Augmented storytelling with AI: how to tell a story without betraying verified facts?

A house passing on its craft for several generations, a science museum explaining a complex phenomenon, a region rich in legends and monuments, a company proud of its history.

All of them have a story to tell. Yet it often lies dormant in archives, booklets or panels that few visitors read to the end.

Generative artificial intelligence changes the game: it can tell, rephrase and answer everyone's questions. But beware, it can also invent a date, an anecdote or a quote.

How can you make the most of the storytelling power of AI without damaging the credibility of a brand or a body of knowledge?

In reality, the success of augmented storytelling depends less on technology than on the quality of the sources it relies on.

What is augmented storytelling?

Augmented storytelling is a story told by an AI from a controlled body of content, which adapts to the question, profile and pace of each visitor.

It differs from an audio guide or a video, which deliver the same content to everyone. It also differs from a general-purpose chatbot, which answers based on what it learned during training, with no guarantee of accuracy.

Visitors no longer follow a fixed path. They ask their own questions, to a character, an artwork, a brand or even the sea.

Which institutions already tell their story with AI?

Several leading cultural institutions have paved the way.

• The Dalí Museum, in Florida, brought Salvador Dalí back to life on screen as early as 2019 with Dalí Lives. Since 2024, with Ask Dalí, visitors pick up a replica of the famous "Lobster Telephone" to ask the artist their questions, in a voice recreated from archival interviews (The Dalí Museum).
• The Metropolitan Museum of Art, in New York, worked with OpenAI to create a conversation with Natalie Potter, a New York socialite whose wedding dress, worn in 1930, was shown in the Sleeping Beauties exhibition in 2024. The character draws on letters, newspaper articles and historical documents selected by the museum team (OpenAI).
• The Pinacoteca de São Paulo launched The Voice of Art in 2017 with IBM and the Ogilvy agency: seven artworks answered visitors, after six months of work with curators and specialists to feed the AI with books, biographies and press archives (Ogilvy).

What these three projects have in common: a body of content built and validated by experts, long before the choice of AI model.

Why can AI betray a story?

A language model produces the most plausible sequence of words. It does not spontaneously tell a verified fact from a credible invention.

Left to itself, it can make several types of errors:

• A hallucination: an invented date, name or anecdote, stated with confidence.
• An anachronism: a historical figure talking about events after their time.
• An off-key tone: a restrained brand that becomes overly casual, a scientist who simplifies to the point of error.
• An off-topic answer: a visitor steers the conversation away and gets an answer unrelated to the place.

This phenomenon lies at the heart of why AI answers vary. For a cultural institution or a brand, every error directly affects its credibility.

How do you keep the story faithful to verified facts?

Reliable augmented storytelling rests on four pillars.

1. A validated knowledge base: archives, scientific publications, interpretation content and brand history, reviewed by curators, historians, scientists or the teams concerned. With RAG (Retrieval-Augmented Generation), the AI answers from these sources rather than from its general knowledge alone.
2. A well-defined character: its tone, its vocabulary, what it knows, what it does not know and what it refuses to invent. An honest character can say it does not know the answer.
3. A consistent embodied voice: cloned or synthetic, the voice is part of the story's identity and must stay true to the world of the institution or the brand.
4. Testing and supervision: trick questions before launch, then regular review of conversations to enrich the knowledge base and correct any drift.

What the Oceans experience shows

In June 2026, for the "Monsters of the Oceans" event at the SEA LIFE Paris Val d'Europe aquarium, Gensai designed a voice installation allowing visitors to question the sea about its coexistence with humans.

The sea answered from a validated knowledge base, in a deep, mysterious voice created through voice cloning with technology from the French company Gradium. The whole system ran on a simple Mac Mini, with no Internet connection, over nine days of operation.

The project is described in detail in the article How to make the oceans speak with frugal AI?.

This architecture is not limited to aquariums. Gensai has the technical stack behind its Augmented Guide product: a sovereign knowledge base, real-time voice, editorial guardrails and multiple interfaces, from kiosks to talking objects, the web or WhatsApp. It can be adapted to a brand, a museum, a science centre or a region. The uses observed at SITEM, at the Louvre, also show the growing interest of cultural venues in these experiences.

Augmented storytelling extends the voice of experts

Augmented storytelling replaces neither the historian, nor the guide, nor the brand.

It extends their voice, makes it accessible at any time, in several languages, at each visitor's pace.

Its value lies entirely in its faithfulness to the sources that feed it.

When it comes to artificial intelligence, technical sophistication never replaces clarity of message.

A brand story, scientific knowledge or heritage to bring into conversation with the public? Gensai supports the creation of the content base, the character and the experience, from the first idea to launch. Talk to the Gensai team.