A zoo audio guide recorded in January will tell visitors about the outdoor penguin exhibit in July, even if the penguins moved indoors six months ago because of a heat management protocol. A botanical garden guide recorded in spring will describe the cherry blossoms to someone visiting in November, staring at bare branches.
This is the fundamental problem with audio guides for living collections. Museums have static objects. Zoos and gardens have organisms that grow, move, breed, go dormant, get sick, migrate between enclosures, and bloom on schedules that don't respect your production timeline. A recorded audio guide is a snapshot. Living collections don't hold still for snapshots.
We haven't worked with a zoo or botanical garden yet at Musa. But we've spent a lot of time thinking about this space, and the more we look at it, the more convinced we are that AI-generated audio guides solve problems here that no traditional approach can touch.
The problem with static guides in living spaces
Traditional audio guide production assumes your collection stays put. You write scripts, record them, produce them in multiple languages, and distribute them. The cost is justified because the content will be valid for years. A Rembrandt doesn't move. A Roman sculpture doesn't go dormant in winter.
Zoos and gardens break this assumption completely.
Animals rotate between enclosures based on weather, breeding programs, veterinary needs, and social dynamics. Plants bloom for two weeks and then don't again for a year. New specimens arrive. Others get transferred to partner institutions. Entire sections close for renovation or seasonal reasons. A butterfly house might be open March through October and shuttered the rest of the year.
A traditional guide can't account for any of this. The economics alone make it impractical: you'd need to re-record content every time something changes, which in a zoo means constantly. So most zoos either don't offer audio guides, or they offer guides that are vaguely correct at best and actively misleading at worst.
The few that try usually produce a generic overview: "In this area, you may see various species of big cats." That's not interpretation. That's a wall plaque read aloud.
What "adaptive living collections" actually means
The phrase sounds like marketing. In practice, it's simpler than it sounds.
An AI-generated audio guide doesn't play recordings. It generates speech in real time from data you provide. If you update the data, the guide updates what it says. No re-recording, no new production cycle, no script revisions.
For a zoo, this means: if the keeper team updates the system at 8 AM to note that the red pandas are in the outdoor habitat today, visitors who arrive at 10 AM hear "The red pandas are in the outdoor habitat today. Look for them in the trees near the back of the enclosure." If it rains and they're moved inside at noon, update the data and the afternoon visitors get the right information.
For a botanical garden: in April, the guide knows the magnolias are in full bloom and directs visitors toward them. In August, it talks about the summer perennials instead. In October, it focuses on autumn colour and seed collection. The guide doesn't need a "spring version" and an "autumn version." It draws from the current state of the collection.
This isn't theoretical. Musa's system already handles exactly this kind of changing content for museums with rotating exhibitions. The pattern is the same: ingest current data, let the AI generate from it. The difference with living collections is that the data changes more frequently, daily or even hourly rather than monthly. But the mechanism is identical.
If you have schedules (feeding times, keeper talks, show times), load them in. The AI will surface whatever is relevant whenever it's relevant. "The sea lion feeding starts in 20 minutes at the main pool" is the kind of thing a live tour guide would say. A recorded guide never could.
Outdoor environments and GPS navigation
Most audio guide systems were designed for indoor spaces. Rooms, galleries, defined paths. Zoos and botanical gardens are something else entirely: 50 to 200 acres of outdoor terrain with winding paths, seasonal closures, and multiple valid routes through the same space.
GPS-based navigation handles this. Visitors get wayfinding that works across the full site, not just inside buildings. The guide can suggest routes based on what's currently open, what's nearby, and what the visitor has already seen. If the African savanna section is closed for maintenance, the guide routes around it without the visitor needing to consult a paper map.
This matters more than it might seem. We've seen research on visitor behavior at large outdoor sites showing that navigation confusion is one of the top reasons people cut visits short. They can't find what they're looking for, they walk past things without realizing they're there, and they default to the most obvious path rather than the most rewarding one. A guide that knows the layout and the visitor's position solves this directly.
Indoor exhibits within zoos (reptile houses, aquariums, tropical houses) can use visual anchor points and floor plans the same way a museum would. The system handles the transition between outdoor GPS and indoor positioning without the visitor doing anything.
Weather and the things you can't control
A family arrives at the zoo on a rainy Tuesday. Half the animals are sheltering. The outdoor bird show is cancelled. The playground is closed.
