AI for Good

How AI Is Preserving Endangered Languages and Cultures

April 7, 2026 · 5 min read

A language dies roughly every two weeks. That is not a dramatic exaggeration — linguists estimate that of the approximately 7,000 languages spoken on Earth today, nearly half are endangered. When the last fluent speaker of a language passes away, an entire way of understanding the world vanishes with them. Centuries of accumulated knowledge, stories, humor, and perspective — gone forever.

For most of human history, there was very little anyone could do about this. Documenting a language is painstaking work that takes years and requires trained linguists. But artificial intelligence is dramatically changing what is possible, and the race to preserve endangered languages is getting a powerful new ally.

Documenting Languages at Speed

Traditional language documentation is slow. A linguist travels to a community, records speakers, transcribes audio by hand, identifies grammatical patterns, builds dictionaries — a process that can take a decade or more for a single language. Given that thousands of languages are at risk, that pace is nowhere near fast enough.

AI speech recognition and natural language processing are accelerating this work dramatically. Modern AI models can listen to recordings of a language and begin identifying patterns — common sounds, word boundaries, recurring phrases — even without any prior knowledge of that language. This does not replace the linguist, but it compresses months of transcription work into days.

Google’s Project Relate and initiatives from organizations like the Endangered Languages Project use AI to help communities create digital records of their languages. Speakers can record themselves telling stories, explaining concepts, or simply having conversations, and AI tools help organize, transcribe, and index this material in ways that make it searchable and usable for future learners.

The key shift is that documentation no longer requires a PhD in linguistics. Community members themselves can use AI-assisted tools to record and preserve their language, putting the power in the hands of the people who know it best.

Translation for Languages With Almost No Speakers

Machine translation systems like Google Translate work well for major languages because they are trained on massive amounts of text — billions of sentences in English, Spanish, Chinese, and so on. But what about a language spoken by 500 people in a remote village, with no written text available online?

This is one of the hardest problems in AI, and researchers are making real progress. A technique called transfer learning allows AI models to leverage patterns learned from well-documented languages to bootstrap translation capabilities for related endangered languages. If you have a strong model for a major language family, it can be adapted to a related minority language with far less training data than starting from scratch.

Meta’s No Language Left Behind project has built translation models covering over 200 languages, including many with very limited digital resources. The models are not perfect, but they are functional — and for languages that previously had zero digital translation capability, even imperfect translation is transformative.

Some projects are taking a different approach entirely. AI systems that work with audio directly — translating speech to speech without needing a written form — are particularly valuable for languages that have no standard writing system. This is a much harder technical problem, but it sidesteps the requirement for text data that simply does not exist for many endangered languages.

Cultural Archiving Goes Digital

Language is just one piece of the puzzle. Culture is expressed through art, music, dance, rituals, craftsmanship, medicine, and countless other practices. Many of these traditions are oral — passed from generation to generation through stories and demonstrations, not written manuals.

AI is helping create comprehensive digital archives of cultural practices. Computer vision can analyze and catalog visual artifacts — textiles, pottery, tools, artwork — identifying patterns, styles, and techniques that help researchers understand and classify cultural traditions. Machine learning models can detect connections between artifacts from different communities, revealing cultural exchange and evolution that was previously invisible.

Music preservation is another area where AI shines. Ethnomusicologists use AI to analyze recordings of traditional music, identifying scales, rhythms, instruments, and vocal techniques. These analyses create detailed records that musicians can use to learn and perform traditional pieces even if the original practitioners are no longer alive to teach them.

Museums and cultural institutions are using AI to digitize and organize vast collections of cultural materials. Handwritten manuscripts can be transcribed automatically. Deteriorated recordings can be enhanced and restored. Photographs can be dated, geolocated, and connected to related materials in other collections. The result is a cultural heritage that is not just preserved but made accessible to anyone in the world.

Oral History: Capturing Voices Before They Are Silent

The most urgent preservation work involves elderly speakers who carry irreplaceable knowledge. When a 90-year-old elder is the last person who remembers a particular creation story, healing practice, or historical account, every day matters.

AI-assisted oral history projects are dramatically increasing the amount of knowledge that can be captured and preserved from individual speakers. Conversational AI systems can conduct structured interviews, asking follow-up questions and probing deeper into topics that surface during a conversation. These systems never get tired, never run out of recording media, and can be used by family members without special training.

After recording, AI tools automatically transcribe the interviews, identify key topics and themes, and cross-reference statements with existing knowledge. The result is not just a recording but a searchable, organized knowledge base that future generations can explore and learn from.

Projects like the First Voices platform for Indigenous communities in Canada and the Living Tongues Institute’s work across Asia and South America are using these approaches right now, with real results. Communities that were watching their linguistic heritage slip away now have tools to fight back.

Why This Matters Beyond Language

Preserving endangered languages is not just academic sentimentalism. Each language encodes unique knowledge — about local plants, weather patterns, ecological relationships, and problem-solving approaches that exist nowhere else. When a language dies, practical knowledge dies with it. Scientists have documented cases where indigenous language terms contain ecological insights that Western science did not discover independently until decades later.

AI cannot bring back a language that is already gone. But it can make the difference between a language that fades into silence and one that survives for future generations to learn, speak, and build upon. That is a use of technology worth celebrating.