doesn't need a separate script for every room. It needs clear instructions: "Explain things the way a friendly older sibling would. Use comparisons to everyday objects. Keep stops under 60 seconds. When discussing dates, relate them to something concrete — 'That's about as long ago as 20 grandma lifetimes.'"
The character does the rest.
Age Ranges That Actually Matter
Not all children are the same audience. That sounds obvious, but most museums treat "kids" as a single bucket. In our experience, three bands cover the realistic range:
Under 5. These children won't use an audio guide independently. They're with a parent who's trying to manage the visit. The best thing you can do for this group is make the adult guide better: shorter stops, awareness that the visitor might be distracted, and practical information like where the family restroom is. Designing a dedicated audio experience for toddlers is wasted effort.
Ages 5-8. This is where character-driven guides shine. Kids in this range respond strongly to personality. They'll listen to a character they like. They'll answer questions directed at them. They want stories, not facts. A pirate who knows about maritime history. An animal character who lives in the museum at night. The character is the hook that keeps them engaged.
Ages 9-12. Older kids can handle more information but still need it delivered differently than adults. They respond to challenge and discovery: "Can you find the hidden symbol in this painting?" They want to feel smart, not talked down to. The tone shifts from storytelling to something closer to a game or a mystery.
Teenagers, in most cases, do fine with the regular adult guide, especially if that guide is conversational and can answer questions. Forcing teens into a "youth" experience usually backfires.
Gamification vs. Storytelling
Museums often ask whether their children's guide should be gamified. Scavenger hunts, point systems, quizzes at each stop. The answer depends entirely on the collection and the museum's goals.
Pure storytelling works best for narrative-rich collections. A history museum walking visitors through a chronological journey. An art museum where each room builds on themes from the last. If the collection already tells a story, the children's guide should follow that story in a voice kids connect with.
Gamification works when the collection is more modular, when objects don't have a strong sequential relationship. A natural history museum where each room is a different ecosystem. A science museum where exhibits are largely independent. Here, giving kids a mission ("Find three animals that use camouflage") creates structure that the collection itself doesn't provide.
Hybrid approaches often work best. A character guides the child through a story, but at certain stops asks a question or poses a challenge. "Count the columns in this room. I bet you can't find them all." The game elements serve the narrative rather than replacing it.
With AI, you don't have to commit to one model at production time. The character's instructions can include both storytelling and game elements, weighted by the museum's preference. Adjusting the balance later takes minutes, not months of re-recording.
Guardrails for Young Visitors
When a child can ask any question and an AI will answer, content safety becomes a real concern. Museums are right to think carefully about this.
The good news: character design is itself a guardrail system. A well-designed children's character has explicit boundaries built into its instructions. It knows to keep language age-appropriate. It knows to redirect questions about violence or death in ways that are honest but not graphic. It knows that when a child asks about a war depicted in a painting, the answer should acknowledge what happened without dwelling on suffering.
These aren't afterthoughts bolted onto the system. They're part of the character definition from the start. And because they operate at the instruction level, they apply to every stop and every language automatically.
A few specific guardrails that matter:
- Vocabulary filtering. The character's instructions specify a reading level. The AI stays within it.
- Topic boundaries. Some subjects need special handling for children. A character serving 5-8 year olds at a site with sensitive historical content can be instructed to focus on daily life and objects rather than violence.
- Redirect behavior. When a child asks something outside the appropriate scope, the character doesn't refuse awkwardly. It redirects: "That's a big question! Let me tell you about something really cool over here instead."
- No external content. The AI answers from the museum's knowledge base only. It doesn't pull from the open internet, so there's no risk of surfacing inappropriate material.
Families Experiencing the Same Exhibit Differently
The most interesting use case is the family's experience together, not any one person's individual guide.
Picture a family of four at a painting. Dad has the standard guide and hears about the artist's technique and historical context. The ten-year-old has the explorer character and gets a challenge: "This painting has a secret. The artist hid something in the background. Can you spot it?" The seven-year-old has the storyteller character and hears a short tale about the people in the painting.
They're all standing at the same object. They can talk to each other about it. The child says "I found the hidden thing!" and the parent, who just heard the context, can engage with that discovery. The experience is shared even though the content is different.
This is nearly impossible with traditional audio guides. You'd need to script, record, and synchronize three separate tracks. With AI characters, each family member simply selects their preferred experience at the start of the visit. The system handles the rest.
Practical Design Tips
These patterns hold for any children's audio guide, AI-driven or traditional:
Keep stops short. 45-90 seconds for ages 5-8. 60-120 seconds for ages 9-12. Adults can tolerate 2-3 minute stops. Kids can't. If you're using AI, build this into the character instructions as a hard limit.
End every stop with engagement. A question, a challenge, something to look for. Passive listening is the fastest way to lose a child's attention. "Before we move on, look at the soldier's shield. What animal do you think that is?" gives the child a reason to keep their eyes up and their brain active.
Use the physical space. The best children's guides point at things. "See the big red painting on your left?" is better than "The next work is Composition in Red by Mondrian." Kids orient visually and spatially, not by artist name.
Let the character have personality flaws. Perfect characters are boring. A character that gets things slightly wrong and lets the child correct it ("Wait, that can't be right, crocodiles don't live in the mountains!") is far more engaging than one that delivers flawless information.
Test with actual children. This sounds obvious. Most museums skip it. Before launch, put the guide in the hands of three kids in your target age range and watch what happens. You'll learn more in 30 minutes than in a month of internal review.
Applying This to Your Museum
Every museum's approach to families will be different. A children's museum has different needs than an old masters gallery. A heritage site telling a single story operates differently than a natural history museum with hundreds of independent specimens.
The point isn't that there's one correct children's audio guide. The character-driven approach, defining a persona with clear rules about tone, complexity, length, and interactivity, gives you more flexibility than scripting a separate track. You can adjust the character's behavior in an afternoon. You can support multiple age ranges without multiplying production costs. You can offer the experience in every language your museum supports.
At Musa, this is built into how the platform works. Museums design characters through our studio, set the rules for how those characters communicate, and the AI museum guide handles delivery across languages and age groups. The character design is the creative work. The scaling is automatic.
If you're thinking about making your museum work better for families, let's talk about what that looks like for your collection.