A recorded guide doesn't know any of this. It cheerfully sends visitors to the flamingo lake, where the flamingos are huddled under a shelter and barely visible.
An AI guide, drawing from current data and conditions, handles this differently. It can acknowledge reality: "The outdoor birds are sheltering today because of the rain, but the tropical house is a great stop right now. It's warm, dry, and the toucans are very active in this weather." That's what a good human guide would do. Adapt.
The same applies to seasonal closures, construction, special events. Anything that changes the experience can be reflected in what the guide says, as long as the information is in the system. And feeding that information in is a data entry task, not a production project.
Conservation messaging that fits, not lectures
Zoos have changed. The modern zoo's primary mission is conservation, and most want their visitor experience to reflect that. But weaving conservation education into an audio guide is hard when every word is scripted. Do you add a conservation paragraph to every stop? That gets preachy fast. Do you create a separate conservation tour? Most visitors won't choose it.
AI-generated guides handle this more naturally. Conservation messaging can be part of the guide's core instructions, a priority built into how it talks about every animal rather than a separate track. When a visitor asks about the snow leopards, the guide talks about the animal and, where it makes sense, mentions that there are fewer than 7,000 left in the wild and that the zoo participates in a breeding program. It doesn't lecture. It includes the information as part of a fuller picture, the way a knowledgeable keeper would.
The guide can also adjust depth based on the visitor's engagement. Someone who asks follow-up questions about the breeding program gets more detail. Someone who just wants to know where to see the tigers gets directions. The conservation content is available without being forced on people who aren't interested. That's better than either ignoring conservation entirely or making every stop feel like a PSA.
Families by default
Zoos are family destinations. This isn't one segment among many; it's the primary audience. Any audio guide that doesn't work for families is failing at the most basic level.
The challenge: a five-year-old and a forty-year-old standing at the same elephant enclosure want very different things. The five-year-old wants to know the elephant's name and whether it likes bananas. The forty-year-old might want to know about the social structure of the herd and the zoo's role in elephant conservation.
With AI character design, you don't produce two separate guides. You design personas. A playful character for young children that tells stories and asks questions: "Can you find the baby elephant? She's the smallest one. Her name is Suki and she was born right here last year!" A more informative persona for adults that goes deeper into ecology and conservation.
Each family member picks their experience. They're all at the same enclosure, talking to each other about what they're seeing, but the guide meets each of them where they are. Kids don't hear watered-down adult content. Adults don't sit through content aimed at six-year-olds.
Musa already does this for museums. A zoo is arguably the best use case for it, because the audience skew toward families is so strong.
The seasonal garden guide
Botanical gardens have a version of this problem that's uniquely their own. The entire point of a garden is that it changes. A garden in May and a garden in September are almost different institutions. The star attractions shift. The routes worth walking shift. The stories worth telling shift.
A traditional guide either ignores this, talking about "the garden" in generic terms, or requires multiple recorded versions tied to seasons. Both options are bad. The generic guide misses the point. The seasonal recordings are expensive and still too coarse: what's blooming in the first week of May and the last week of May aren't the same.
An AI guide that draws from current horticultural data gives visitors something no garden has offered before: interpretation that matches exactly what they're seeing. Not "in spring, you might see..." but "the tree peonies in this bed opened three days ago. This is their peak week." That specificity turns a pleasant walk into something informative.
Gardens already track this data internally. Horticulture teams maintain records of what's blooming, what's been planted, what's about to peak. Feeding that data into a guide system isn't extra work. It's making existing knowledge accessible to visitors for the first time.
What it takes to get started
We'll be direct: we haven't deployed at a zoo or botanical garden yet. What we have is a system built for content that changes, tested across museums with rotating exhibitions, seasonal programming, and constantly updating collections.
The architecture maps cleanly onto living collections. Data ingestion, real-time generation, GPS navigation, multilingual support, family-friendly character design: all of this exists today. The adaptation for zoos and gardens is about data sources (keeper reports, horticultural records, animal schedules) rather than new technology.
If you run a zoo or garden and you've given up on audio guides because the content goes stale too fast, or because the production costs don't justify the result, or because you've never found a guide that could handle the fact that your collection is alive, this is worth a conversation.
We think this is one of the strongest use cases for AI-generated audio guides, precisely because the limitations of traditional guides are most obvious here. A painting doesn't move. An elephant does. The guide should know the difference.
If you're interested in what this could look like for your site, we can talk it through.